University of Ghana http://ugspace.ug.edu.gh
REGIONAL INSTITUTE FOR POPULATION STUDIES
UNIVERSITY OF GHANA
LEGON
GENDER AND PERCEPTIONS OF PERSONAL SECURITY
IN GHANA
BY: CLEMENT SEFA-NYARKO
(10319459)
THIS DISSERTATION IS SUBMITTED TO THE UNIVERSITY OF GHANA,
LEGON, IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE
AWARD OF MA POPULATION STUDIES DEGREE.
SEPTEMBER 2016
University of Ghana http://ugspace.ug.edu.gh
Acceptance
Accepted by the College of Social Sciences, University of Ghana, Legon, in partial
fulfilment of the requirements of the degree of Master of Arts in Population Studies
Supervisor:
……………………………………………………………
Dr. Naa Dodua Dodoo
Date: …………………………………………………......
ii
University of Ghana http://ugspace.ug.edu.gh
Declaration
I, Clement Sefa-Nyarko, hereby declare that except for references to other works, which
has been duly acknowledged, this dissertation is the results of my own research and that
it has neither in part nor in whole been presented elsewhere for another degree.
Signed: …………………………………………………………………………
Clement Sefa-Nyarko
iii
University of Ghana http://ugspace.ug.edu.gh
Table of Contents
Contents
Acceptance .....................................................................................................................ii
Declaration ................................................................................................................... iii
Table of Contents .......................................................................................................... iv
List of Tables ................................................................................................................ ix
List of Figures ................................................................................................................ x
Abstract ......................................................................................................................... xi
CHAPTER ONE ............................................................................................................ 1
INTRODUCTION ......................................................................................................... 1
1.1 Background to the Study...................................................................................... 1
1.2 Statement of the Problem ..................................................................................... 3
1.3 Research Questions .............................................................................................. 7
1.4 Objectives ............................................................................................................ 7
1.5 Rationale for the study ......................................................................................... 7
1.6 Clarification of Terms .......................................................................................... 9
1.7 Organization of the Study .................................................................................. 10
CHAPTER TWO ......................................................................................................... 12
LITERATURE REVIEW ............................................................................................ 12
2.1 Personal Security Defined.................................................................................. 12
2.2 Theories and Contexts of Gender, Violence and Personal Security .................. 14
iv
University of Ghana http://ugspace.ug.edu.gh
2.2.1 Patriarchy, Masculinity, Violence and Personal Security ........................... 14
2.2.2 Personal Insecurity of Women .................................................................... 17
2.2.3 Personal Insecurity of Men ......................................................................... 21
2.3 Conceptualizing Gender and Perception of Personal Security: Fear of Crime and
Social Disorder......................................................................................................... 24
2.3.1 Other Determinants of Perceptions of Personal Security ........................... 27
2.4 Perceptions of Personal Security and Complexities in Measurement ............... 29
2.5 Conceptual Framework ...................................................................................... 30
2.5.1 Dependent Variable: Perception of Personal Security ................................ 32
2.5.2 Independent Variables ................................................................................ 32
2.5.3 Intermediate Variables 1: Fear Facilitating Factors .................................... 33
2.5.4 Intermediate Variables 2: Fear Inhibitors ................................................... 34
2.6 Hypotheses ......................................................................................................... 34
CHAPTER THREE ..................................................................................................... 35
METHODOLOGY ...................................................................................................... 35
3.1 About the GLSS 6 Data ..................................................................................... 35
3.2 Methods of Analysis .......................................................................................... 37
3.2.1 Dependent Variable .................................................................................... 37
3.2.2 Intermediate Variables ................................................................................ 38
3.2.3 Intermediate Variables 1: Fear Facilitators ................................................. 39
3.2.4 Intermediate Variables 2: Fear Inhibitors ................................................... 40
3.2.5 Independent Variables ................................................................................ 43
v
University of Ghana http://ugspace.ug.edu.gh
3.2.6 Independent Variable: Gender .................................................................... 43
3.2.7 Control Variables: Other Socio-Demographic Variables ........................... 43
3.3 Methods of Analysis .......................................................................................... 45
3.4 Data Limitations................................................................................................. 48
CHAPTER FOUR ........................................................................................................ 49
UNIVARIATE ANALYSES ....................................................................................... 49
4.1 About Ghana (Profile of Ghana) ........................................................................ 49
4.2 Socio-Demographic Characteristics of Respondents ......................................... 50
4.3 Fear Facilitators ................................................................................................. 53
4.4 Fear Inhibitors .................................................................................................... 54
4.5 Perception of Personal Security ......................................................................... 55
CHAPTER FIVE ......................................................................................................... 58
BIVARIATE ANALYSES .......................................................................................... 58
5.1 Introduction ........................................................................................................ 58
5.2 Gender, Fear facilitators, Fear Inhibitors & Perception of Personal Security ... 58
5.2.1 Gender and Perception of Personal Security .............................................. 58
5.2.2 Gender and the Fear Facilitators (Perception of Social Disorder) .............. 60
5.2.3 Gender and the Fear Inhibitors (Those Variables that are Continuous) ..... 61
5.2.4 Gender and Trust in Neighbours ................................................................. 61
5.2.5 Gender and Confidence in Public Security Services .................................. 65
5.2.6 Gender and the Other Non-Continuous Fear Inhibitors .............................. 66
vi
University of Ghana http://ugspace.ug.edu.gh
5.3 The Control Variables and the Perception of Personal Security........................ 68
5.4 Intermediate Variables and Perception of Personal Security ............................. 72
5.4.1 The Fear Facilitators and Perception of Personal Security ......................... 72
5.4.2 The Fear Inhibitors ...................................................................................... 74
CHAPTER SIX ............................................................................................................ 78
MULTIVARIATE ANALYSES ................................................................................. 78
6.1 Introduction ........................................................................................................ 78
6.1.1 Regression Method Used ............................................................................ 79
6.2 Meeting Assumptions of Multiple Regression Analysis ................................... 80
6.3 Regression Diagnostics ...................................................................................... 84
6.4 The Results and Checking of Hypotheses ......................................................... 87
6.4.1 Gender ......................................................................................................... 87
6.4.2 The Fear Facilitators (Social Disorder Factors) .......................................... 87
6.4.3 Control Variables ........................................................................................ 91
6.4.4 Fear Inhibitors ............................................................................................. 92
6.5 Discussion .......................................................................................................... 93
CHAPTER SEVEN ..................................................................................................... 97
SUMMARY, CONCLUSIONS AND RECOMMENDATIONS................................ 97
7.1 Summary ............................................................................................................ 97
7.2 Conclusions ........................................................................................................ 98
7.3 Recommendations .............................................................................................. 99
vii
University of Ghana http://ugspace.ug.edu.gh
APPENDIX A ............................................................................................................ 102
Additional Tables ................................................................................................... 102
APPENDIX B ............................................................................................................ 105
Additional Charts and Graphs ................................................................................ 105
APPENDIX C ............................................................................................................ 109
Summary of Variables Measurements ................................................................... 109
REFERENCES .......................................................................................................... 112
viii
University of Ghana http://ugspace.ug.edu.gh
List of Tables
Table 2.1 Instances of Violence against Women in their Life Cycle
Table 4. 1 Demographic and Other Socio-Geographic Characteristics
Table 4.2: Descriptive Statistics of Respondents’ Accounts of Fear Facilitators
Table 4.3: Descriptive Statistics of Fear-Inhibitor Scales
Table 4.4: Categorical (Dichotomous) Fear Inhibitors
Table 4.5: Summary Statistics for the Perception of Personal Security Scale
Table 5.1: Mann-Whitney U-Test Ranks Table for Gender and Perception of Personal Security
Table 5.2: Test Statistics for Mann-Whitney U Test
Table 5.3: Gender and the Occurrence (or Perception) of Fear Facilitators among Respondents
Table 5.4: Pearson Chi-Square Tests for Gender and the Fear Facilitators
Table 5.5: Tests for Normality of Some Fear Inhibitors and Gender
Table 5.6: Distribution of Initial Perception, Confidence and Trust Scales by Gender
Table 5.7: Independent Samples t-Test for Gender and Trust in Neighbours
Table 5.8: Cross Tabulations for Gender and each of the Dichotomous Fear Inhibitors
Table 5.9: Chi-Square Tests for Gender and each of the Dichotomous Fear Inhibitors
Table 5.10 Descriptive Statistics of Perception of Personal Security and Dichotomous Control
Variables
Table 5.11: Spearman’s Rho Correlation Matrix
Table 5.12: ANOVA Output for Religion and Perception of Personal Security
Table 5.13: Descriptive Statistics for Fear Facilitators and Perception of Personal Security
Table 5.14: Table for Some Fear Inhibitors (Dichotomous) and Perception of Personal Security
Table 6.1: Model Summary of the Regression Models
Table 6.2: ANOVA Test for the Model
Table 6.3: Hierarchical Regression Models Predicting Perception of Personal Security
Table 6.3 (Cont.): Hierarchical Regression Models Predicting Perception of Personal Security
Table 1A: Item-Total Statistics for the Variables in the Perception of Personal Security Scale
Table 2A: Item-Total Statistics for Variables in the Public Security Service Trust Scale
Table 3A: Item-Total Statistics for Variables in the Trust for Neighbours Scale
Table 4A: Descriptive Statistics for Test of Association between Perception of Personal
Security and Region of Residence, Religion and Ethnicity
Table 5A: Multiple Comparisons Table (Post-Hoc Test) for Region of Residence and
Perception Scale
Note: Tables 1A to 5A can be found in the Appendix A
ix
University of Ghana http://ugspace.ug.edu.gh
List of Figures
Figure 2.1: Conceptual Framework
Figure 4.1: Histogram Representation of the Perception of Personal Security of Respondents.
Figure 4.2: Modified Box Plot for Perception Scale
Figure 5.1: Box plot Distribution for Gender and Trust in Neighbours
Figure 5.2: Line Graph of Perception of Personal Security and Confidence in Public Security
Services and Trust in Neighbours.
Figure 1B: Histogram Representation of the Confidence in Public Security Scale
Figure 2B: A Histogram Representing Trust in Neighbours and Community Members
Figure 3B: Modified Box Plot for Confidence in Public Security Services by Gender
Figure 4B: A Box Plot for Trust in Neighbours Scale
Figure 5B: Box Plot for Gender and Perception of Personal Security
Figure 6B: A Scatter Plot of Standardized Residuals (*ZRESID) and Standardized Predicted
Value (*ZPRED)
Figure 7B: Normal Probability Plot of the Standardized Residuals
Figure 8B: Normal PP-Plot of Standardized Residuals
Note: Figures 1B to 8B are all found in Appendix B.
x
University of Ghana http://ugspace.ug.edu.gh
Abstract
Peace and security are sine qua non for development; and they are intrinsically linked
to demographic change. People are likely to move out en masses where there is
extremely low personal security, real or perceived. Even in non-conflict societies,
personal security is especially compromised through crime, persecution, domestic
violence and public disorder. Perceptions of personal security determine people’s
participation in social, economic and political activities, and have an impact on their
physiological and psychological wellbeing. Ghana is a typical politically stable country
that had a relatively lower score on personal safety in the 2013 Ibrahim Index of African
Governance, compared with its own score on other indicators of governance. None of
the explanations offered for Ghana’s lower score on personal safety has considered
gender as an important demographic indicator. Men and women perceive crime,
violence and threats differently based on their socialisation, which makes it imperative
that any discourse on personal security takes into account gender dynamics in society.
The system of patriarchy which is endemic in almost every society, coupled with
cultures of masculinity and femininity, play significant roles in the way men and
women perceive their personal security. This study uses a four-item scale from the sixth
edition of the Ghana Living Standards Survey to provide gender-sensitive explanations
for gender differences in the perception of personal security in Ghana. The results are
consistent with existing literature, especially when fear facilitators and fear inhibitors
are introduced into the analysis. Residential arrangements and type of place of
residence, among others, were found to predict perceptions of personal security
significantly. Fear inhibiting factors, like reliable security services and responsive
governance, were found to provide positive associations with perceptions of personal
security, with women responding more positively. The thesis argues that focusing on
improving fear inhibitors can significantly enhance perceptions of personal security of
people, and can reduce gender inequality gaps.
xi
University of Ghana http://ugspace.ug.edu.gh
CHAPTER ONE
INTRODUCTION
1.1 Background to the Study
The seminal Human Development Report of 1994 confirmed that Peace and security
are sine qua non for development (UNDP, 1994). Subsequent studies have agreed with
this, and established that peace and security are not only compromised in the context of
conflict, but also in non-conflict societies (Moser and McIlwaine, 2004; Paris, 2001 and
2004; Ayres, 1998; Alkire, 2003), where threats to human security abound. Personal
security is an aspect of human security, and is especially compromised by crime and
other forms of violence in many non-conflict developing countries. It is defined as ‘the
safety of human beings ... and the ways in which they are enabled to live a full and
productive life with dignity, without feeling afraid or at risk’ (Muggah, 2012: para 5).
In many developing countries, gender is an important dimension of the discourse on
personal security. Gender advocacy approaches have varied over the years, starting with
Women in Development (WID) and Gender and Development (GAD) approaches in
the 1980s that emphasized sexual and reproductive rights of women. Through
discussions at the 1994 Cairo International Conference on Population and Development
(ICPD), the 1995 Fourth World Conference on Women in Beijing, and the 2000 World
AIDS Day campaign programmes; it became apparent that ‘Men make a difference’1
in the discourse on gender equality and personal security. This was a milestone in the
discourse on gender, a shift from women-centred approach to gender (Dodoo and Frost,
2008; Dodoo, 1998; Gill, n.d.).
Personal security, like education, reproductive and health rights, is linked to collective
social goods since its absence restricts the freedom of both men and women to access
basic social and economic needs, thereby affecting quality and quantity of life. This
study hopes to explore men’s and women’s’ perceptions of security at home and in their
neighbourhood. Such gender sensitive analysis is useful in capturing the security
dimensions of inequality and social injustice, and ensuring that threats that afflict
1
The 2000 World AIDS Day Campaign had the theme: ‘Men make a difference’.
1
University of Ghana http://ugspace.ug.edu.gh
individuals and communities are not overlooked (Blanchard, 2003; Caprioli, 2005;
Harcourt, 2009).
Ghana is a politically stable country but has deficit in personal security. The 2013
Ibrahim Index of African Governance ranked Ghana as the 7th on the continent with
overall governance score of 66.8 percent. The country was ranked 6th on Safety and
Rule of Law, scoring 70.8 percent, much higher than the West African average of 54.9
percent. However, Ghana was ranked 16th on Personal Safety, scoring a low of 52.5
percent (IIAG, 2013).2 When people feel unsafe in their homes or neighbourhood, they
also feel insecure to engage in productive active activities. People’s ability to move
about freely, access education, access healthcare, and engage in productive economic
activities depend greatly on their perception and experience of personal security.
Individuals assess their personal security by assessing their immediate personal safety
conditions. They ask themselves whether they are able to stay home alone, walk in their
neighbourhood alone at night, or walk alone to school, workplace or market places
without compromising their safety.
In patriarchal societies, unequal gender roles, rampant inappropriate sexual behaviours,
and gendered socializations have impact on the way men and women assess their
security (Kameri-Mbote, 2004). Men and women experience and perpetuate violence
differently, whether in the domestic or public space (UNODC, 2011; UNDP, 2012;
Grossman and Owren, 2008; Bradshaw 2004). Socially constructed differences must
therefore be critically assessed in the discourse on security, especially in sub-Saharan
Africa, in order not to mask differentiated perceptions and experiences of (in)security
by men and women. Such differentiation ought to be done with caution in order not to
accept gender stereotypes and injustices as indubitable norms. Gender roles, norms and
stereotypes abound in Ghana (Gender Policy of the YMCA, 2011). Women’s work load
in domestic chores is approximately 25% greater than men (World Bank Gender
2
This is notwithstanding the fact that the 2012 Afrobarometer reported that 84 percent of Ghanaians felt safe at
home and 81 percent felt safe in their neighbourhood. The discrepancy between the Afrobarometer and the IIAG
report could be because data for the former was collected in 2011, which was one year prior to the general election
in Ghana; whilst data for the latter was collected in 2012, which was an election year with its attendant potential of
post-election violence.
2
University of Ghana http://ugspace.ug.edu.gh
Review 2002). Whilst women spend about two and half hours a day on unpaid chores
at home, men spend about 40 minutes a day (GSS, 2009). This study hopes to factor
these into the analysis in subsequent chapters.
This dissertation problematises (not glorifies) masculinity, and hopes to measure
gendered perceptions of personal insecurity in Ghana after controlling for other
variables like place of residence, region of residence and age. The theory of patriarchy
understood within the context of socially constructed hierarchy of masculinity and
femininity would be interrogated to aid the analysis. The power relations created
between men and women socializes the former to be aggressive and confident in public,
whilst the latter is said to be receptive and submissive to the protection of men. Men
are socialized into masculinities, and are expected to be ‘aggressive, active, life-taking
and violent’ (Harders, 2011: 141); whilst women are socialized into femininities, and
are expected to be ‘life-giving, passive, peaceful and caring’ (ibid). A gendered
relationship is created as part of socialization. This affects people’s daily experiences
of oppression and fear, impacting on their perceptions of personal security. The
dissertation recognizes that perception of (in)security does not equate to real experience
of (in)security. However, it works on the assumption that when people feel insecure,
the quality of their lives and productivity are affected.
The study would therefore interrogate the differences in perception of security using
datasets from the 6th round of the Ghana Living Standards Survey, which had a module
on Governance, Peace and Security. Explanatory variables would be used to understand
underlying causes and consequences of the perceptions of personal insecurity. The rest
of this chapter provides the statement of the problem, objectives of the study,
hypotheses, theory, and rationale for the study.
1.2 Statement of the Problem
In the past two decades, Ghana has emerged as an example of political stability and
good governance in Sub-Saharan Africa (Throup, 2011: 1). According to the 2013 Mo
Ibrahim Index on African Governance, Ghana is doing well on most indicators of
governance including rule of law, national security, participation, and human rights.
The country has a robust political system that thrives on democracy. Ghana’s high
overall score (66.8 percent) and ranking (7th) on governance, and higher score (70.8
3
University of Ghana http://ugspace.ug.edu.gh
percent) and ranking (6th) on Safety and Rule of Law are not consistent with its low
score on personal safety (16th with 52.5 percent) (IIAG, 2013). This is indicative of
structural challenges which affect the daily realities of people. The following are
reasons offered by some scholars to explain the low sense of personal security in Ghana.
First, the fierce political rivalry that has become part of Ghana’s democratic
dispensation sometimes turns violent during electioneering campaigns (Throup, 2011),
affecting people’s personal security. Second, socio-economic factors afflicting the
country in the last few years increase crime and other anti-social activities, especially
in urban areas (Opoku, 2012), thereby compounding safety dilemmas of individuals in
their homes and public spaces. This nexus is what Odumosu (1999: 71) referred to as
the ‘social cost of poverty’ in his article on the rate of crime in Nigeria. Third, an
emerging sector of the economy, commercial oil extraction, has the potential of
increasing trends of personal insecurity, considering trends in Chad, Nigeria and
Angola. The main report of the sixth round of the Ghana Living Standards Survey has
indicated that the oil extraction region of Ghana (Western Region) was the most unsafe
region between 2012 and 2013 due to higher rates of crime, attacks and insecurity (GSS,
2013: 177).
Fourth, ethnic and religious rivalries, which have claimed many lives in parts of the
country, especially the northern region (Mahama, 2009; Odotei and Awedoba, 2006: 5)
compound the problem of personal insecurity, due to occurrences of clashes and
violence. Fifth, bulging youth unemployment that breeds grievances, discontent, and
competition over few available job opportunities (Amankrah 2008) expose the youth to
manipulation by the older generation. Just as it is in other African countries like Kenya,
politicians and other influential persons entice the youth with rewards in order to help
them consolidate their hold on to power and influence, sometimes through the use of
violence (Kolo, 2010; Henningsen and Jones, 2013; Anderson, 2002)). Murunga (2006)
has argued, however, that any assessment of youth impacts on society should be on how
to harness their potentials for nation building, and not on the few instances where their
potentials have been allowed to get destructive due to bad policies and influences.
Despite the dominance of gender in the demographic and socio-economic discourse of
nations (UNFPA, 2003; Parmar et al, 2014; Neubert and Taggar, 2004), limited
4
University of Ghana http://ugspace.ug.edu.gh
attention has been given to gender as an important determinant of personal insecurity
in Ghana. The issue of gender and security in Ghana is not entirely terra incognita.
Existing literature pays attention to gender and human security in general, without
exploring the personal security dimensions of men and women in the face of increasing
crime, violence and abuse. For instance, gender insecurity as a result of domestic and
sexual violence has been widely studied, spearheaded by feminists and other activists
who are passionate about the rights of women in society (Ampofo, 1993; OfeiAboagye, 1994; Ardayfio-Schandorf, 2005; Cantalupo et al, 2006; Osam, 2004;
Tenkorang et al, 2013). Some work has also been done on gender and human security,
which is more concerned with survival strategies of women, especially those in rural
areas affected by famine and harsh climatic conditions (Pearl and Dankelman, 2008;
Glazebrook, 2011; Wrigley-Asante, 2008). In the context of conflict and post-conflict
societies, gender and insecurity discourses have been limited to sexual abuses by
soldiers, neglect of the security needs of women, and under-representation of women
in post-conflict peace building (Mama, 2014; Mama and Okazawa-Ray, 2012; Alaga,
2011; UNSSRTF, 2012; UNDP, 2012; UN Women, 2011).
However, no attention has been paid to the dilemmas facing both men and women in
societies plagued by some degree of crime and violence, including Ghana. Such
analysis can provide some feedback in holistically explaining Ghana’s low index of
personal safety. Structural factors perpetuating gender imbalances that subject men and
women to specific safety threats ought to be studied for deeper understanding of
security dilemmas of the state. This would help identify sources of structural violence
and structural inequality, or at least provide new insight into social relations that could
have impact on national cohesion. This study is aimed at treating perceptions of
personal security as an indicator of structural stresses suppressing the development and
wellbeing of members of society. The power relations between men and women in
patriarchal societies affect their respective experiences and perceptions of security
(Kameri-Mbote, 2005: 3). It is therefore helpful to have a gender-sensitive framework
for understanding personal security, and to explore underlying causative factors.
Global crime statistics show that men are both the major perpetrators and victims of
public crime or public violence, whilst women are major victims of domestic violence
5
University of Ghana http://ugspace.ug.edu.gh
(UNODC, 2011: 57-63; See also Harders, 2011). In 2008 for instance, 83% of victims
of homicide outside the home were men, suggesting a typical pattern of ‘men killing
men’ (UNODC, 2011: 63). Many acts of violent crime are committed within male
groups (Farr and Gebre-World, 2002), explaining why many prison inmates are men
(Harder, 2011: 143). Men dominate the military and security services, but are also
major victims of war and violence (The Rwandese and Bosnian experiences typify the
fact that men are major victims of war). Globally, up to 50% of women have
experienced domestic violence (Kishor and Johnson, 2004). According to the United
Nations Office of Drugs and Crime (UNODC) report on Latin America, women are
most likely to be murdered at home, whereas men are most likely to be murdered
outside the home (UNODC, 2011). A 1998 survey in Ghana showed that one in three
women had experienced violence at the hands of their current or recent partner (CokerAppiah and Cusack, 1999); and in 2010, some 109,784 cases of violence against women
and children were reported at the Domestic Violence and Victims Support Unit
(DOVVSU) of the Ghana Police Service (Ghanaweb, 2010). It is a truism, therefore,
that experiences of personal security vary between men and women. Whether they
perceive and express threats to their personal security differently is not well explored.
This is important because it provides a sense of how people can freely go about their
productive and livelihood activities at home and in their communities where they live.
This study is designed to provide a gender-sensitive analysis of perceptions of personal
security in Ghana, something which has been overlooked over the years. The study
acknowledges that perceptions of personal security do not translate into real
experiences of harm or direct threats to personal security. However, perceptions of
insecurity cannot be taken lightly, irrespective of the cause. Such perceptions have an
impact on access to health, education and quality of life, which makes the study very
timely at a time when the United Nations is making preparations to extend the mandate
of the MDG regime beyond 2015. Perceptions of personal security, experiences of
discrimination and inequality, and the freedom to live a fulfilled life are interlinked;
and these have an impact on Ghana’s ability to attain the MDGs.
6
University of Ghana http://ugspace.ug.edu.gh
1.3 Research Questions
Gender-sensitive assessment of personal security is important, since gendered roles,
relationships, opportunities and mobility vary significantly in Ghana, especially across
various regions and residential areas. In order to understand and offer explanations for
variations brought forth as a result of such gendered roles and relationships, the
following questions will be answered in the study:
How do perceptions of personal security vary in Ghana?
Why do variations exist in perceptions of personal security in Ghana?
1.4 Objectives
According to Kameri-Mbote (2004: 1), ‘gender as a relational concept contributes
significantly to an understanding of the causes and impacts of insecurity, and is also a
critical factor in the search for solutions to insecurity’. The power relations between
individuals and how their roles are defined in society are expected to have significant
impact on access to security.
Generally, this study aims to interrogate demographic predictors and consequences of
personal safety in Ghana. Specifically, it aims to
1. explore gender-sensitive perceptions of personal security in Ghana;
2. identify the determinants of perceptions of personal safety for men and women
in Ghana; and
3. inform policy on gender and security in Ghana.
1.5 Rationale for the study
Since gendered roles and relationships in many developing countries have a direct
bearing on wellbeing and safety of men and women, this dissertation analyses the
impact these differences have on perceptions of personal safety. The study is relevant
and timely for five reasons:
First, paying attention to gender-sensitive perceptions of personal security in Ghana
would provide useful feedback for government and policy makers on livelihood
strategies for men and women, which could also be a good indicator for gender equality,
providing another dimension for the evaluation of MDG goals 1 to 3. Men and women
7
University of Ghana http://ugspace.ug.edu.gh
have almost 50 percent representation of the population of Ghana, and so it is important
to pay attention to the differences in perception of personal security.
Second, perception of personal insecurity has impact on how people access health,
education, and basic social amenities. It also determines the degree of freedom that is
enjoyed by individuals. Efforts made by the United Nations, the international
community, and the government of Ghana aimed at improving the quality of life of
Ghanaians would be futile unless citizens live and perceive their personal security as
indubitable. Since men and women have nuanced experiences and perceptions of their
social realities, a study of their perceptions of personal security would help determine
the state of their wellbeing. This would help inform policy, especially in the natural
resource exploration sector, where employment opportunities are male-oriented.
Third, human security is directly premised on personal security, since vulnerability to
personal insecurity undermines human development (UNDP, 2012; UNDP, 1994).
According to UNDP, ‘people who fear for their lives cannot plan for a long-term and
more prosperous future’ (UNDP, 2012; 13). Personal insecurity undermines the
building of trust between individuals, their community and the nation, thereby
undermining efficiency of labour and productivity. National, regional and international
security will be unsustainable unless personal security is tackled as a fundamental
determinant of security (Kameri-Mbote, 2004: 3). It is therefore important to
understand the dynamics of gender and its impact on perceptions of personal security,
since the male or female, boy or girl, forms the fundamental building block of any
society. There cannot be any better early warning system for the security of the state
than to look at it in the prism of gender.
Fourth, the World Bank identifies security as an important pillar for poverty reduction;
and poverty reduction is described to include ‘reducing vulnerability’ to shocks,
disasters, health catastrophes and threats to personal safety (World Bank, 2001). Since
gender is a determinant of vulnerability in sub-Saharan Africa, a measure of gendersensitive personal security could expand the knowledge base of this nexus, and inform
policy. This is because perceptions of security have impact on access to health,
education and quality living of both men and women, without which would impede the
attainment of the Millennium Development Goals (MDGs); especially those goals
8
University of Ghana http://ugspace.ug.edu.gh
aimed at improving living conditions of peoples and eliminating major global and
gender inequities (UNCHS, 2003; Abama and Kwaja, 2009; see also Mudege et al.,
2008). Moreover, if the United Nations Security Council Resolution 1325 (UNSCR,
1325) is to make sense in post-conflict peace building, the security needs of both men
and women must be made priorities in non-conflict societies. This is because many
conflict and post-conflict gender-based atrocities are mere transfer of cultural norms
and practices from non-conflict contexts to conflict and post-conflict situations.
Fifth, for the first time in Ghana, data are available for measuring perceptions of
personal security through the introduction of the Governance, Peace and Security
module in the Ghana Living Standards Survey. This must inspire the interest of
demographers in exploring the nexus between gender, migration and socio-economic
characteristics on the one hand and perceptions of peace and security on the other hand.
It is important to know how these perceptions can influence demographic change and
vice versa; and this dissertation sought to do just that.
1.6 Clarification of Terms
Gender: Gender is a social construction of roles and relationships between men and
women of different age groups. These roles and relationships are constructed based on
biological categorizations of sex, and could change over time (World Bank, 2002: 1;
Moser, 1993: 230.)
Human Security: Human security is defined as safety from pervasive threats like hunger
and oppression, and protection from sudden disruption to daily patterns of human lives
(UNDP, 1994). It is a post-Cold War concept of security that took over from the statecentred concept of security that prevailed at the time. Human security encompasses
freedom from livelihood threat (like shortage of food, water, energy, economic),
freedom from threats to survival (death, injury and ill-health), and freedom from
indignity (absence of human rights and prevention of participation).
Perception: Perception is a mental process through which individuals interpret and
organize experiences and events to produce a meaningful world. A person’s
consciousness and acceptance of the event plays important part in the perception
9
University of Ghana http://ugspace.ug.edu.gh
process. Perception is subjective, and may not correspond to reality (Pickens, 2005:
52ff).
Personal safety: Personal safety denotes a protection from danger, risk, harm or injury
which have unintended or random sources. Such random or unintended sources include
accidents, environmental conditions, and other incidents occurring as result of one or
more coincidences. The relationship between personal safety and personal security is
directly proportional, but indirectly proportional to risk or danger. This is because
weakness in personal security increases risk, thereby decreasing personal safety (Byres
and Cusimano, 2010: para 3).
Personal security: Personal security is ‘the safety of human beings ... and the ways in
which they are enabled to live a full and productive life with dignity, without feeling
afraid’ of crime, violence, attacks, risks or other intended harms (Muggah, 2012: para
5). Personal insecurity occurs when a person or household is unable to have a fulfilling,
productive and dignifying life as a result of the fear of violence deliberately inflicted
for some gains or to create harm. There are two main determinants of personal security:
freedom from violence and freedom from the fear of violence or danger. The prevalence
of crime, unfavourable political conditions, and certain suppressive socio-cultural
practices perpetuate violence and fear, thereby creating personal insecurity.
State or National Security: The concept of national security is concerned with
protecting national borders, ensuring collective safety, and preventing sectarian,
regional and national conflicts.
1.7 Organization of the Study
The next chapter reviews existing literature on the relationship between gender and
perceptions of personal security. It would explore theories like patriarchy and
masculinity and how they perpetuate perceptions of personal (in)security among men
and women; and would expound on the conceptual framework for the study. Chapter
Three explains the methodology adopted for this study; and is followed in Chapter Four
by a profile of Ghana and demographic characteristics of persons who responded to the
module on Governance, Peace and Security in the 2012/2013 Ghana Living Standards
Survey. Chapters Five and Six deal with bivariate and multivariate analyses, and
10
University of Ghana http://ugspace.ug.edu.gh
explore the relationship between the various set of variables – independent, control,
intermediate and dependent. The final chapter provides a summary for the study,
concluding remarks, and recommendations for the entire dissertation.
11
University of Ghana http://ugspace.ug.edu.gh
CHAPTER TWO
LITERATURE REVIEW
2.1 Personal Security Defined
The concept of human security subsumes personal security (UNDP, 1994), as it
identifies ‘the protection of the individual as the starting point for political thinking and
practice’ (Werthes et al., 2011: 6; See UNDP, 1994: 22-23). Human security is a
safeguard of the ‘vital core of all human lives [against] critical pervasive threats’ that
hamper human fulfilments (Alkire, 2003: 2; see UNHSC, 2003) like hunger and
oppression, and protection from sudden disruption to daily patterns of human lives
(UNDP, 1994). Some scholars have proposed three distinctions of human security:
livelihood security, which includes food, water, energy, and economic provisions;
survival security, which includes absence of threat of death and injury or presence of
health; and dignity, which includes full human rights and participation in political life
of the state (Grossman and Owren, 2008: 25). Others have made two distinctions:
‘freedom from fear’ and ‘freedom from want’ (Acharya, 2001: 448; UNCHS, 2003).
Personal security is a survival security, freedom from fear (Alkire, 2003; Muggah,
2012; UNDP, 1994; Acharya, 2001; UNCHS, 2003), and is a phenomenon directly
related to experiences of individuals and households (UNDP, 1994; Muggah, 2012;
Acharya, 2001), unlike the concept of national security which is concerned with
protecting national borders, ensuring collective safety, and preventing sectarian,
regional and national conflicts (Matthews, 1989; Booth, 1991).3 Personal security is a
fundamental building block upon which a nation could achieve sustainable
development and stability, since according to the United Nations, ‘the world can never
be at peace unless people have security in their daily lives’ (UNDP, 1994: 1). Personal
security then becomes ‘the safety of human beings ... and the ways in which they are
3
In an era of globalization and interdependence of nations, human security, national security and global security are
intrinsically linked (Paris 2001: 98). Rotberg (2004: 5) and Piazza (2008: 85) have shown that where human security
is severely compromised, like the cases of Somalia, Afghanistan and some parts of Sudan, local and transnational
crime rate increases. Human security is therefore a critical dimension for the safety of individuals, the security of
nations and the international community.
12
University of Ghana http://ugspace.ug.edu.gh
enabled to live a full and productive life with dignity, without feeling afraid or at risk’
(Muggah, 2012: para 5).
Personal insecurity occurs when a person or household is unable to have fulfilling,
productive and dignifying life as a result of the fear of violence or threats (Moser and
McIlwaine, 2004; Moser and Rodgers, 2005). There are two direct determinants of
personal insecurity – experience of victimization and fear of any harmful event like
crime (Muggah, 2012; Moser and McIlwaine, 2004; Moser and Rodgers, 2005). The
prevalence of crime, unfavourable political conditions, corruption, human rights
abuses, widening gap of socio-economic inequality, natural disasters and certain sociocultural practices undermines personal security, and instil fear that could destabilize an
entire nation through the loot seeker (greed) and justice seeker (grievance) models of
conflict (Stewart, 2010; Keen, 2012; Collier and Binswanger, 1999; Collier 2000;
Berdal, 2005; Ballentine and Nitzschke, 2003; Ballentine, 2003).
Violence forms the core of personal insecurity, and is defined as ‘the exercise of power
that is invariably used to legitimize the use of force for specific gains’, socially
acceptable or not (Moser and Rodgers, 2005: 4). Violence could be ‘direct’ or ‘indirect’
(Alkire, 2003: 29), and its limit of acceptance could be part of the political, institutional,
economic and social frameworks of societies (Moser and Rodgers, 2005: 5; Galtung,
1969, 1985). Direct violence are those actions and inactions like domestic violence,
theft, rape, kidnapping, that are deliberately caused by state or non-state actors like
criminal gangs, thieves, rebels and spouses, with the primary aim of causing fear, panic
or injury in order to gain some benefit (Alkire, 2003). Indirect violence are structural
ones like corruption, non-responsive state institutions, vandalism or mob-actions, and
may be defined as ‘actions by groups, or systems or institutions whose threat to
[personal] security is a by-product of an action taken for a different primary purpose’
(Alkire, 2003: 29). All these affect perceptions of personal security.
13
University of Ghana http://ugspace.ug.edu.gh
2.2 Theories and Contexts of Gender, Violence and Personal Security
2.2.1 Patriarchy, Masculinity,4 Violence and Personal Security
The human species naturally bonds and builds relationships (Frith, 2013; Tomasello
and Vaish, 2012), the processes of which give them power to violate others, but also to
be violated by others (Armstrong and Bernstein, 2008; Castells, 2007). Societies
regulate such relationships through cultural, political and institutional norms and
practices; and coalesce power around a central point, the state (McAdam, 1982), and
around other multiple sources through socialisation (Armstrong and Berstein, 2008). In
almost all human cultures, one group exhibit some form of domination over the other(s)
through institutionalized power relationships, most of which turn out to be exploitative
(Francis, 2004; Eisler, 1990; Armstrong and Bernstein, 2008; Harders, 2011).
Patriarchy5 is an example of such systems created to regulate the relationship between
men and women (Holland, 2006; Walby, 1991; Kambarami, 2006; Sultana, 2011),
whilst masculinity and femininity are polar characteristics (stereotypes) that are
groomed and exhibited by men and women respectively (Kambarami, 2006; Sultana,
2011). Masculinity and femininity are perceived as polar concepts in many societies,
one being the opposite of the other (Swatay, 2012; Bhasin, 2004). Men are expected to
dominate, control and order, whilst women are expected to be submissive, controlled
and take orders. Both patriarchy and masculinity systematically undermine femininity
and the place of women in society (Swatay, 2012; Bhasin, 2004; Holland, 2006; Walby,
1991), whilst femininity condones and accepts the underprivileged place of women
(Kambarami, 2006; Eves, 2010; Kambrami, 2006). Together with masculinity and
femininity, the concept of patriarchy can help explain gendered inequality, differences
in experiences of violence and perceptions of personal security.
4
The emergence of homosexuality into dominant western socio-cultural discourses is radically shifting the concept
of masculinity and femininity from the male-female dichotomy to more nuanced and inclusive categorizations. Male
and female members of society can each exhibit feminine and masculine traits, according to this emerging trend of
socialization. This discourse, however, dwells on dominant cultural traits in developing countries, which persists
with the male-female categorizations of masculinity and femininity.
Patriarchy originates from ancient Greece, literally meaning ‘rule of fathers’. It was adopted in Europe in the
Seventeenth Century to justify feudal systems, aimed at nurturing and maintaining ‘principles of patrilineal descent
and inheritance’ (Edstrom et al. 2014: 3).
5
14
University of Ghana http://ugspace.ug.edu.gh
Works by Murshed (2002: 389) and Gurr (1970) argue that inequality in all shapes and
forms are acts of discrimination that create insecurity, grievances and rebellion if they
become systemic. Beyond the household level, some scholars have found a strong
connection between gender inequality and outbreaks of intrastate conflicts, an ultimate
source of personal insecurity (Caprioli, 2005; Harcourt, 2009). Countries with higher
levels of gender inequality are more likely to be involved in interstate disputes and more
likely to rely on violence to settle those disputes’ through the cultivation of ‘cultural
norms that legitimize violence’ (Caprioli, 2005: 162-163). This is consistent with
findings in Galtung (1985), which identified violent behaviour as the end product of
structural and cultural systems that condone and propagate violence in society.
Structural violence is understood as the systematic exploitations embedded into social
order, which is justified and legitimized by cultural violence that manifests itself in
politics, religion, ideology, language and art (Galtung, 1975: 265).
Some scholars argue that patriarchy creates and sustains structural suppressive systems
that condone vices and injustices against women (Meintjes et al., 2001; Francis, 2004;
Holland, 2006; Walby, 1991), and against other men who are considered to be
‘subalterns’ (Carrigan et al., 1985) or ‘unmarked male’ (Connell, 2005) due to certain
traits they posses which are not masculine enough. This affects perceptions, and
sometimes experiences, of personal security of men and women, especially in subSaharan Africa where little institutional frameworks exist to protect the interest of the
vulnerable members of society. Among women of the Kasena-Nankana district of
Ghana for instance, the ‘fear of violence is neither rare nor unfounded’, since women
perceive themselves to have ‘no power over [their] husbands’, who are culturally
allowed to beat their wives and have unlimited access to them at all times (Bawah et
al., 1999: 56-57). A Ugandan woman accounts: ‘sometimes, we are seriously wronged
by other people, usually men’, who are often protected by ‘their fellow rich and
powerful’ men (UN Women, 2011: 10). This fits Galtung’s notion of cultural violence,
which ‘makes direct and structural violence look, even feel, right – or at least not
wrong’ (Galtung, 1990: 291). Through all recorded history, security systems for women
have been ‘partial….elusive and mundane’ (Sylvester, 1994: 183; see also Blanchard,
2003: 1290), appending women’s security to men and grossing over their specific
15
University of Ghana http://ugspace.ug.edu.gh
security needs (UNDP, 2002; Estes et al., 2012; Kameri-Mbote, 2005; Cammack, 2000;
Kambarami, 2006; McDowell and Pringle, 1992).
However, all men do not reap the dividends of patriarchy and masculinity equally. Men
who exhibit traits of femininity or inadequate masculine traits are filtered out and
discriminated upon, especially in developing countries (Connell and Messerschmidt,
2005; Cornwall and Lindisfarne, 1994). Masculinity is a cultural ideal for male
members of society that creates a hierarchy or power relations among them based on
who is tough, has physical power, strength and ability to dominate (Harders, 2011: 143;
Page, 2010; Edstrom et al., 2014). Patriarchy dwells on masculinity to provide
privileges to ‘some men’, subordinating all others (Edstrom et al., 2014: 3). Irrespective
of their level on the masculinity strata, all men are postulated to be superior to women
(Edstrom et al., 2014; Harders, 2011; Eves, 2010; Page, 2010), and in many African
societies, ‘men are the dominant decision makers’ (Bawah et al, 1999: 61).
Some scholars argue that violence is intrinsically built into masculinity, and is sustained
by competition among men to reap the dividends of patriarchy (Harders, 2011; Eves,
2010; Page, 2010; Lo, 2010). Masculinity assumes that ‘rage is an acceptable response
to frustration; that vengeful violence is justified; and that to acknowledge loneliness,
pain, vulnerability, or to seek assistance is a sign of inadequacy’ (Page, 2010: 2). A
study in the Balkans in 2007 titled, Exploring dimensions of masculinity and violence,
found that men have generally been socialized to exhibit violence tendencies. For
instance, ‘A man should defend his honour, protect his family, and not let his children
or his wife be insulted’, said an adolescent male (Eckman et al., 2007: 17). Said another:
‘You must be protective of the opposite sex and family: Otherwise you are not a real
man’ (ibid); and yet another: ‘You have violence, and you have self-defence. Selfdefence is justified. We gather and we defend a friend. I am proud of that…’ (ibid). To
be a protector also implies to use force to protect if that becomes an option. Men are
not expected to exhibit emotions, not to cry, or to show any sign of weakness (Goodey,
1997; Eckman et al., 2007), as for instance, this statement from an adolescent: ‘typical
man usually doesn’t cry, but when he is drunk, it is acceptable’ (Eckman et al., 2007:
21).
16
University of Ghana http://ugspace.ug.edu.gh
Such masculine traits suppress men’s need for intimacy, sharing and self-disclosure,
which leads to frustrations that are vented out through different forms of violence, like
sexual violence, crime and victimization (Philaretou, 2001). It could also affect the
health of men through aggravated stress and cardiovascular diseases, with some studies
suggesting that men develop heart diseases 7 to 10 years earlier than women and thus
affect their life expectancy (Fodor and Tzerovska, 2004; Sharp, 1994), although the
overall rate of death from cardiovascular disease is the same for both men and women
(Maas and Appelman, 2010; Emslie, 2005; Vaccarino et al, 1999). The Word Health
Organisation estimates that 3.8 million men and 3.4 million women die from heart
diseases each year (Mackay and Mensah, 2004). This man does not cry syndrome can
have an impact on the quality of the measure of perceptions of personal security, since
men generally are not willing to express the extent of their personal insecurity.
Critics of the theory of patriarchy have argued that it is ahistorical and does not
desegregate the different experiences of women and men found in different cultures
and societies (Barrett, 1980; Rowbotham, 1981). Walby (1991: 2) disagrees with the
latter argument and insists that the concept captures ‘the depth, pervasiveness and
interconnectedness of different aspects of women’s subordination’, and could be varied
to account for the differences in approach to gender biases over different epochs.
Charvet (1982) argues to complement the position of Walby (1991), and explains that
the strength of patriarchy as a culture lies on its ability to draw on the authority of
tradition and universality, making injustices look almost right, or at least not wrong.
The next two sections explore fundamental basis for personal insecurity for both men
and women, both of which have patriarchy as the source. This underscores the fact that
patriarchy undermines the wellbeing of both men and women, although the impact is
felt most by women.
2.2.2 Personal Insecurity of Women
In the private space – in families and in households – women are the major victims of
violence and men are the main perpetrators (UNODC, 2011). This results from gender
roles and power relations that confine women to the domestic sphere (Alaga, 2011;
Harders, 2011; Kameri-Mbote, 2004; GSS, 2009; Agarwal, 1997); and which make
them exclusively dependent on men, especially in developing countries, where
17
University of Ghana http://ugspace.ug.edu.gh
marriage and other social norms restrict the ‘reproductive and sexual behaviours’ of
women (Horne et al., 2013: 503; Caldwell and Caldwell, 1987). ‘Limited educational
and training opportunities for women’ reinforce their dependence on men (Tenkorang
et al., 2013: 4), increases fertility, and exposes them to domestic violence and sexually
transmitted diseases (Tenkorang et al., 2013; Horne et al., 2013; Bawah et al, 1999).
Educational level of women, their socio-economic status, and place of residence
determines their degree of toleration of violence at home (Tenkorang et al., 2013;
Bradley, 2001; Bawah, 2002). More educated women are less likely to be maltreated
by their partners due to their increased bargaining power, and they are less likely to
tolerate gender-based violence. Women with many children are not able to compete
with men in the labour force, since much of their productive life is spent reproducing
and caring for children (Tenkorang et al., 2013: 4). This male dominance in family
arrangements puts women at a disadvantage. Horne et al. (2013) found some correlation
between payment of bride wealth and violence against women in Ghanaian homes.
Patriarchy and masculinity perpetuate violence against women (Harley, 2005), and
femininity condones such practices (Eves, 2010; Kambarami, 2006). All these affect
their general perceptions of personal security.
Shona men of Zimbabwe, for instance, ‘are socialized to view themselves as
breadwinners and heads of households whilst [women] are taught to be obedient and
submissive housekeepers’ (Kambarami, 2006: 2). In Papua New Guinea (PNG), boys
are ‘socialized from early childhood to interact assertively and aggressively with others’
(Eves, 2010: 53); whilst girls are taught to submit to their male counterparts at all times.
This makes women subservient to men, constantly living under the threat of violence
both at home and in the neighbourhood (Eves, 2010; AI, 2006; NRI and JAG, 2006).
About 67 percent of women in rural areas and 56 percent in urban areas in Papua New
Guinea have ever been abused by their husbands, according to available national figures
(Bradley, 2001; See also Eves, 2010; 54). In Kenya, 42 percent of some 612 women
surveyed in one district reported that they have ever been beaten by their partner, many
of whom reported repeated beatings (UNICEF, 2000: 5). A survey in 1998 showed that
a third of Ghanaian women interviewed had experienced violence at the hands of their
current or previous partners (Coker-Appiah and Cusack, 1999). In the Kasena-Nankana
district of Ghana, the following expression by an opinion leader captures the aggressive
18
University of Ghana http://ugspace.ug.edu.gh
attitude of men towards the sexuality of women: ‘if a woman is not experiencing her
menses and is not sick, she has no right to refuse sex…so if she refuses to have sex…I
will beat her’ (Bawah et al., 1999: 57). In India, up to 45% of married men
acknowledged that they have ever abused their wives physically, and 20% of men in
Thailand acknowledged the same (UNICEF, 2000: 5). Such gender-based violence is
not only prevalent in developing countries, according to the United Nations figures, but
also exists in developed countries. Up to 59% of Japanese women reported experiencing
physical abuses from their partners in a survey in 1993 (ibid); and in the United States,
28% of women reported at least one event of physical abuse from their partners, in a
national survey (ibid). These experiences of personal insecurity at home also affect
personal insecurity in the neighbourhood, as would be seen later in this section.
The United Nations further describes violence against women in prevailing patriarchal
systems as a ‘life cycle’ event globally. Table 2.1 summarizes the graduated cycles of
violence which women are exposed to throughout their life across the world.
Thus, women go through various degrees of violence and injustice that transcend any
age group. Right from conception, females begin to face the brunt of patriarchal
discrimination, through sex selectivity as is practiced in China and other Asian
countries. The impact of these on perceptions of personal insecurity is ‘comparable to
torture in both its form and severity (UNICEF, 2000: 4), but is difficult to measure
quantitatively. Proxies like suicide and wellbeing have been helpful in assessing this
impact.
Femininity is a socialization that condones these forms of violence and threats. In Papua
New Guinea for instance, some women think that ‘husbands don’t hit their wives for
no reason’ (Eves, 2010: 57). Such justifications are also common in Ghana (Horne et
al., 2013; Ofei-Aboagye, 1994; Amoakohene, 2004), perpetuated by certain cultural
practices and gender roles in marriages that make it impossible for women to defend
their rights in the face of threats to their personal security at home (Ofei-Aboagye, 1994;
Amoakohene, 2004; Bowman, 2003; Heilman, 2010; Horne et al., 2013).
19
University of Ghana http://ugspace.ug.edu.gh
Table 2.1 Instances of Violence against Women in their Life Cycle (UNICEF, 2000: 5)
PHASE
Pre-birth
Infancy
Girlhood
Adolescence
Adulthood
Elderly
and
TYPE OF VIOLENCE
Sex-selective termination of pregnancy;
Defects from battering pregnant woman due to birth outcomes.
Female infanticide
Physical, sexual and psychological abuse of girls
Child marriage; female genital mutilation;
Physical, sexual and psychological abuse;
Verbal aggression and constant humiliation;
Traditional practices harmful to female, like genital mutilations;
Incest; child prostitution and pornography
Dating and courtship violence, like date rape and physical deformities
Economically coerced sex, like school girls ‘sugar daddies’ for support;
Incest; rape; sexual harassment;
Traditional practices harmful to female, like genital mutilations;
Denial of access to funds and labour force participation restrictions;
Trafficking of women;
Dowry abuse and murder;
Verbal aggression and constant humiliation;
Threats to take away custody of children.
Forced ‘suicide’;
Homicide of widows for economic reasons;
Sexual, physical and psychological abuse.
A quantitative study by Horne et al. (2013) found some statistically significant social
support for men who beat their wives for sexually-related disagreements in the North
Tongu district of the Volta Region of Ghana. Although some of these findings appear
to be outdated and may not represent current realities, the principles underlying
patriarchy, masculinity and femininity transcend many epochs in history (Edstrom et
al., 2014; Holland 2006; Weber, 1978; Pateman, 1988; Engels, 1972).
Such experiences of personal insecurity at home also affect perceptions of personal
security outside the home. About 29% of Estonian women aged between 18 and 24
years, and 52% of those above 65 years, live in fear of domestic violence (UNICEF,
2000). In Papua New Guinea (PNG), feeling of insecurity at home is found to affect
women’s ability to move freely to assess public services like education, healthcare and
market centres (Eves, 2010; AI, 2006; NRI and JAG, 2006). Moreover, a woman who
has ever been abused is 12 times more likely to commit suicide than one who has never
been abused (UNICEF, 2000; UN, 1998). In the United States, up to 50% of women
who have been abused are likely to commit suicide (UNICEF, 2000).
20
University of Ghana http://ugspace.ug.edu.gh
Nevertheless, some scholars provide counter arguments to the dominant masculinityviolence discourse (West et al., 1978; Wilson, 1983; Gelles, 1972). They argue that
only ‘impaired’ masculinity is expressed with violence (West et al. 1978). ‘Problematic
childhood’, where there is no love and attention, prevent some men from acquiring
‘normal form of masculinity’, leaving ‘them unprepared to deal with the stresses of life’
(West et al., 1978, as paraphrased by Walby, 1991: 130). This makes them extremely
sensitive and unable to have measured control over social interactions with women and
with society in general, leading them to engage in anti-social activities like rape, theft
and wife beating. This kind of explanation is consistent with Freud’s psychoanalytic
projection of childhood anomalies into adult life (Freud 1965). Wilson (1983) and
Gelles (1972) have used class analysis from the Marxist perspective to provide support
for this abnormal masculinity explanation. They argue that male violence throughout
history has been most common in places where there is economic stress. In those places,
unemployed men and those men at the lower strata of society are the most culpable in
reported cases of crime and violence (Wilson, 1983 and Gelles, 1972). However, this
is an overgeneralization, since not all men with problematic childhood end up being
violent; and not all violent men had problematic childhood. There are also many more
unreported cases of rape and domestic violence from homes which may be classified as
upper class homes (Russell, 1982). Contemporary scholars have argued that new forms
of plural and sometimes competing masculinities have emerged in developing countries
as a result of formal education, labour migration and religion, which tend to openly
detest violent tendencies (Eves, 2010). In spite of these changes, all forms of
masculinity make ‘legitimate’ claim of power over the feminine (Eves, 2010: 52).
2.2.3 Personal Insecurity of Men
In spite of the dividends of patriarchy to men, and because of the burden of masculinity
– breadwinner, protector of households, provision of fatherly guidance to children, need
to exhibit powerful and assertive tendencies – some men who are not able to live up to
expectations of masculinity become antisocial and stressed out, with accompanying
feelings of inadequacy and low self-esteem (Lo, 2010; Harders, 2011; Eves, 2010;
Page, 2010; Eckman et al., 2007). These increase the likelihood for men ‘to perpetuate
violence’ (Eckman et al., 2007: 8). Some have argued that due to dwindling number of
women in Asia as a result of sex selectivity at conception, many men are denied wives,
21
University of Ghana http://ugspace.ug.edu.gh
and their ‘sexual frustrations’ are channelled into antisocial activities and sometimes
‘extreme nationalism’ (Lo, 2010: 3). Some scholars have proposed that the very nature
of masculinity requires close institutional and social management, lest its destructive
tendencies are unleashed into a self-destructive mode (Knight and Tullberg, 2011; Gill,
n.d;). Evidence of this is found in Nairobi Slums in Kenya (Dodoo et al., 2003). Poverty
compels men into drinking habits, in order to be insulated from frustrations of living
below masculine expectations:
Many men deliberately drink because, in this highly patriarchal society, they feel ashamed at
not being able to fulfil their ‘manly’ obligations as providers, because of unemployment or low
wages. (Dodoo et al., 2003)
This brings to the fore, the debate on whether masculinity is in crisis in developing
countries due to changing socio-cultural norms, persistently harsh socio-economic
conditions that do not allow men to live up to their socially-constructed responsibilities,
and assertive feminist ideologies (Clare, 2000; Seidler, 2005) that put men on their toes
without being prepared to deal with such challenges.
Thus, violence and crime are coping strategies for some men who are plagued with
unemployment, divorce, social stresses and poverty. Masculinities have generally
viewed the use of violence as possible option for resolving problems (UNICEF, 2000;
Borwankar et al., 2008); whilst femininity condones violence (Eves, 2010; Kambarami,
2006). The United Nations homicide statistics concludes, for instance, that ‘crime…is
typically a male activity’ (UNODC, 2011: 63). Yet, whilst 90% of people convicted of
crime globally are men, 82% of victims of murder are also men, even after accounting
for domestic violence which is mainly men aggression against women (ibid). By
engaging in violence and crime, they are able to convince themselves that they are also
powerful and confident, in the midst of perceived insurmountable challenges, thereby
consolidating their masculine position. Violence could be unleashed directly; through
economic crimes like robbery and theft, social crimes like rape and sexual exploitation,
and other forms of public and domestic mayhem (Walby, 1991); or indirectly through
institutional exploitations, corruption and mob justice (Walby, 1991; Alkire, 2003).
These notwithstanding, both men and women generally suffer the consequences of
insecurity that arise from crime and violence (UNDP, 2012: 13; Kameri-Mbote, 2005:
2).
22
University of Ghana http://ugspace.ug.edu.gh
Masculinity also directly increases the vulnerability of men (Grossman and Owren,
2008). As part of the cultures associated with masculinity, men become ‘overconfident’
and presumptuous in the face of risky situations in public; whilst women ‘are more
willing to adapt their strategies and behaviour’ (Grossman and Owren, 2008: 9).
Consequently, more men become victims of disasters and public violence because first,
men are socialized to be more adventurous than women, bringing them face-to-face
with disaster when it strikes; and second, men underestimate risks and only realize the
extent of exposure to risk when it is too late to escape (Bradshaw, 2004; UNODC,
2011).
The ‘men don’t cry syndrome’ (Eckman et al., 2007: 21; Goodey, 1997; Goldberg,
1979), where men ought not to admit fear or express public emotions, also affect the
ability of men to report of victimization perpetrated against them by both men and
women (Dutton and Nicholls, 2005). Studies have found that women have up to 10
times chance of reporting partner violence than men (Brown, 2004; Stets and Straus,
1992). A study in the United Kingdom also found that even when men report of
domestic violence to the police, arrest of the women perpetrators is only 60% of the
time, compared to 75% of the time when women report of being victimized (Brown,
2004: 34). The combined effect of the unwillingness of men to report domestic violence
and the reluctance of the police to arrest female perpetrators has meant that fewer
female perpetrators of violence get into the national statistics (Brown, 2004; Statistics
Canada, 2003). Statistics in Ghana shows that the trend is changing, though slowly. A
news item that made headlines in national dailies on 23rd March 2015 was captioned:
‘518 Men report wives to DOVVSU for beating them’. In that story, the Ghana Police
Service in Tamale Reported that ‘there is an increase in the number of men now coming
out boldly to report their abusive wives.’ The story stated that ‘In 2009, 71 cases were
reported by the men, 47 in 2010, 44 in 2011, 77 in 2012, 147 in 2013 and 132 in 2014’.
By contrast, almost 2000 cases of domestic violence were reported by women (The
Chronicle, Monday March 23, 2015). Dutton and Nicholles (2005) insists that the
gender differentials in domestic violence that places women as the perpetual victims
needs a rethink; implying that both men and women suffer domestic violence to almost
equal proportions. Some scholars have rebutted such claims of pre-emptive domestic
attacks on men by women, arguing that women only exhibit violence against men in
23
University of Ghana http://ugspace.ug.edu.gh
self-defence, even when the female violence is disproportionately deployed (Walker,
1984 and Saunders, 1988).
Vulnerabilities exist in both femininity and masculinity, even though the proportions
may be skewed against women due to the protection of men’s interests by patriarchy.
Such vulnerabilities are exploited by both men and women, leading to domestic
violence and abuses that also affect their general perceptions of security at home and in
the neighbourhood.
2.3 Conceptualizing Gender and Perception of Personal Security: Fear of
Crime and Social Disorder
Perception of personal insecurity could be situated in the fear of crime literature (Warr,
2000), since fear of crime is defined as ‘negative emotional reactions generated by
crime or symbols associated with crime’ and violence (Ferraro and La Grange, 1987:
72). Fear of crime is a driving force that can affect perception of personal security of
people, and this is aptly captured by the Business Day publication of December 3, 2006
as follows:
The most destructive characteristic of crime is the fear it places in our souls. The greater part of
that fear arises from the fact that crime is an amorphous presence, its sources only dimly
imagined. The fear sits on our shoulders and follows us everywhere. (Roberts, 2010: 250)
Fear is defined as the ‘institutional, cultural and psychological repercussion of violence’
(Kruijt and Koonings, 1999: 5). It generates and is generated by perceptions and
realities of ‘insecurity and vulnerability’ (Moser and Rodgers, 2005: 4; Arriagada and
Godoy, 2000) through a closed cycle where one leads to the other and is in turn
regenerated by the other.
Perception of personal insecurity and fear of crime are both emotional dispositions
which include mistrust for the other, fear of unknown persons, perceived risk at home
and in the neighbourhood, concern about downturn of neighbourhood safety conditions,
and deteriorating morality of a community (Warr, 2000). Two important defining
parameters stand out in this explanation. First, there is the emotional response to fearproducing situations; and second, there is the rational assessment of the risks for
victimization (Ferraro and LaGrange, 1987; Ferraro, 1995; Rountree and Land, 1996).
These define perceptions of personal security.
24
University of Ghana http://ugspace.ug.edu.gh
Findings on gender and perceptions of personal insecurity have been inconclusive.
Scarborough et al (2010) found that women are more likely to express higher perception
of personal insecurity than men. Schafer et al. (2006) on the other hand, found that both
men and women express relatively similar perceptions of personal insecurity: women,
for personal safety and personal victimization reasons; and men, for property
victimization reasons. The low sense of personal safety expressed by women is
especially due to their risk of exposure to sexual violence (Warr, 1984; Riger et al.,
1978; Stanko, 1990), which transcends all other forms of threat (Warr, 1984). This is
especially because even when a woman has been sexually abused, there is the likelihood
that she would be accused of inviting such violence upon herself due to her actions and
inactions. The fear of property victimization expressed by men is due to their need to
be breadwinners and protectors of the household (Schafer et al., 2006).
Closely related to perception of personal insecurity is the broken windows thesis. The
broken windows perspective looks at the relationship between weakening social order
and crime (Wilson and Kelling, 1982). It proposes that there is a direct relationship
between social disorder and crime. Social problems like poverty, unemployment,
broken homes, and overcrowding in communities have positive correlation to
perceptions of insecurity and fear of crime (Sampson and Raundebush, 2004; Wyant,
2008). Social disorder, like public display of alcoholism, dilapidated structures,
numerous abandoned cars, indiscriminate deposit of garbage, high rates of loitering,
and high child delinquency, can have two effects on personal security of both men and
women in society:
First, it creates fear in community members (Scarborough et al., 2010) since they know
that they may not have social support in their times of need. This motivates people or
groups to put up defences for protection, legitimate or otherwise, thereby increasing the
sense of fear and insecurity for the socially vulnerable. This is likely to embolden
violent men, for instance, to be more aggressive towards those they can easily
overpower and exploit (Garofalo, 1981; Goodey, 1997; Madriz, 1997; Stanko, 1990);
or to encourage gated communities and high-walled houses (Jackson, 2006; Moller,
2005; Valji et al., 2004; Roberts, 2010). This can have a self-fulfilling effect of mistrust,
25
University of Ghana http://ugspace.ug.edu.gh
since those who are capable of securing their own safety create ‘barriers’ and
misconceptions that inhibit social cohesion (Valji et al., 2004).
Second, such visual cues of social disorder give a signal to criminal offenders that there
is no social cohesion in the community, that the residents are indifferent to their
surroundings, and that they lack any collective strategy to deal with criminals and those
who breach social norms (Sampson and Raudenbush, 2004; Wilson and Kelling, 1982).
Scarborough et al. (2010) found a strong correlation between perception of social
disorder and higher perception of personal insecurity.6
Since women are socialized through femininity to be very sensitive of their
physiological and social vulnerabilities (Scott, 2003; Goodey, 1997; Madriz, 1997;
Stanko, 1995); to feel that their security needs are appendages to that of men
(Kambarami, 2006; Kameri-Mbote, 2005; Cammack, 2000; Riger, et al. 1978); and that
the wellbeing of their children is exclusively theirs (Gilchrist et al., 1998); any increase
in social disorder is likely to affect their perceptions of personal insecurity. This is
consistent with the findings of Schafer et al. (2006), in which women were found to be
more sensitive of their personal safety. Men, on the other hand, have been socialized to
be protectors, breadwinners, and property owners; so they have their own sense of
insecurity in the midst of social disorder, which some scholars have referred to as ‘Fear
of property victimisation’ (Schafer et al., 2006: 289; McGarrell et al., 1987). That is,
whilst women demonstrate perceptions of personal insecurity due to the possibility of
being raped in the event of violence like burglary (Warr, 1984; Ferraro, 1996; Fisher
and Sloan, 2003); men express similar perceptions of insecurity for loss of property
(Schafer et al., 2006).
Whilst the broken windows hypothesis proposes that the police and other national
institutions should move to instil social order to curb domestic violence and other
criminal activities (Hinkle and Weisburd, 2008), some scholars who belong to the
Collective Efficacy Perspective school have proposed that enhancing social cohesion
6
It however could not prove that such fear could actually lead to crime as proposed by the broken window
hypothesis.
26
University of Ghana http://ugspace.ug.edu.gh
could be more effective if it is done independent of formal social or national institutions
like the police (Wyant, 2008; Sampson and Raudenbush, 2004; Sampson et al., 1997).
This is because the continuous presence of the police and other security agencies is
itself a source of insecurity due to the signal it sends that there are problems in the
community (Hinkle and Weisburd, 2008). The capacity of the security agencies would
also be stretched beyond limits if they have to be visible in every single community
(Bursik and Grasmick, 1993). Enhancing the capacity of both men and women through
education and social support systems could reduce gender inequality and social disorder
(Buchmann and Hannum, 2001; Ogundari and Aromolaran, 2014; Kuenzi, 2005;
Gyimah, 2009), and may be a better option in dealing with the broken window problem.
2.3.1 Other Determinants of Perceptions of Personal Security
Place of residence and safety of neighbourhoods also shapes people’s perception of
security (Brunton-Smith and Jackson, 2012; Ferraro, 1995; Sampson et al., 1997;
Jackson, 2005; Tulloch, 2003). Studies have shown that it is not only events in
immediate neighbourhood that determine people’s perceptions of security, but the
wider social and geographical context of the state and region, since the definition of a
person’s neighbourhood is more complex than administrative demarcation of regions
and states. There is a ‘spill over effect’ of threats, public crime and violence beyond
administrative boundaries (Brunton-Smith and Jackson, 2012: 56), and therefore
beyond the home and immediate neighbourhood. People are likely to have a higher
sense of insecurity, even when their homes and immediate social environment are
without crime and disorder, but share boundaries with geographical areas that have
higher rates of social disorder and crime.
Commercial exploration of natural resources also accounts for some variation in
perceptions of personal security. Le Billon (2001) argues that natural resource
availability in a country heightens insecurity through three dynamics: First, if the
resource is lootable (easily stolen), like Diamonds, crime increases, and those who wish
to rebel against the state can easily have access to funds to start a rebellion (See also
Ross, 2004). Second, the cost-benefit analysis becomes favourable to engage in crime
or rebellion where benefits from these resources can easily be accessed illegally; and
third, the benefit that may accrue to power wielders in a stable political environment
27
University of Ghana http://ugspace.ug.edu.gh
becomes lower than an unstable context, especially in sub-Saharan Africa where there
are weak institutions (Le Billon 2001: 574), thereby discouraging any serious
commitment to peace deals when armed conflicts arise. Both the greed and grievances
explanations (Stewart, 2010; Keen, 2012; Collier and Binswanger, 1999; Collier, 2000;
Berdal, 2005; Ballentine and Nitzschke, 2003; Ballentine, 2003) fit well in the nexus
between natural resources and personal security, since crime rate, depletion of the
environment, and general livelihood challenges increase with commercial exploration
of natural resources (UNEP, 2011). The Niger Delta region of Nigeria is a typical case,
where commercial crude oil exploration has resulted in massive destruction of the
ecosystem in the Ogono area (Steiner, 2010; Vidal, 2010), increased militia clashes
with government security agencies, maritime insecurity, hostage taking for ransom,
general disorder, and destruction of properties (Ejibunu, 2007). Such clashes of power
and exploration processes exclude women; especially since exploration of most natural
resources are all men affairs. This makes women passive players in resource-rich
regions of any country. This potentially heightens their sense of personal insecurity.
Scarce or depleting natural resources can affect personal security as well, with some
studies predicting that inadequate rainfall that leads to drought and economic hardships
can increase the probability of armed violence by about 50 percent (Edward et al.,
2003). Women are disproportionately put at disadvantage in such situations, since they
are socialized to do domestic chores like cooking, washing and fetching firewood in
many developing countries. Depletion of renewable resources, especially, affects the
place and wellbeing of women. In 2003, prolonged droughts in Darfur in Western
Sudan led to struggles over scarce resources like water and renewable resources for
subsistence, which increased rates of crime and personal insecurity; and exposed many
women to rape, violence and sexual abuse. This became a battle for survival between
non-Arabs and Arabs led by the Janjaweed militia (Sharaky, 2005; Hagan, 2007;
Giddens, 2011). This natural-resource-violence discourse remains controversial, since
countries like Japan have attained political stability without natural resources, and
Norway has attained same with natural resources.
28
University of Ghana http://ugspace.ug.edu.gh
2.4 Perceptions of Personal Security and Complexities in Measurement
Perception of security is difficult to measure, since perception is subjective and exposed
to biases (Pickens, 2005). Perception is contingent on memory, time interval between
exposure to events and recall, and visibility of event. Perceptions of personal insecurity
are borne out of fear, especially when people believe they themselves, the society or
the state have lost control over their immediate social environment (Lewis and Salem,
1986; Tulloch, 2003). This may not always correlate to real threat to security (Pickens
2005), although the mere fact that there is perceived threat or fear cannot be discounted
(Brunton-Smith and Jackson, 2012; Ferraro, 1995; Jackson, 2005). Perceptions of
personal insecurity, whether borne out of real threat or perceived threat, have significant
impact on people’s behaviour (Brunton-Smith and Jackson, 2012; Tulloch, 2003); and
could affect their mobility, education, productivity, health and wellbeing (Jackson and
Stafford, 2009; Stafford et al., 2007). Studies have found that people with greater
perceptions of insecurity (fear) are 1.93 times more likely to have depression than
people who have lower sense of personal insecurity (Stafford et al., 2007). People with
higher perception of personal insecurity exercise less, socialize less, and are more
confined to themselves than those who have higher perception of security, which have
impact on their health and mental status. A study among slum dwellers in Nairobi found
that the fear of being raped, murdered or attacked affected the movement of adolescents,
which in turn affected their education (Mudege et al., 2008). A primary school pupil,
for instance, recounted why she dropped out of school:
I used to walk to school through some risky area called Rurii. Sometimes people used to be
murdered in that area… The road which is secured was a long distance to school; and also my
mother had no money for my bus fare….Children who were late for school were punished and
sent back home. I was in a dilemma…. I started missing school; I continued missing school
until I completely dropped out of school. (Mudege et al., 2008: 106)
In another instance, one of the students recounted:
You meet people and they rape you. Now you lose hope and say, ‘ah! I will never go to school
if that school is what got me raped. (Mudege et al., 2008: 107)
Yet, perception of insecurity may not necessarily correspond to real threat or
victimization, since perception is affected by time, memory, proximity of other cases
of violence, and subjective interpretation of events. The disconnection that can exist
between perception of insecurity and immanence of real threat can be exemplified by
29
University of Ghana http://ugspace.ug.edu.gh
two South American surveys conducted by the UNDP. Five out of ten South Americans
perceived that security in their countries had deteriorated in 2012. However, eight (8)
out of ten (10) people felt safe in their neighbourhood in Honduras, where murder rate
is highest in the world (86.5 murder cases per 100, 000 people). In Chile, where murder
rate is lowest in the region (2 murder cases per 100, 000 people), perception of safety
is lower than in Honduras, with seven (7) out of ten (10) people feeling secure in their
neighbourhood (UNDP, 2013).
Stiglitz et al. (2010) aptly captured this disconnection between perception of insecurity
and actual experience of victimisation in the Commission on the Measurement of
Economic Performance and Social Progress report:
Countries with a higher share of people reporting fear of crime do not experience a higher
victimization while, within countries, older and richer people feel more unsafe than younger
and poorer people, despite being less likely to be victims of crime. (Stiglitz et al., 2010: 53)
2.5 Conceptual Framework
The focus of this study is to understand the factors that drive variation in perceptions
of personal security expressed by men and women (See Figure 2.1 below).
A number of studies have been done on gender-based perceptions and fear of crime
(Schafer et al., 2006; Scarborough et al., 2010; Ferraro and LaGrange, 1998; Acierno
et al., 2004). There is increasing evidence that violence against women, perpetuated by
patriarchy, increase fear (Horne et al., 2013; Ardayfio-Schandorf, 2005; Lannert et al.,
2014; Walby, 1991; see contrasts by Dutton and Nicholls, 2005). Some few studies
have looked at the anxieties of men due to the burden of masculinity (Bawah et al.,
1999; Eckam et al., 2007; Swatay, 2012). Others have found that fear may not be borne
out of direct threat, but out of altruistic concern for other ‘vulnerable’ household
members like children (Warr and Ellison, 2000).
30
University of Ghana http://ugspace.ug.edu.gh
Intermediate
Fear Facilitators
Dependent
Independent
Gender
Tension or conflict in community
Experience of robbery
Experience of attacks or threats
Control
Spouse
Co residence
Age
Residential status
Region of
residence
Household size
Economic
activity
Education
Religion
Perceptions
of personal
security
(scale)
Fear Inhibitors
Confidence in pub security service
Trust for neighbours scale
Government responsiveness
Private Security Index
Figure 2.1: Conceptual Framework
[Adapted from McGarrell et al (1997), Schafer et al (2006) and Scarborough et al (2010)]
The work of McGarrel et al. (1997) is particularly useful for this study, since it
considered the impact of demographic, neighbourhood and social conditions on
perceptions of personal security. In their model of fear, they used a combination of
social order, social disorder and community concern hypotheses, which were later
explored with gender dynamics by Schafer et al. (2006). Together with what has
become known as the indirect victimization model of fear (Garofalo and Laub, 1978;
Will and McGrath 1995), this study derives its conceptual framework. This indirect
victimization model counters the argument that victims of violence are most likely to
have higher personal insecurity (Skogan and Maxfield, 1981); and argues that people
who feel more vulnerable in society, like women and elderly, are likely to feel more
insecure (See also Wyant, 2008).
31
University of Ghana http://ugspace.ug.edu.gh
2.5.1 Dependent Variable: Perception of Personal Security
Measuring perception of personal security is difficult, since there is no conceptually
definite distinction between fear of specific crime and general sense of personal
insecurity (Ferraro, 1995; Ferraro and LaGrange, 1987; Warr, 2000). Perception of
personal insecurity, like fear of crime, covers ‘a wide range of subjective and emotional
assessments and behavioural reports’ (DuBow et al., 1979: 1). Since perception of
security is difficult to measure directly, some indirect approaches have been used.
Kaplinsky (2000), for instance, uses well-being of the individual as an indirect measure
for it; Mudege et al. (2008) use actions and behaviours of parents and students in
response to threat to measure it; whilst Gallup and Barometer surveys use behaviours
like “walking alone at night” or “women walking unaccompanied at night” to measures
it. Although Ferraro and LaGrange (1987) refers to this traditional ‘fear of walking
alone at night measure’ as vague and formless, which does not measure any specific
form of threat, it is useful in measuring a general sense of perceptions of personal
security over time and across regions (Roberts, 2010).
2.5.2 Independent Variables
A person’s real or perceived physical, social and demographic vulnerability can affect
the perception of security (Schafer et al., 2006; Goodey, 1997; Skogan and Maxfield,
1981). Persons who are socially vulnerable and disadvantaged, like women in the
context of patriarchy and the elderly, are overwhelmed by fear when they live or work
in places which have higher rates of crime, social disorder and lower social support
(Wyant, 2008). Some studies found strong correlation between gender and personal
safety (Scarborough et al., 2010; Warr 1984), although Schafer et al. (2006) found no
significant relationship between gender and safety in general. Whilst studies found that
men have lower sense of security when it comes to safety of properties, especially due
to patriarchal social arrangements that apportion property acquisition to men (Schaffer
et al., 2006); Others contend that women always exhibit the lowest sense of security,
since they live in the fear that they may be sexually abused (Warr, 1984; Farraro, 1996;
Fisher and Sloan, 2003). There are others who believe that men, especially young men,
are mostly victims of public crime, and so exhibit the greatest sense of personal
insecurity in places of social disorder (Rand, 2008). Education has also been found to
32
University of Ghana http://ugspace.ug.edu.gh
be a significant predictor of fear. As level of education increases, perception of personal
insecurity decrease (Scarborough et al., 2010) for both men and women. This is
probably because certain perceptions and misconceptions about safety in
neighbourhoods are cleared with higher education. Limited education for women
(Tenkorang et al., 2013) reinforces their dependence (Ogundari and Aromolaran, 2014;
Kuenzi, 2005; Gyimah, 2009), increasing their chance of feeling insecure in the absence
of rigorous social support systems. Minority residents and low-income individuals have
lower sense of personal security (Garofalo, 1981; Will and McGrath, 1995). Other
demographic characteristics that could affect perceptions of security include place of
residence, household size, and marital status.
2.5.3 Intermediate Variables 1: Fear Facilitating Factors
Certain behaviours, attitudes and perceptions have the capacity to reduce perception of
security (McGarrel et al., 1987; Schafer et al., 2006). Social disorder and perceptions
thereof induce individual concerns for communal integrity, leading people to think that
the community no longer has grip over its environment. Frequency of violence,
hooliganism, theft, illegal activities and child destitution degrades people’s perception
of safety; and could affect social integration. Skogan (1990) calls this the spiral of
decadence, which creates anxiety and relapse into conditions of social disorder.
Although a much contested thesis, the broken window thesis helps explain this kind of
connection (Wilson and Kelling, 1982; see also Scarborough et al., 2010; Skogan and
Maxfield, 1981). When a person perceives that the frequency of disorder in a
community is higher, such perception could be translated into real personal insecurity.
This is what has been termed as the risk of violence in this study.
There is also the altruistic nature of expression of personal insecurity, where the
presence of children and other vulnerable members of society (Warr and Ellison, 2000;
Gilchrist et al., 1998) can lead people to rate their perception of security lower. Women
with many children are unable to compete fully in the labour force since they spend
much of their productive life bringing forth children and caring for them. This confines
them mostly to the home, and might affect their perception of personal security when
the community is thought to be unsafe. Personal experiences of violence, attacks and
robbery can affect people’s perception of safety (Skogan and Maxfield, 1981), although
33
University of Ghana http://ugspace.ug.edu.gh
this has been contested in recent times. Those who are the most at risk of gang attacks
like young men, for instance, are found to have lower sense of insecurity, in contrast to
those who are least at risk of attack, like women and elderly (Garofalo and Laub, 1978).
2.5.4 Intermediate Variables 2: Fear Inhibitors
Some social factors act as buffer against perceptions of insecurity (McGarrell et al.,
1987). These factors can counter the effects of fear facilitators like disorder, crime and
other forms of social ills. These factors reassure residents that they are together in
dealing with communal stresses that threaten their very survival. Social cohesion, trust
in relatives and neighbours, assurance that the government has individual’s interest at
heart, and assurance of efficient state security services can boost such feelings of
personal security. Availability of neighbourhood watch scheme and other private
initiatives to boost personal security could also have impact on perception of personal
security. This category of variables is important for informing policies aimed at
targeting wellbeing of persons in society.
2.6 Hypotheses
The following hypotheses are informed by three major scholarly perspectives. The first
is that men are mostly perpetrators of crime and violence in developing countries
(UNDP, 2012; Kameri-Mbote, 2005; UNODC 2011), and that women’s fear of sexual
assault heightens their sense of insecurity (Warr, 1984). Second, patriarchal societies
protect the social, economic and political interest of men and leave women vulnerable
(Cain et al., 1979; Chesler, 1991; Hunnicutt, 2009; UN Women, 2011). Third,
commercial exploration of crude oil and other natural resources could breed insecurity
and political instability in developing countries in the absence of robust institutions
(Auty, 1993: 1; Le Billon, 2001: 562; Ross, 2012: 5; Humphreys, 2005: 27; Collier et
al., 2003).
1. Women are more likely to feel insecure than men in Ghana;
2. People who stay with their spouses are more likely to have higher perception of
personal security than those who do not stay with spouses.
3. Residents of resource-rich regions are less likely to feel secure than residents of
non-resource-rich regions.
34
University of Ghana http://ugspace.ug.edu.gh
CHAPTER THREE
METHODOLOGY
3.1 About the GLSS 6 Data
Surveys and censuses are the main sources of national demographic data in Ghana, as
is the case in many other developing countries. The Ghana Living Standards Survey
part 6 (GLSS 6) is the source of data for this study. GLSS 6 was conducted between
October 2012 and October 2013, by the Ghana Statistical Service (GSS). The twelvemonth data collection period was divided into 10 cycles of 35 days each. Some 18,000
nationally representative households in 1,200 enumeration areas were selected for the
survey. The response rate was 93 percent. Weights have been computed into the data
for analysis since according to the Ghana Statistical Service, ‘disproportionately larger
samples from regions with smaller populations were drawn’ to account for certain
socio-demographic and ethnic dynamics in regions which are sparsely populated (GSS,
2014: 207), especially in the three northern regions. This means that ‘each sample
household did not have the same chance of selection into the survey’ (ibid). Thus, the
weights were computed by GSS to account for the different probabilities of selection
of each household. These weights were based on regional variations in population size,
composition and characteristic, as was declared in the 2010 Population and Housing
Census.
In the sampling process, the country was divided along the borders of the ten already
existing administrative regions; and then further sub-divided into urban and rural strata.
In order to account for varying degrees of livelihood strategies and geographical
differences across the country, further subdivisions were made according to the three
ecological zones – Coastal, Forest and Savannah. Due to the unique nature of the
Greater Accra region, as the national capital and most industrialized city in the country,
it is always treated as a separate geographical category, often referred to as the Greater
Accra Metropolitan Area (GAMA) in the data. The Confidence Interval chosen for the
study is 95 percent, suggesting an error margin of 5 percent.
A new module was introduced into the GLSS 6 – the Governance, Peace and Security
Questionnaire – which is the primary source of data for this study. The Governance,
35
University of Ghana http://ugspace.ug.edu.gh
Peace and Security module collected information from household heads or their
designated representatives, unlike the other modules that collected data for all
household members and household characteristics. This is mainly a perception module,
soliciting information on how respondents perceive certain safety, security and
governance stimuli in their communities. The unit of analysis in this study, therefore,
is the household head. The total sample size is 18,000 since that is the total number of
households sampled for the survey. There were five other modules in addition to the
Governance, Peace and Security Questionnaire, which together comprise GLSS 6. The
GLSS was first conducted in Ghana in 1987, with subsequent ones in 1998, 1991/92,
1998/99 and 2005/06. The 2012/2013 round is the sixth and latest one, and is the only
one yet to have introduced a module on Governance, Peace and Security.
The Governance, Peace and Security module had sections soliciting information on
experiences of theft in households, sexual offences against respondents, their account
of violence and security in the community, their perceptions of safety, their accounts of
peace and social cohesion, their engagement in political activities, and their assessment
of national governance. Information solicited on all the various sections were used in
this analysis, except the sexual assault section. Information from this section was
excluded because only 2 percent of respondents reported that they have ever been
sexually assaulted, whilst the rest said they have never been sexually assaulted.
Nevertheless, it is important to note that out of a total count of 331 respondents who
reported that they have ever been sexually abused, almost a third (65.9 percent) were
females. On its own merit, this could have interesting dynamics on perception of
personal security; but the relatively smaller numbers could not allow for further analysis
in the overall model for this study.
The study is not limited to particular regions or districts; but covers the perspectives of
all the household heads or their designated representatives who took part in the survey.
All these respondents shall together be referred to in this study as household heads,
although in some instances, representatives of household heads responded to the
questionnaires on their behalf where they were absent. Statistical Package for the Social
Sciences (SPSS) was used for the entire analysis, as would be seen in subsequent
36
University of Ghana http://ugspace.ug.edu.gh
chapters. All the files in the Governance, Peace and Security folder prepared by the
GSS were merged using Household Identification Number (HID) as the ‘Key Variable’.
3.2 Methods of Analysis
3.2.1 Dependent Variable
A four item additive scale was used to operationalise the outcome variable, the
perception of personal security in Ghana. This perception measures the general sense
of contentment and peace of mind that people have in their daily lives, homes,
neighbourhood and community. Different stimuli may affect people’s perceptions
about safety in walking alone at night in their neighbourhood, their comfort in staying
alone at home at night, general contentment about safety during ordinary daily
activities, and their perception about household safety. These are expected to be distinct
but interrelated stimuli. Therefore, no single variable on safety is expected to capture
the overall perception of personal security of individuals. That is why four questions
have been combined into a scale to measure perception of personal security. To ensure
that the four variables for the scale were not arbitrarily chosen, a reliability analysis test
was done (Alpha=0.857).7
Generally, a Cronbach’s Alpha that is greater than 0.7 indicates a statistically reliable
coherence between the variables that are to be used for the scale.8 Table 1 of Appendix
A shows Item-Total Statistics for the four variables. The ‘Cronbach’s Alpha if Item
deleted’ values range from 0.809 to 0.828, indicating that deleting any of the four
variables would reduce the reliability of the scale in representing the individual
variables. Each of the four questions or variables therefore complements the accuracy
of the scale in measuring the perception of personal security. The questions were
selected from the Safety section of the module, and include the following:
Do you feel safe walking down your street alone in your neighbourhood at night?
How safe do you feel when you are at home alone after dark?
7
On SPSS: Analyse>Scale>Reliability Analysis_Statistics: Scale if Item deleted.
8
To double-check the compatibility of these four variables for the scale, a factor analysis was conducted, which
generated a single component with eigenvalue 3.233.
37
University of Ghana http://ugspace.ug.edu.gh
In your daily life, do you feel very safe or not safe at all?
Would you say your household is safe from crime and violence at home?
Each of the questions had the following answer options: 1=Very safe; 2=Fairly safe;
3=Safe; 4=A bit unsafe; 5=Not safe at all. The direction of these response options was
reversed and spread along a continuum from 0 to 100,9 using the ‘Recode into different
variables’ command on the Transform menu of SPSS. This process also transformed
each of the questions from ordinal or ranked variables into continuous variables. In
order to create the composite scale for perception of personal security, all four variables
were combined into a single scale. The ‘Compute Variable’ command was used to
create average scores for the four variables. This scale is what shall be termed
subsequently in this study as the Perception Scale. This is the Perception of Personal
Security Scale.
This perception scale varies from ‘Very high perception of personal security’ (a score
of 100) to ‘Very low perception of personal security’ (a score of 0). In other words, in
the continuum from 0 to 100, lower values represent higher sense of personal insecurity,
and higher values represent higher sense of personal security.
3.2.2 Intermediate Variables
Two sets of intermediate variables are used to ascertain the pathways through which
perception of personal security is influenced by gender. These two sets are fear
facilitators and fear inhibitors. The fear facilitators are expected to test how certain
stimuli that instil fear could have direct bearing on perceptions of personal insecurity;
whilst the fear inhibitors would test the pathways through which certain social
phenomena that reduce fear could have influences on perception of personal security.
These two intermediate variables are discussed below.
9
In the recode, the ‘Not Safe At all’ option (originally coded as 5) was coded as 0, indicating that there
is absolutely no sense of safety for the respondent; whilst the ‘Very Safe’ option (originally coded as 1)
was coded as 100, indicating absolutely high sense of safety.
38
University of Ghana http://ugspace.ug.edu.gh
3.2.3 Intermediate Variables 1: Fear Facilitators
A number of factors are predicted to have negative correlation to perception of personal
security, especially due to their potential to instil fear in people. These factors include
proximity to some form of social disorder, the presence of children in the household,
personal experience of crime, and personal experience of attack, assault or threat.
Proximity to some form of social disorder increases a person’s perception of risk to
violence. This is consistent with the broken windows hypothesis, in which disorder has
both psychological and pragmatic impact on perception of security (Skogan, 1990;
McGarrel et al., 1997). Both the intensity and timing of the events of social disorder
determine how perception of personal security is affected. Such stimuli of social
disorder do not have to affect the respondent or the household directly to influence
perceptions, but could be very influential even if the event occurred in the community,
town or neighbourhood.
The first variable measures whether respondents perceive that there is tension or
conflict between different groups or social actors in the community. A dummy variable
was created for this question, with 1 representing ‘Yes, there is tension or conflict’; and
0 representing ‘No, there is no tension or conflict’. The first category is a combination
of varying degrees of ‘Yes’, based on time and intensity. The second category is a
simple denial of any form of tension or conflict in the community.
The presence of a child in a household is another fear facilitator that could affect
perception of personal security. It would be interesting to see how this affects the
perception of personal security of men and women, since some studies found altruistic
feeling of insecurity due to the presence of children at home. There were 11, 498
children under five years in the total household membership of 72,372, representing
15.9 percent of all members in households sampled for the Living Standards Survey.
However, due to the household arrangement in Ghana, where household heads may not
be direct biological parents of children in the household, this variable was dropped from
the list of variables used for the analysis.
39
University of Ghana http://ugspace.ug.edu.gh
Two other variables were used to separately measure personal experiences of violence
– Personal experiences of robbery and attack or threat. These are direct manifestation
of social disorder in the neighbourhood. The first question is: ‘Over the past five years
(since 2008), did anyone steal or try to steal anything from you or other members of
your household?’ The second variable is paraphrased as: In the past 12 months, have
you or a member of your household been attacked, assaulted or threatened at home or
elsewhere in the community by someone in ways that frightened you? Each of the
variables was recoded into dummy variables. The recoding merged personal
experiences with household experiences into a single category: ‘Yes, myself or a
member of my household’ [1=‘Yes, myself or a member of my household’; 0=No,
neither myself nor a member of my household]. Three main factors influenced the
decision to combine these categories. First, the responses for the household experiences
were negligible (2.5 percent), especially for the attack or threat variable. Second, the
researcher assumed that ties between household members in many parts of Africa is
very strong, making emotional responses to violent experiences a shared household
phenomenon. Third, violent acts like robbery and threats affect the collective wellbeing
of household members. Consequently, it is sometimes difficult to draw a distinct line
between them.
3.2.4 Intermediate Variables 2: Fear Inhibitors
A number of social processes and events are expected to enhance the perceptions of
personal security of individuals. These factors are predicted to boost perception of
personal security, acting as buffer against perception of personal insecurity.
The first fear inhibitor is the expression of confidence in public security services. This
is expected to boost the perceptions of personal security. A three-item additive scale
was used to operationalise confidence in public security services (Alpha=0.764). Table
2 of Appendix A presents the Item-Total Statistics for the reliability analysis. The three
questions were chosen from the Violence and Security section of the module; and they
are:
In general, how much confidence do you have in the public security services to protect
you, your household and your property from crime and violence? [1=Extremely
40
University of Ghana http://ugspace.ug.edu.gh
confident; 2=Somewhat confident; 3=Not very confident; 4=Not at all confident;
5=Don’t know10]
In general, do you agree with the statement ‘public security services address the
problems submitted to them in an effective manner’? [1=Strongly agree; 2=Somewhat
agree; 3=No opinion; 4=Somewhat disagree; 5=Strongly disagree; 6=Don’t know]
What level of confidence do you have in the police? [1=Very confident; 2=Confident;
3=Not confident; 4=Don’t Know]
These questions were recoded and transformed into a reverse continuous direction from
0 to 100; where 0 indicates no confidence in public security services or total
disagreement with a statement about the effectiveness of the security services. On the
other extreme, 100 represent highest confidence or total agreement with their
effectiveness. All the ‘Don’t Know’ responses were given midway values, since they
did not fit into any of the extreme values. There were less than ten non-responses, which
were all added to the third option category of ‘not confident’. The additive scale is
continuous, and ranges from 0 (No trust in the public security services at all) to 100
(Complete trust in public security services).
The second inhibitor to fear is social cohesion. This is expressed in terms of trust for
neighbours and community members. A six-item additive scale was developed for this
social cohesion (Alpha=0.904). Table 3 in Appendix A displays the Item-Total
statistics. Questions about trust for family or household members were excluded from
this scale, since that would bias the scale. The six selected variables asked whether
respondents trust people from their neighbourhood/village, ethnic group members, nonethnic group members, people from other religions, people who belong to the same
group/organization, and business owners in the community with whom they interact
regularly. There were four response options for each of them (1=All the time; 2=Most
of the time; 3=Sometimes; 4=Not at all]. There were some few non-responses, less than
ten, which were all added to the last category of ‘not at all’. The non-response in each
category was less than six, which were each added to the fourth category of ‘not at all’.
A scale from 0 to 100 was created for these six variables, after reversing the direction
The ‘Don’t Know’ options were recoded as system missing or no-responses, since they could not fit anywhere in
the continuum of the scale.
10
41
University of Ghana http://ugspace.ug.edu.gh
of the ordinal responses. It ranged from ‘No trust in neighbours at all’ (0) to ‘Constant,
unreserved trust in neighbours’ (100). Table 3 of Appendix A presents the Item-Total
Statistics.
The third inhibitor to fear is people’s perception of government’s responsiveness to
their needs, especially regarding policy implementation. This variable inquired: ‘To
what extent does the government take into account concerns voiced by you, your
household or people like you when making changes in rules, laws and policies affecting
your business/activities?’. The original question had five ranked responses, ranging
from ‘Always’ to ‘Never’. A dummy variable was created to combine all the variance
of government accountability to ‘Yes, the government takes my concerns into account’
(coded as 1); and ’No, government does not take my concerns into account’. The ‘Don’t
Know’ responses were counted as negative response, based on the assumption that if
the government actually took the concerns of the respondents into account, they would
have had some interactions already. The lack of knowledge was assumed to mean one
of two things: either the government did not engage with them, or the respondents did
not bother to engage with the government. Their concerns were therefore not factored
into government policies and legislations. There were seven non-responses and two
error entries. These nine were also added to the ‘No’ category.
The fourth of these fear inhibitors is the availability of at least one private security
gadget, service or phenomenon. An index was created for six variables that measured
the security upon which respondents relied upon to protect themselves and their
households. These are special door locks, special window/door grilles, neighbourhood
watch schemes, security guard, dogs and barbed wire. A single variable was created for
all these questions, where anyone who relied on at least one of these items or services
was coded as 1. Those who did not rely on any of these items or services were coded
as 0. Thus, a dummy variable for private security index was created [1=Availability of
at least one private security service; 0=no private security measure at all].
42
University of Ghana http://ugspace.ug.edu.gh
3.2.5 Independent Variables
This section discusses the approach for measuring gender, the main independent
variable, and other explanatory variables which were controlled in the process of
analysis. These controlled variables include age, marital status, and place of residence.
3.2.6 Independent Variable: Gender
Gender is a socially constructed phenomenon which socializes people to take up and
accept certain roles and relationships models based on sex. Male and female members
of society are socialized to take up models that categorize them into men and women.
Since almost all males in Ghana are socialized to take up roles of men and almost all
females are socialized to take up roles of women, the variable on ‘Sex’ was recoded as
‘Gender’, measured as ‘men’ and ‘women’. A dummy variable was created [1=Men
and 0=Women].
3.2.7 Control Variables: Other Socio-Demographic Variables
Some socio-demographic factors which are expected to affect perception of personal
security were used as control variables. These include age, marital status, education,
religious affiliation and employment status. Geographic factors like place of residence,
ecological zone and region of residence could also affect perceptions of personal
security.
Age: Age was measured as ordinal variable, and re-categorized as less than 20, 20 to
44, 45 to 64 and 65+ years. This categorization was done to find out how the youth,
active adult cohort, older adult cohort and elderly or retired cohorts perceive their
personal security. The 65+ year category was used as the reference category.
Spousal co-residence for those in sexual unions: Those who are married or in
consensual unions and co-reside with spouses were separated from all other persons
who did not reside in the same household with spouses. Couples who reside in the same
household are expected to feel more secured than those who do not co-reside. A dummy
variable was also created for this variable [1=Stay together with spouse in same
household; 0=Do not stay with spouse in same household]. Those who did not stay in
same household with spouse also included everybody else who was not married or in
regular sexual union. Some 49 non-responses were added to the never married cohort.
43
University of Ghana http://ugspace.ug.edu.gh
Residence: Dummy variable was created for place of residence (1=Urban; 0=Rural).
Region of residence: The ten regions of Ghana was recoded into four categories
[1=Natural resource endowed regions; 2=Dry savannah, resource scarce regions;
3=Greater Accra Metropolitan Area (GAMA); and 4=Other Regions]. Three of the ten
regions of Ghana explore natural resources in commercial quantities. These are the
Asante Region (Gold), Eastern Region (Diamond and Timber), and Western Region
(Crude Oil and Timber). These three regions were categorised as Natural Resource
Endowed regions. This category was used as the reference category. The Dry savannah
regions comprised of the Northern, Upper East and Upper West regions. These are
mostly deprived areas, with only one season of rainfall throughout the year, unlike the
other regions in the south that has two rainy seasons each year. They form a dry
savannah belt, and are mostly hard pressed for both renewable and non-renewable
natural resources. These are resource scarce regions; and are also replete with sporadic
outbreaks of chieftaincy and ethnic conflicts. The Greater Accra Region was
categorized as one category, due to its unique urban and political position in national
affairs. The three other regions – Central, Volta and Brong Ahafo – were categorized
as ‘Other regions’.
Household size: Household size was recoded into five ordinal categories: one member,
two members, three members, four members, and five or more members of the
household. The first category was used as the reference category. This category is
expected to shed light on how household size influence perception of personal security.
Educational level: Educational level was measured as the highest level of formal
schooling completed by the respondents at the time of the interview. This was grouped
into four categories [1=No education; 2=Primary education; 3=Secondary education;
4=Post-secondary education]. The recoding was done in such a way that only those who
completed primary 6 were categorized as ‘primary education’. Similarly, all those who
did not complete Junior Secondary School (JHS), Middle school or Senior Secondary
school (SHS) were placed in the category of primary school. Only those who finished
school at JHS 3 or JHS 4, SHS 3, Middle school stage 4, Secondary school stage 5 or
above and other vocational and technical schools were placed in the category of
‘Secondary education’. Primary education was used as the reference category. This was
44
University of Ghana http://ugspace.ug.edu.gh
a filter question for those who ever attended school. Therefore, all the non-responses
were treated as those with no education at all.
Employment status: The question for measuring employment status asked about any
type of work at all engaged by the respondent in the previous days that had fetched
some money, profit or pay to him or her. The question is: ‘Did [you] do any work for
pay, profit, and family gain; or did [you] produce anything for barter or home use during
the last 7 days even if it was for only one hour?’ A dummy variable was created for this
variable (1=Yes; 0=No).
Religion: Religion was recoded into four categories: 1=Catholics/protestants; 2=Islam;
3=Pentecostal/Charismatic; 4=Other. Religion could be an important determinant of
perception of personal security due to the prominence of religion in people’s life in
Africa, and the varying support systems available to adherents of the different religious
groups.
Appendix C has a summary of all measurements outlined in this section.
3.3 Methods of Analysis
The SPSS is the statistical software used for all the analysis. Various levels of analysis
were used: univariate, bivariate and multivariate.
Univariate analysis is the first stage, where variations in all the variables were explored.
Histograms have been used to display variation in all the continuous variables.
Frequency tables have been used to express variations in all the variables, whilst bar
charts and pie charts are used to display variation in some of the ordinal and nominal
variables.
The second stage is the bivariate analyses, which explored interconnections and
relationships between the variables in pairs. These are relationships mainly between the
independent, intermediate and or dependent variables, through the use of IndependentSample T-Tests, Analysis of Variance (ANOVA), and Chi-Square tests.
Independent-Sample T-Tests are expected to assess relationships between the
dichotomous and continuous variables. Before the Independent-Sample T-Test was
used, some few assumptions proposed by Field (2009) and Laerd Statistics online were
45
University of Ghana http://ugspace.ug.edu.gh
explored. First, only independent variables that are dichotomous and dependent
variables that are continuous were used for this test. Second, the GLSS data ensured
that there is independence of observation between groups in the sample. Gender, for
instance, passed all the two assumptions which are applicable, since it is dichotomous,
and none of those who responded to the questionnaire as men also responded as women.
Third, there were no significant outliers with the continuous variables, and box plots
have been used to verify that this assumption was not violated. Fourth, each variable
which is treated as outcome variable should be approximately normally distributed for
each category of the explanatory variable. The Kolmogorov-Smirnov test of normality
has been used to verify this for all the applicable bivariate analyses, since we have large
sample size more than 2000 people.11 Finally, homogeneity of variances was ensured,
through the use of the Levene’s test for homogeneity of variances. Only the trust in
neighbours scale met these assumptions when it was paired with gender.
The Independent-Sample T-Test could not be used to test associations between the main
independent and dependent variables – gender and perception of personal security. This
is because the normality assumption could not be met. The normality assumption could
not be met because the Kolmogorov-Smirnov test showed a p-value of 0.000, signifying
that there is significant deviation from the mean. The perception scale did not, therefore,
meet the normality criteria. Leard Statistics online and Field (2009) suggest that
Wilcoxon signed-ranked Test and the Mann-Whitney U-Test is a preferred test for
comparing mean under such circumstances, since they are non-parametric tests, and do
not require as many assumptions as the parametric ones. According to Field (2009:
540), ‘these tests are the non-parametric equivalent of the independent t-test’. These
non-parametric tests rely on ranks within the data to run the tests, not on the actual
values. Since the Mann-Whitney U-Test does not require any normality, it was used to
test association between gender and perceptions of personal security. The mean rank
was used to report this output, since the box plot showed that there are differences in
the shapes of the distribution between men and women (See Figure 5 of Appendix B).
11
The Shapiro-Wilk test of normality could have been used as well, if we had a sample size of less than 2000
46
University of Ghana http://ugspace.ug.edu.gh
For similar reasons, the association between gender and confidence in public security
services was also tested with Mann-Whitney U-Test. The median was used to report
the results, since the box plot showed that there are similarities between the shapes of
the distribution for men and women (See Figure 3 of Appendix B).
The Mann-Whitney U-Test was used for testing association between three of the control
variables (which were dichotomous) and perception of personal security. These are coresidence with spouse, economic activity and type of place of residence. This was due
to violation of non-significant outliers, normality and homogeneity of variance
assumptions, for which reason the Independent-Sample T-Test could not be used.
ANOVA was also used to test association between independent variables which are
nominal and dependent variables which are continuous. All the assumptions for the
Independent-Sample T-Test are applicable here, except for the fact that independent
variables in the case of ANOVA can be nominal or ordinal, not necessarily
dichotomous (Field, 2009; Laerd Statistics online). All the assumptions were met for
ANOVA as well. These variables whose associations were tested with ANOVA include
religion and perceptions of personal security, and marital status and perception of
personal security. The assumptions of normality and homogeneity were violated, since
perception of personal security is not normally distributed. However, some scholars
have proposed that F is robust against non-normal and violation of homogeneity if the
sample sizes are similar, and if the means are similar (Field, 2009; Wilcox, 2005; Glass,
1966).
Chi-square tests were done for all the dependent and independent variables which are
dichotomous. These include gender and experience of robbery, and gender and
experience of attacks or threats.
The final stage of analysis was the multivariate stage, where all the variables were put
together into models, to ascertain the extent of their association with perception of
personal security. Four linear regression models were developed. The first model was
a multiple linear regression that looked at the relationship between gender alone and
the perception of personal security. The second model introduced the fear facilitators.
The control variables were introduced in the third model; whilst the final model
47
University of Ghana http://ugspace.ug.edu.gh
included all the predictor variables – gender, control variables, fear facilitators, and fear
inhibitors. Assumptions of normality and homogeneity were assumed due to the large
sample size of the data (Field, 2009).
3.4 Data Limitations
Some important characteristics like education, marital status, employment status and
religious affiliations were not collected as part of the Governance, Peace and Security
module, which made it difficult to make broader analyses with demographic variables.
More expert statistical computations were needed to merge variables from the other
modules in the GLSS 6.
Although household heads were expected to be the main respondents, the multiple
visitations for compiling the data for the Living Standards Surveys has meant that
respondents changed hands some few times. In some instances, household members
other than household heads responded to the questions. This could potentially bias the
results. Moreover, the respondents provided responses for themselves to a large extent,
which means that their perceptions of personal security cannot represent the households
per sey; but their individual perspectives.
48
University of Ghana http://ugspace.ug.edu.gh
CHAPTER FOUR
UNIVARIATE ANALYSES
4.1 About Ghana (Profile of Ghana)
Ghana is one of the most politically stable countries in sub-Saharan Africa. Since 1992,
the country has had six free and fair democratic elections, and has had three different
governments ascending the reins of power through peaceful transitions. The latest
Ibrahim Index of African Governance ranks Ghana as 7th in overall best practices in
Governance in Africa. It also scored highly in rule of law and participation indicators.
The population of Ghana is 25,658,823 (GSS, 2010). The male population is about 49
percent of the total population, whilst female population is about 51 percent. There are
ten administrative regions of Ghana and 170 districts across the country. Ghana
discovered crude oil in commercial quantities in 2007, which became part of its list of
other natural resource commodities that support its national budget – gold, diamond,
cocoa, timber. Revenue from gold, cocoa and timber especially form a substantial
chunk of the country’s Gross Domestic Product.
The national average of households headed by males is 69.5 percent, compared to 30.5
percent headed by female (GSS, 2014: 190).
The gender inequality gap in the country has been consistently narrowed over the years,
with Ghana being the only country in West Africa to attain optimum Gender Parity
Index (GPI) at primary school, around 0.97. GPI at higher levels of education has also
increased, presently stagnating at 0.92, much higher than the sub-Saharan African
average. Gender equality in labour force participation is still low. The employment to
population ratio in the country is about 67.4 percent; 69.0 percent for males and 65.6
percent for females. Majority of Ghanaians work in the private informal sector (86.2
percent); 91 percent of female and 81 percent of male are employed in this sector. The
proportion of male who work in the government and private formal sector is twice the
proportion of female. Unemployment rate is higher in urban areas (18.2 percent) than
in rural areas (8.0 percent) due to increasing drift in rural-urban migration without
corresponding expansion of industry and social facilities.
49
University of Ghana http://ugspace.ug.edu.gh
4.2 Socio-Demographic Characteristics of Respondents
The unit of analysis is the household heads in each of the 18,000 households surveyed
for the Ghana Living Standards Survey. The total number of valid responses for this
study was 15,756, representing 87.53 percent of total surveyed household heads. It is
this total number of respondents who responded to the set of questions used for the
Perception scale. Table 4.1 shows the summary of their characteristics.
Table 4. 1 Demographic and Other Socio-Geographic Characteristics
Characteristics
Gender
Categories
Women
Men
Spousal Co-residence
Do not stay with spouse
Stay with spouse in same HH
Age group of respondents
<20 yrs
20 - 44 yrs
45 - 64 yrs
65+ year
Household size
1 member
2 members
3 members
4 members
5+ members
Religion
Catholics/Protestants
Islam
Pentecostal/Charismatic
Others
Completed
Level
of No Education
Education
Primary
Secondary/ Vocational/ Technical
Tertiary/ Post-Secondary
Economic activity in last 7 No economic activity
days
Engaged in economic activity
Type of Place of Residence Rural
Urban
Region of Residence
Natural resource endowed regions
Dry savannah, resource scarce reg
Greater Accra Region
Other regions
Total
[Source: Generated from Ghana Living Standards 2012/2013]
Frequency
8353
7403
8069
7687
841
9565
3686
1664
2671
1837
2218
2365
6665
4743
3901
4668
2444
6567
2039
6235
915
3560
12196
8801
6955
5040
4419
1806
4491
15756
Percent
53.0
47.0
51.2
48.8
5.3
60.7
23.4
10.6
17.0
11.6
14.1
15.0
42.3
30.1
24.8
29.6
15.5
41.7
12.9
39.6
5.8
22.6
77.4
55.9
44.1
32.0
28.0
11.5
28.5
100
Out of the total number of respondents, 53 percent were women and 47 percent were
men. This indicates a Sex Ratio (SR) of 88.5, a little less than the national average of
95.2 (GSS, 2013: 59). This proportion is not consistent with the fact that there are many
male household heads than women household heads in Ghana. This disparity is because
many of the persons who actually took part in the survey were female household
50
University of Ghana http://ugspace.ug.edu.gh
members – spouses, divorced, separated or widowed – who were available at home
during interview visitations. It probably indicates that women are more confined to the
homes than men in many parts of Ghana. A closer look at the marital status question
provides some answers to this dilemma.
More than half the respondents were married, followed by the never married cohort
(21.5 percent) and those who were previously married but were separated, divorced or
widowed (18.4 percent). Those in consensual unions were the least of the cohorts,
forming 8.5 percent. Altogether, those who were married and cohabiting with their
spouses were 48.8 percent of the total number of respondents. The rest (51.2 percent)
were either staying without a spouse because they were not currently in marriage (or
cohabiting).
Majority of respondents (60.7 percent) were in the active working age group, from 20
years to 44 years. This was followed by those in the older-working and early-retired
classes, 45 years to 64 years (23.4 percent). Adolescents were the least represented,
being 5.3 percent of total respondents. This lower adolescent cohort is understandable,
since many household heads in Ghana are expected to be older persons with families
and properties of their own. Adolescents can represent the household in rare cases,
where both parents are deceased or in sojourn, and where there are no other adults
present. In some cultures, women delegate adolescent males to represent the household
where the older male head of the household is not available.
Household size is big in many households across the survey areas. Some 42.3 percent
of households had more than 5 members. This was followed by single-member
households (15 percent). Four-member households and three-member households
followed closely (15.0 and 14.1 percent respectively), and then two-member
households (11.7 percent). This is consistent with existing literature, which shows that
households in sub-Saharan Africa are big due to persistence of extended family
systems.
Orthodox Christians and Pentecostals were in the majority, forming three out of five of
total respondents. Adherents of Islam were about a quarter of the total number of
respondents.
51
University of Ghana http://ugspace.ug.edu.gh
Many of the respondents had “no education” (41.7 percent), meaning that they either
never entered the classroom, or did not complete schooling up to class six. Those who
completed primary school but did not complete Junior Secondary School (JSS/JHS) or
Middle School were also classified as completing only primary school (12.9 percent).
Those who completed Junior Secondary School and above but not beyond Senior
Secondary School (SHS/SSS) or other second cycle vocational and technical
institutions were classified as completing secondary school, and these formed 39.6
percent of respondents. Respondents who had tertiary or post-secondary education like
nursing trainees, teacher trainees and polytechnic graduates were the least of the total
number (5.8 percent).
More than three-quarters of respondents were engaged in some kind of economic
activity. In the seven days preceding the days of the interviews, only 22.6 percent of
respondents had not engaged in any economic activity that was expected to yield some
profit or monetary gains. The rest engaged in economic activities.
In terms of geographic location, about six out of ten respondents resided in rural areas,
with the remaining four out of ten residing in urban areas. For the regional distributions
category, residents from natural resource-rich regions were more (32 percent), followed
closely by those from other regions (28.5 percent), and then those from dry savannah,
resource-scarce regions (28 percent). Respondents residing in Greater Accra were about
a tenth of the total number. The resource-rich regions comprised of the Asante Region
that is endowed with Gold; the Western Region that is endowed with crude oil and
timber; and the Eastern Region that is endowed with Diamond and timber. The dry
savannah regions included the Northern Region, Upper East and Upper West regions.
These have dry climatic conditions in most part of the year, and have scarce renewable
and non-renewable natural resources like forests and water bodies. The other regions
comprised of the Brong Ahafo Region, Volta Region and Central Region. The Greater
Accra Region is treated separately, due to its distinct urban, industrial and demographic
characteristics. It is the capital city of Ghana, the most densely populated, and the most
industrialized region, according to the 2010 Population and Housing Census.
52
University of Ghana http://ugspace.ug.edu.gh
4.3 Fear Facilitators
Three fear facilitators are outlined in the conceptual framework. These are people’s
perception of tension in the community, their experience of robbery, and their
experience of attacks or frightening threats. These are all categorical variables, and
Table 4.2 displays the frequencies of responses.
Tension or conflict in community was used as proxy for measuring social disorder.
Asked whether they thought that their communities have manifestations of social
disorder, 16.5 percent said that their communities were plagued with conflicts or that
there was tension between different groups of people. The rest (85.5 percent) said there
was no tension or conflict at all in their communities.
Experience of robbery is a direct manifestation of social disorder on individuals or their
households. More than a third of respondents (36.5 percent) had been victims of robbery
in the five years that preceded the survey. Such robbery incidence occurred to them or
their household members. Majority of respondents (63.5 percent), had never been
robbed or had their property stolen over the period.
A second direct manifestation of social disorder is experience of personal attacks or
threats that significantly frighten individuals and or members of their households. Only
about 1 out of 10 respondents said that members of their household or themselves have
been victims of such attacks and threats in the 12 months that preceded the interview.
The rest (91.1 percent) had never had such experience over the one year period.
Table 4.2: Descriptive Statistics of Respondents’ Accounts of Fear Facilitators
Fear facilitators
Tension or conflict in
neighbourhood
Experience of robbery at home
Experience of attacks or
frightening threats
Total
Categories
No
Frequency
13158
Percent
83.5
Yes
No
Yes
No
Yes
2598
10006
5750
14350
1406
15756
16.5
63.5
36.5
91.1
8.9
100
[Source: The Ghana Living Standards Survey 2012/2013]
53
University of Ghana http://ugspace.ug.edu.gh
4.4 Fear Inhibitors
Four variables were used to measure fear inhibitors. Two of them were measured as
continuous variables – confidence in public security services and trust for neighbours.
These were scales which were developed from multiple variables. The other two fear
inhibitors were dichotomous – government responsiveness and availability of any
private security measure that provides some assurance of security at home (Private
Security Index).
The statistical summary of the continuous variables are displayed in Table 4.3, which
includes the mean, the standard deviations and standard errors about the means. In
either scale, 0 represents very low confidence or trust and 100 represents very high
confidence or trust.
Table 4.3: Descriptive Statistics of Fear-Inhibitor Scales
Fear inhibitor scales
Mean
Confidence in Public 72.52
Security Services
Trust in neighbours scale 43.99
Medi
an
77.78
Min.
0.00
Max.
100.0
0
38.89 0.00
100.0
0
[Source: The Ghana Living Standards Survey 2012/2013]
Std.
Dev.
22.19
Std.
Err
0.18
23.77
0.19
Total
N
1575
6
1575
6
About half the total number of respondents had very high (77.78 score) confidence in
the public security services to protect them from crime, threats and violence. The
average confidence in the efficiency and effectiveness of the police to protect people
and to deal with any exigency that may arise in their communities was 72.52 on the
scale from 0 to 100. The graphical distribution (histogram) of the confidence scale can
be found in Figure 1 of Appendix B. The standard deviation for the confidence in public
security services scale was 22.19 scores, with a standard error of 0.18 scores about the
mean. A modified box plot and stem-and-leaf plot showed that all observations below
11 percent were outliers or extreme cases (See Appendix B). These are 298 cases,
representing 1.9 percent of total number of respondents. These outliers have been
maintained for variations during the bivariate and multivariate analysis stage, to find
out whether their omission or inclusion could have any impact.
There was low level of trust for community members with whom they interact on
regular basis. About half the total number of respondents had 38.89 score of trust in
54
University of Ghana http://ugspace.ug.edu.gh
their community members. The average score for trust for community members was
44. The graphical distribution (histogram) of the trust for neighbours scale can be found
in Figure 2 of Appendix B. The standard deviation for the trust in neighbours was 23.77
scores, with standard error of 0.19 scores about the mean. A modified box plot showed
that there were no outliers for the trust for neighbours scale (See Appendix B).
Table 4.4 presents descriptive statistics of the other two fear inhibitors measured as
dichotomous variables. Almost four out of ten respondents said government takes their
concerns and that of their household members into account when formulating and
changing policies and or laws. Some 61.2 percent said their concerns are never taken
into account when government is effecting changes to existing laws, formulating new
law or implementing policies.
In terms of the availability of private security services at home, about a third of
respondents said they rely on at least one private security service at home. These include
special door locks, special window or door grilles, security guard, neighbourhood watch
schemes, dogs and barbed wires.
Table 4.4: Categorical (Dichotomous) Fear Inhibitors
Fear Inhibitors
Government responsive
to concerns
Private Security Index
Categories
No
Yes
No private security
At least one private security
gadget or service exist
Total
Frequency
9650
6106
9899
5857
Percent
61.2
38.8
62.8
37.2
15756
100
[Source: Ghana Living Standards Survey 2012/2013]
4.5 Perception of Personal Security
This is the outcome variable for the study, and provides a scale of increasing scores of
perception of personal security, from 0 to 100. Table 4.5 below provides the summary
statistics. About half the total number of respondents had perception of personal
security score of 93.75. The average score of perception of personal security among the
respondents was 82.5. This means that majority of respondents had little or no fears at
all to move about alone at night, to stay alone at home at night, or to go about their daily
activities without significant fear of compromising their safety. Figure 4.1 presents
graphical representation of the distribution with the help of a histogram. The standard
55
University of Ghana http://ugspace.ug.edu.gh
deviation from the mean was 22.29 scores, with standard error of 0.18 scores. A
modified box plot of the distribution indicates that all observations that recorded less
than 20 scores on the perception scale are outliers (See Figure 4.2). These are 185 in
number, according to a stem-and-leaf plot, representing 1.2 percent of the total number
of respondents. Different models would be run with and without these outliers to
ascertain their significance in influencing the data.
Table 4.5: Summary Statistics for the Perception of Personal Security Scale
Mean Media Min.
Max. Std.
n
Dev.
Perception Scale for Personal 82.49 93.75 0.00
100.0 22.29
Security
0
[Source: Generated from the Ghana Living Standards Survey 2012/2013]
Std.
Err.
0.18
Total
N
1575
6
Figure 4.1: Histogram Representation of the Perception of Personal Security of
Respondents.
56
University of Ghana http://ugspace.ug.edu.gh
Figure 4.2: Modified Box Plot for Perception Scale
57
University of Ghana http://ugspace.ug.edu.gh
CHAPTER FIVE
BIVARIATE ANALYSES
5.1 Introduction
In this chapter, the strength of the relationship among the variables and the others are
tested. The strength of relationship between gender (main independent variable) and
perception of personal security (outcome variable) is tested. Gender is also tested
against each of the intermediate variables. Similarly, all the control variables are tested
against the outcome variable.
5.2 Gender, Fear facilitators, Fear Inhibitors & Perception of Personal
Security
5.2.1 Gender and Perception of Personal Security
The Mann-Whitney U-Test was used to compare means of men and women (Gender)
with regards to their perception of personal security. This test is best for comparing two
independent categories (men and women) when the outcome variable is either ordinal
or continuous, but does not meet the normality assumption necessary for IndependentSample T-test (Laerd Statistics online).12 The problems of non-normality and
significant outliers have been explained earlier in chapter four. Cases were excluded
listwise, descriptive statistics and quartiles were checked, and the Mann-Whitney U
option was highlighted. Two box plots generated for the distribution did not show that
the shapes for the perception of male and female were similar (See Figure 5 of
Appendix B)13. The median was nevertheless used for reporting this test. Table 5.1
below shows the actual figures for the Mann-Whitney U test, showing the mean ranks.
The corresponding test statistics are shown in Table 5.2. The mean scores are shown in
Table 5.6 ahead.
12
Analyse>Nonparametric tests>Legacy dialogs>2 independent
scale>Grouping variable=Gender(0,1)>Mann-Whitney U-Test…
samples>Test
variables=Perception
13 Analyse>Descriptive statistics>Explore>Dependent=Perception scale; Factor list=Gender>Plots>’Factor level
together’; ‘Stem-and-leaf’; ‘histogram’; ‘normality plots with tests’…
58
University of Ghana http://ugspace.ug.edu.gh
Table 5.1: Mann-Whitney U-Test Ranks Table for Gender and Perception of Personal Security
Perception Scale for Gender
Personal Security
Women
Men
Total
N
Median
8353
7403
15756
93.75
100
Mean
Rank
7606.99
8184.86
Sum of Ranks
63541146.50
60592499.50
Table 5.2: Test Statistics for Mann-Whitney U Test.a
Tests
Mann-Whitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
a. Grouping Variable: Gender
Perception Scale for Personal Security
28650000
63540000
-8.474
0.000
Men (Median = 100) generally have higher perception of personal security than women
(Median = 93.75). This difference in perception of personal security between men and
women is statistically significant at 99 percent confidence interval, judging from the
test statistics from the Mann-Whitney U-Test above (U = 28650000, p = 0.000, z = 8.474, r = -0.068).
The size of the effect could also be computed using the equation provided by Field
(2009: 550)
r
= z/√N,
Where z = z-score provided by the Mann-Whitney test, and N=sample size.
= -8.474/√15756
= -0.068
The size of the effect is very slight.
The significant difference in mean was expected. It is consistent with both the literature
on patriarchal disadvantages of women and the fear of crime. Women are expected to
feel and express their sense of personal insecurity, especially if they perceive higher
occurrence of social disorder. Men are also generally expected to hype their sense of
security. Many research findings conclude that all things being equal, women would be
more fearful of crime and violence than men.
59
University of Ghana http://ugspace.ug.edu.gh
5.2.2 Gender and the Fear Facilitators (Perception of Social Disorder)
The descriptive statistics for the association between gender and the fear facilitators (or
social disorder stimuli) are shown in Table 5.3 below, with corresponding chi-square
statistics in Table 5.4. The perception of conflict in the neighbourhood was the same
for both men and women across the country, representing one out of six of the total
number of respondents. Men experienced attacks or frightening threats (9.2 percent)
more than women (8.7 percent) in the 12 months preceding the interview; but this
difference was not significant (p = 0.210).
Table 5.3: Gender and the Occurrence (or Perception) of Fear Facilitators among Respondents
Fear Facilitators
Category
Gender
Women
Men
Frequency Percent
Frequency Percent
Tension or conflict No
6961
83.3
6197
83.7
in neighbourhood
Yes
1392
16.7
1206
16.3
Experience
of No
5433
65.0
4573
61.8
robbery at home
Yes
2920
35.0
2830
38.2
Experience of attack No
7630
91.3
6720
90.8
or threats at home
Yes
723
8.7
683
9.2
Total
8353
100
7403
100
[Source: Generated from the Ghana Living Standards Survey 2012/2013]
It is the difference in the experience of robbery at home that was found to be significant
at 99 percent confidence interval (p = 0.01). 38.2 percent of men reported of
experiencing robbery at home in the last five years, which was significantly more than
the 35 percent of women who reported same. These findings are counterintuitive.
Table 5.4: Pearson Chi-Square Tests for Gender and the Fear Facilitators
Fear facilitators
Tension
or
neighbourhood
Gender
conflict
in Chi-square
df
Sig.
Experienced robbery at home
Chi-square
df
Sig.
Experienced attack or threats Chi-square
that frighten
df
Sig.
0.399
1
0.528
18.110
1
0.000*
1.571
1
0.210
60
University of Ghana http://ugspace.ug.edu.gh
It was expected that men would experience robbery, threats and attacks more than
women, due to their socialization. Men are socialized to be property owners, and so
should have had more items stolen than women. Moreover, men were expected to be
attacked and threatened more than women, since they are socialized to be more
adventurous, daring, outgoing and aggressive than women. It is possible nevertheless,
that the same socialization which vilifies men who express their fear publicly also
stopped them from reporting their experiences of robbery and violence. Women, on the
other hand, were expected to express more fear at home, according to the extant
literature. But this was not the case with the finding. Further multivariate analysis in
the next chapter should throw further light on these dynamics.
5.2.3 Gender and the Fear Inhibitors (Those Variables that are Continuous)
Two of the fear inhibitors are continuous variables – confidence in public security
services and trust for neighbours. The best instrument of analysis for gender
(dichotomous) and each of the continuous variables is the Independent-Samples T-Test.
However, the confidence in public security scale failed to satisfy all three assumptions
of no significant outliers, normality and homogeneity of variance (See explore outputs
in previous chapter and Table 10 below).14 Another method, the Mann-Whitney U-Test,
would be used for the confidence in public security scale later in this section.
5.2.4 Gender and Trust in Neighbours
Trust in neighbours, on the other hand, satisfied all the assumptions for an IndependentSample T-Test. The Independent-Sample T-Test or Two Sample T-Test is used to
explore whether there is significant difference between the means of two unrelated
groups. Here, we test whether there is significant difference in the extent of trust that
men and women have for their neighbours. The confidence interval for this test is 95
percent, implying an error margin of 0.05 (α-level = 0.05). The null hypothesis (Ho) is
that the means of men and women are the same:
Ho: ẋ1 = ẋ2
14
A closer look at the box plots also confirmed that observations for confidence in public security services are
actually truly non-normal.
61
University of Ghana http://ugspace.ug.edu.gh
The Alternative hypothesis (HA) is that the means are not the same.
HA: ẋ1 ≠ ẋ2
Before the Independent-Sample T-Test was conducted for gender and trust in
neighbours, some three assumptions were verified. Tests were conducted for outliers,
normality and homogeneity of variances.
The first assumption that was verified was a check for outliers. There were no outliers
for trust in neighbours, as has already been established in the previous chapter. The
second assumption that was verified was the normality of the distribution of trust for
each category of gender. Since the sample size is large, the Kolmogorov-Smirnov test
was conducted.15 Confidence interval of 95 percent was chosen; whilst ‘factor level
together’, ‘stem-and-leaf plot’ and ‘normality plots with test’ were the options checked
for the plots. The test of normality in Table 5.5 shows that the data significantly deviate
from a normal distribution, since all the p-values are less than 0.05. This means that
based on the statistics from Kolmogoro-Smirnov test of normality, Trust for neighbours
scale did not satisfy the test of normality, the same way confidence in public security
services did not satisfy the test. However, there are exceptions to the rule; especially
when the sample size is big, decisions should be made after other graphical outputs like
box plots, histograms and Q-Q plots have been analysed (Field 2009). The histogram
and box plot for trust in neighbours displayed some normality. The box plot is shown
in Figure 5.1 below.
Table 5.5: Tests for Normality of some Fear Inhibitors and Gender
Some Fear Inhibitors
Confidence in Public
Security Services
Trust in neighbours scale
Gender
Women
Men
Women
Men
a. Lilliefors Significance Correction
Kolmogorov-Smirnova
Statistic
df
0.145
8353
0.151
7403
0.130
8353
0.127
7403
Sig.
0.000
0.000
0.000
0.000
Analyze>Descriptive Statistics>Explore…>Dependent=Confidence in public security services/Trust in
neighbours; Factor list=Gender…
15
62
University of Ghana http://ugspace.ug.edu.gh
Figure 5.1: Box Plot Distribution for Gender and Trust In Neighbours
The distribution for the trust in neighbours is closer to normality, as shown in the box
plot in Figure 5.1. Field (2009) advises that when the sample size is very large, even
the slightest deviations might lead to an output of non-normality, for which reason some
discretion is needed in using the results of Komogorov-Smirnov to decide on tests of
correlation. Due to the near normality of gender and trust in neighbours, it was further
considered for the Independent-Sample T-Test.
Further test was done to verify homogeneity of variance. The Levene’s test statistics
was used.16 The null hypothesis for the Levene’s test is that the variances in different
groups are equal. In other words, the difference between the variances is zero.
Ho: S12 = S22
or
S12 – S22 = 0
The Alternative hypothesis proposes a counter argument and argues that the variances
are not equal. When the p-value is significant (p<0.05), it means that the null hypothesis
is rejected. In that case, it can be concluded that the variances are not the same, and that
there is no homogeneity of variance. If, however, the p-value is not significant (p>0.05),
16
Analyze>Descriptive statistics>Explore>Dependent=Trust in neighbours; Factor list=Gender>Plots=Factor levels
together+Stem-and-leaf+histogram+Normality plots with tests+Untransformed…
63
University of Ghana http://ugspace.ug.edu.gh
it means that the null hypothesis is accepted, with the conclusion that there is
homogeneity of variance.
When the ‘untransformed’ radio command was selected, the Levene’s test of variances
had a p-value greater than 0.05 based on mean (p = 0.948). This means that
homogeneity of variance for gender and trust in neighbours is verified (F(1, 15754) =
0.004, p>0.05, not significant). The Independent-Sample T-Test can, therefore, be done
to test the association between gender and the trust in neighbours.17 The results in
Tables 5.6 and 5.7 show that trust in neighbours is not significantly different for women
(43.65+/-23.76 scores) and men (44.38+/-23.77 scores), t (15754) = -1.928, p = 0.054.
There is no significant difference, therefore, between the trust that men and women
have for other members of the community. This is because the p-value is greater than
0.05.
Table 5.6: Distribution of Initial Perception, Confidence and Trust Scales by Gender
Variables
Description
Perception Scale for Personal Frequency
Security
Mean
Standard Deviation
Standard Error of Mean
Confidence in Public Security Frequency
Services
Mean
Standard Deviation
Standard Error of Mean
Trust in neighbours scale
Frequency
Mean
Standard Deviation
Standard Error of Mean
Gender
Women
8353
80.98
23.09
0.25
8353
72.98
21.68
0.24
8353
43.64
23.76
0.26
Men
7403
84.19
21.23
0.25
7403
72.00
22.75
0.26
7403
44.38
23.77
0.28
[Source: The Ghana Living Standards Survey 2012/2013]
17
Analyse>Compare means>Independent-Sample T-Test>Test variables=Trust in neighbours; Grouping
variable=Gender(0,1)>95%CI; Exclude cases listwise…
64
University of Ghana http://ugspace.ug.edu.gh
Table 5.7: Independent Samples Test for Gender and Trust in Neighbours
Trust in
neighbours
scale
Levene's Test
for Eq. of
Variances
F
Sig.
Equal
0.004
variances
assumed
Equal
variances not
assumed
0.948
t-test for Equality of Means
T
df
Sig.
(2tail)
Mean
Diff
Std.
Error
Differen
ce
95% CI of the
Diff
Lower Upper
1.92
8
1.92
8
1575
4
0.054
0.3793
5
1.4748
1
0.012
34
1553
0
0.054
0.7312
3
0.7312
3
0.3793
7
1.4748
3
0.012
37
A closer look at the means can also attest to this, since they are almost the same, 43.65
scores for women and 44.38 scores for men, with similar standard deviations of 23.8
scores for each. There is therefore no evidence that gender explains the phenomenon of
trust for other members of the community.
5.2.5 Gender and Confidence in Public Security Services
The Mann-Whitney U-Test was used to check for association between gender and
confidence in public security services.18 This is because the variables did not meet the
conditions of no significant outliers, normality and homogeneity of variance, as has
been explained earlier. These are not important assumptions for Mann-Whitney U-Test,
which is a non-parametric test. The box plots showed that the shapes for men and
women are similar (See Figure 3 in Appendix B). Therefore, we could compare the
medians instead of the means. The median is the best measure for the Mann-Whitney
U-Test. Since SPSS compares means by default with the Mann-Whitney U-Test, the
18
Analyse>Nonparametric tests>Legacy dialogs>2 independent samples>Test variables=Confidence in public
security services>Grouping variable=Gender (0,1)>Mann-Whitney U-Test…
65
University of Ghana http://ugspace.ug.edu.gh
‘Explore’ command was first used to reset this default from the mean to the median so
that the median would be known for reporting.19
The resulting output showed that there was no significant difference between the means
and the medians. Here, only the median is reported. Expression of confidence in public
security services for women (Mdn = 77.78 score) did not vary significantly from that
for men (Mdn = 77.78 score). According to the statistics, U = 30420000, z = -1.781, ns,
r = -0.0142. The size of the effect (r) is computed by the equation provided by Field
(2009: 550):
r
= z/√N,
Where z = z-score provided by the Mann-Whitney test, and N=sample size.
= -1.781/√15756
= -0.0142
The value of r is very slight. Thus, there is no difference at all in the extent to which
one differs from the other.
No significant difference in the expression of confidence in public security services was
recorded between men and women. In other words, both men and women have similar
level of confidence in public security services to provide them with security, and to
resolve security concerns when they arise.
5.2.6 Gender and the Other Non-Continuous Fear Inhibitors
In order to test the relationship between gender and each of the other fear inhibitors,
chi-square analyses were done. These fear inhibitors are four in number, and include
government responsiveness to private concerns, neighbourhood security schemes in
community, dogs at home, and the use of special door and window grilles. Tables 5.8
and 5.9 show the descriptive statistics and the chi-square test outputs respectively.
19
Analyse>Comparemeans>Means>Dependentlist=Confidenceinpublicsecurityservices;Independent=Gender>
Options>Transfer median into the ‘Cell Statistics:’ box and deselect mean, number of cases and standard deviation…
66
University of Ghana http://ugspace.ug.edu.gh
Table 5.8: Cross Tabulations for Gender and each of the Dichotomous Fear Inhibitors
Fear Inhibitors
Category
Gender
Women
Frequency Percent
Men
Frequency Percent
No
Yes
No
Yes
5266
3087
8015
338
63.0
37.0
96.0
4.0
4384
3019
7095
308
59.2
40.8
95.8
4.2
No
Yes
Special windows or door No
grilles
Yes
Total
6560
1793
7541
812
8353
78.5
21.5
90.3
9.7
100
5792
1611
6669
734
7403
78.2
21.8
90.1
9.9
100
(Dichotomous)
Government responsive
to concerns
Availability of
neighbourhood watch
scheme
Dogs at Home
[Source: The Ghana Living Standards Survey 2012/2013]
Only concerns for government responsiveness showed any significant difference
between men and women. The results showed that at α-level of 0.01, there was a
significant relationship between gender and perception of government responsiveness
to the needs of individuals and households. 37 percent of women and 40.8 percent of
men said government took their concerns or the concerns of their households seriously.
Table 5.9: Chi-Square Tests for Gender and each of the Dichotomous Fear Inhibitors
Fear Inhibitors (Dichotomous)
Government responsive to concerns
Gender
Chi-square
df
Sig.
Availability of neighbourhood watch Chi-square
scheme
df
Sig.
Dogs at Home
Chi-square
df
Sig.
Special windows or door grilles
Chi-square
df
Sig.
24.179
1
0.000*
0.130
1
0.719
0.203
1
0.652
0.167
1
0.683
This means that men had a more positive appraisal of government responsiveness to
their needs and the needs of their households than women. It means that women
generally have lower appraisal of government responsiveness to their needs and the
needs of their household members. This is consistent with extant literature on
patriarchy, where the interests of women are never of prime concern in many societies.
Since men also generally have higher perception of personal security, it would be
67
University of Ghana http://ugspace.ug.edu.gh
interesting to see how this indicator impacts on that perception in multivariate analysis
in the next chapter.
For all the other categorical fear inhibitors, there were no significant relationships.
There was no significant relationship between gender and availability of any of the
private security factors. The difference between men and women who live in homes
protected by neighbourhood watch schemes, dogs and special doors was not significant.
Many households were protected by dogs (22 percent for both men and women). The
use of dogs is a common practice in Ghana, not necessarily for security purposes, but
dogs are seen as pets in many households in Ghana. About a tenth of respondents rely
on special windows and door grilles; whilst some five percent live in communities with
neighbourhood watch schemes.
5.3 The Control Variables and the Perception of Personal Security
The Mann-Whitney U-Test was used to test the association between the means of three
of the control variables for the perception of personal security. The cross-tabulations
for these are shown in Table 5.10. Since the perception of personal security is not
normally distributed, as has been explained in earlier sections, all the subsequent tests
of association with dichotomous variables have been done with the Mann-Whitney UTest.20 The three continuous variables are co-residence with spouse, engagement in
economic activity, and type of place of residence. Box plots for each of them found that
their shapes were not similar. Therefore, the means are used for comparison instead of
the median.
Co-Residence: Perception of personal security was statistically significantly higher for
those who shared residence with their spouses (Mean = 83.07) than for those who did
not share residence with their spouses or were not married (Mean = 81.94). The MannWhitney U-Test, U = 3001000, z = -3.744, p<0.01. This means that co-residence has
statistically significantly higher influence on people’s feeling of safety in their homes
and neighbourhood.
20
Analyse>Nonparametric tests>2 Independent Samples…>…
68
University of Ghana http://ugspace.ug.edu.gh
Table 5.10 Descriptive Statistics of Perception of Personal Security and Dichotomous Control
Variables
Variables
Categories
Spousal co-residence
No
Yes
Economic activity in No economic activity
last 7 days
Engaged
in
economic
Type of
Residence
Place
activity
of Rural
Urban
Perception Scale
Mean
Std. Dev.
Std Err
81.94
83.07
80.76
22.53
22.03
22.6
0.25
0.25
0.38
82.99
22.18
0.20
84.47
79.99
21.48
23.05
0.23
0.28
[Source: Generated from the Ghana Living Standards Survey 2012/2013]
Economic activity: Perception of personal security was statistically significantly higher
for those who engaged in economic activities in the seven days preceding the interview
(Mean = 82.99) than for those who did not engage in any economic activity (Mean =
80.76). The Mann-Whitney U-Test, U = 20310000, z = -6.221, p<0.01. This means that
economic activity provided significant explanation for perception of personal security
among respondents.
Type of place of residence: Perception of personal security was found to be statistically
significantly higher for rural dwellers (Mean = 84.47) than for urban dweller (Mean =
79.99). The Mann-Whitney U-Test, U = 26710000, z = -14.635, p<0.01. The size of
the effect (r) is -0.117, which is quite weak, but higher than the r for the other two
control variables tested earlier. Rural dwellers, therefore, have significantly higher
perception of personal security compared to their counterparts in urban areas. This
finding is consistent with extant literature, which posits that uncontrolled urbanization
in many parts of the developing world increases anti-social activities, which makes
urban dwellers feel more insecure than rural dwellers.
Age, household size and completed level of education were compared with perception
of personal security. Spearman’s Rho (ρ) correlation method was used to measure these
associations,21 due to the large sample size (Field 2009). As a nonparametric test, the
21 Analyse>Correlates>Bivariate>Drop all relevant variables into the ‘variables:’ box….>Select Spearman and
deselect all others…
69
University of Ghana http://ugspace.ug.edu.gh
only assumption required for this test is for the independent variable to be ordinal,
whilst the dependent variable is also either ordinal or continuous. All three control
variables – age, household size and level of education – are ordinal, whilst the
perception of personal security is continuous. Cases were excluded analysis by analysis.
There would not be much difference whether cases were excluded listwise or analysis
by analysis, since there are no missing variables. Table 5.11 below presents the output
for the Spearman’s Rho correlation analysis.
Table 5.11: Spearman’s Rho Correlations Matrix
Age group of
respondents
Age group Household
of
size
respondent
s
1.000
-0.111**
Correlation
Coefficient
Sig.
(2.
0.000
tailed)
N
15756
15756
Household size
Correlation
-.111**
1.000
Coefficient
Sig.
(20.000
.
tailed)
N
15756
15756
Level of
Correlation
-0.202**
-.142**
Completed
Coefficient
Education
Sig.
(20.000
0.000
tailed)
N
15756
15756
Perception Scale Correlation
0.041**
-0.004
for Personal
Coefficient
Security
Sig.
(20.000
0.604
tailed)
N
15756
15756
**. Correlation is significant at the 0.01 level (2-tailed).
Level of
Completed
Education
-0.202**
Perception
Scale for
Personal
Security
0.041**
0.000
0.000
15756
-0.142**
15756
-0.004
0.000
0.604
15756
1.000
15756
0.006
.
0.473
15756
0.006
15756
1.000
0.473
.
15756
15756
Age was the only one among the three which had a positive significant correlation with
perception of personal security, with correlation coefficient of 0.041 (ρ = 0.041),
p<0.01. That means that at 99 percent confidence interval, age is positively correlated
to perception of personal security. Perception of personal security therefore increases
with age. Older persons are thus, expected to have higher perception of personal
security. None of the other two variables – household size (ρ = -0.004, p = 0.604) and
level of education (ρ = 0.006, p = 0.473) – had any significant correlation with
perception of personal security.
70
University of Ghana http://ugspace.ug.edu.gh
The next sets of variables to be checked for association are nominal measures. These
are region of residence and religion. Analysis of Variance (ANOVA) is the best test for
checking association between nominal and continuous variables. Table 4 of Appendix
B presents the descriptive statistics. Although all tests for normality and homogeneity
of variance with perception of personal security scores have failed, some scholars agree
that with large samples, these can be violated for ANOVA, since it is a robust. Findings
from Field (2009: 360), Wilcox (2005) and Glass (1966) suggest that when the means
of the categories are similar, normality may be violated; and that when sample sizes are
similar, homogeneity of variance assumption may be violated for these three categories.
Field (2009: 359), however, cautions that the ‘violation of the assumption of
independence is very serious’. The independence of the groups is assured in this data,
as collected by the Ghana Statistical Service.
One-way ANOVA was done for each of the variables.22 Turkey post-hoc tests were
done, together with output options of description and homogeneity of variance. It has
adequately been explained that although these tests violate the normality and
homogeneity assumptions, there are enough scholarly support for them to be carried
on. The Null hypothesis for ANOVA is based on the assumption that the means of the
groups for each categorical variable relative to perception of personal security are the
same; and the alternative hypothesis is that the two means are not the same.
Ho: ẋ1 = ẋ2 = ẋ3 = ẋ4 …. ẋn
HA: The means of the two variables are not the same
Where
ẋ1-n
are the mean scores of perception of personal security for each of the
categories of the nominal variables.
Religion: The test for association between religion and perception of personal security
showed that at least one of the means is significantly different (F(3, 15752) = 17.964,
p<0.01), thus, rejecting the null hypothesis that the means are the same. Table 5.12
below shows the ANOVA output.
22
Analyse>Compare Means>One way ANOVA>Post hoc: Turks>Options: Descriptive, homogeneity of variance
71
University of Ghana http://ugspace.ug.edu.gh
Table 5.12: ANOVA Output for Religion and Perception of Personal Security
Between
Groups
Within Groups
Total
Sum
of df
Squares
26698.307
7803528.243
7830226.551
Mean Square F
3
8899.436
15752
15755
495.399
Sig.
17.964
0.000
Post hoc test showed (see also Table 4 of Appendix A) that Catholics/protestants (83.43,
sd = 21.53, p < 0.01) had statistically significantly higher perception of personal
security
than
adherents
of
Islam
(80.36,
sd
=
22.90),
but
not
for
Pentecostals/Charismatics (82.59, sd = 22.54) and those from other religions (83.88, sd
= 22.07). Adherents of Islam, on the other hand, had statistically significantly lower
perception of perception of personal security (p < 0.01) than Catholics/Protestants,
Pentecostals/Charismatics and adherents of other religious denominations. There was
no other statistically significant difference between any of the other groups.
Region of Residence: The ANOVA test of association between region of residence and
perception of personal security showed that there is significant difference between at
least one of the means (F(3, 15752) = 89.296, p < 0.01). The following are the results
of the post-hoc test. Perception of personal security for residents of natural resource
endowed regions (86.24, sd = 21.31, p < 0.01) was statistically significantly higher than
those residing in Dry Savannah region (79.47, sd = 22.85), Greater Accra region (79.18,
sd = 21.42) and other regions (82.58, sd = 22.52). There was also statistically
significantly higher perception for those in other regions (p < 0.01) as compared with
those from Dry Savannah and the Greater Accra regions. There was, however, no
statistical significance between perception of personal security for respondents residing
in Greater Accra and those in Dry Savannah regions (see Table 5 of Appendix A for
the post-hoc test output).
5.4 Intermediate Variables and Perception of Personal Security
5.4.1 The Fear Facilitators and Perception of Personal Security
All the fear inhibitors were measured as dichotomous variables; and since there is no
normality in the distribution of the perception of personal security, the Mann-Whitney
U-Test, which is a nonparametric test, would be used for comparing their means. Before
the Mann-Whitney U-Test was performed, the explore command was used to assess the
72
University of Ghana http://ugspace.ug.edu.gh
descriptive statistics for the means. Histogram distributions showed that the shape of
the distributions of each pair of groups were similar for all the fear inhibitors. Their
descriptive statistics are displayed in Table 5.13.
Those who reported that there was tension or conflict in their community (Mean = 76.00
scores) had statistically significantly lower perception of personal security than those
who said there is no tension (Mean = 83.77 scores): U = 14070000, z = -15.184, p <
0.01, r = -0.121.23 This means that at 99 percent confidence interval, there was
significant difference between the means of those who consider that they lived in
communities with rampant cases of conflict and tension. People who live in places with
tension (perceived or real) have lower perception of personal security than those who
live in places without tension or conflict. This is consistent with extant literature.
Table 5.13: Descriptive Statistics for Fear Facilitators and Perception of Personal Security
Variables
Response
Tension or conflict in
neighbourhood
Experience of robbery
at home
Experience of attacks
or threats that frighten
Total
No
Yes
No
Yes
No
Yes
Perception Scale for Personal Security
Mean
Std Dev. Std Err of Frequency
scores
Mean
83.77
21.38
0.19
13158
76.00
25.49
0.50
2598
83.63
21.61
0.22
10006
80.51
23.31
0.31
5750
82.73
22.12
0.18
14350
80.09
23.87
0.64
1406
15756
[Source: The Ghana Living Standards Survey 2012/2013]
Those who experienced robbery at home in the past (Mean = 80.51 scores) had
statistically significantly lower perception of personal security than those who have
never experienced robbery (Mean = 83.63 scores): U = 26540000, z = -8.613, p < 0.01,
r = -0.069.24 This means that at 99 percent confidence interval, there is statistically
significant difference between persons who have ever experienced robbery at home in
the past and those who have never experienced robbery. Persons who have never
experienced robbery at home are likely to have higher perception of personal security
than those who have ever experienced robbery at home. The literature on this is
23
The size of the effect, r = z/√N
24
The size of the effect, r = z/√N
73
University of Ghana http://ugspace.ug.edu.gh
inconclusive, since some studies found that persons who have experienced personal
violence do not necessarily report higher levels of fear. Further analysis is expected to
clarify this.
Those who experienced attacks or frightening threats in the past (Mean = 80.09 scores)
had statistically significantly lower perception of personal security than those who
never experienced any such threats or attacks (Mean = 82.73): U = 9466693.50, z = 4.064, p < 0.01, r = -0.032.25 This means that at α-level of 0.01, persons who have ever
experienced frightening threats or attacks have statistically significantly lower
perception of personal security than those who have not. Those who have never
experienced these frightening threats, therefore, have higher perception of personal
security than those who have experienced any threats or attacks.
The fear facilitators have all shown some statistically significant association to
perception of personal security. The presence of any of these fear inhibitors is expected
to increase people’s fears, and to reduce perception of personal security.
5.4.2 The Fear Inhibitors
Two of the fear inhibitors were measured as continuous variables, whilst the other two
were measured as dichotomous variables. The Pearson’s correlation analysis was used
to test association between perception of personal security and the two continuous
variables – confidence in public security services and trust in neighbours. The results
showed that there is a positive association between perception of personal security and
the two other scales as shown in the Figure 5.2 below.
25
The size of the effect, r = z/√N
74
University of Ghana http://ugspace.ug.edu.gh
Figure 5.2: Line Graph of Perception of Personal Security and Confidence in Public
Security Services and Trust in Neighbours.
There was a statistically significantly positive association between perception of
personal security and confidence in public security services (r = 0.166, p < 0.01).
Although this association is weak, increased confidence in public security services is
statistically significantly associated with higher perceptions of personal security. The
coefficient of determination is low nevertheless (r2 = 0.02756). This means that
confidence in public security services explains perception of personal security by only
2.76 percent. Other variables in the model explain perception of personal security by
97.24 percent.
There was also statistically significantly positive association between perception of
personal security and trust for neighbours (r = 0.172, p < 0.01). This correlation is also
weak; yet, increased trust in community members is significantly associated with
increased perceptions of personal security. The coefficient of determination is low (r2
= 0.02958). This means that trust in community members explains perception of
personal security by 2.3 percent. Thus, other variables in the model explain up to 97
percent of perception of personal security.
75
University of Ghana http://ugspace.ug.edu.gh
The means of the other two fear inhibitors are displayed in Table 5.14. Since the
perception of personal security scale is not normal and does not satisfy the homogeneity
of variance assumption, as has been elaborately explained earlier, the Mann-Whitney
U-Test would be used for this bivariate analysis. This is possible because this test is
nonparametric. The means shall be used for reporting the test.
Table 5.14: Table for Some Fear Inhibitors (Dichotomous) and Perception of Personal Security
Some Fear Inhibitors
Categories
(dichotomous)
Government responsive to No
concerns
Yes
Private security at home
No private security
gadget at home
At
least
one
private
security
gadget or service
at home
Total
Perception Scale for Personal Security
Mean
Std
Std Err Frequency
Dev.
of Mean
80.97
84.89
83.61
22.66
21.48
21.95
0.23
0.27
0.22
9650
6106
9899
80.59
22.75
0.30
5857
15756
[Source: The Ghana Living Standards Survey 2012/2013]
The perception of personal security for those who said government was responsive to
their concerns (Mean = 84.89) was statistically significantly higher than those who said
government was not responsive to their concerns (Mean = 80.97): U = 26500000, z = 11.340, p < 0.01, r = -0.090.26 This means that at α-level of 0.01, those who perceived
the government to be responsive to their needs also had positively higher perception of
personal security than those who did not perceive the government as responsive to their
needs. This was an expected outcome as a fear inhibitor.
The perception of personal security of those who had at least one private security
enhancer at home (Mean = 80.59) was statistically significantly lower than those who
had no private security enhancer (Mean = 83.61): U= 26560000, z = -9.370, p < 0.01, r
= -0.075. That means that contrary to expectation, those who had at least one enhancer
of private security at home had statistically significantly lower perception of personal
security than those who had none. This could be attributable to the fact that people who
26
r is the size of the effect of the association, which is computed as z/√N
76
University of Ghana http://ugspace.ug.edu.gh
decide to adopt private security measures may already be living in very insecure
communities, where crime rate and violence may be rampant. There is also the
possibility of mistrust that results in societies where there are high walls and extreme
protection. The extant literature supports this self-fulfilling effect of mistrust that results
when people try to build walls and fences around themselves.
Further rigorous regression analyses have been done in the next chapter, to explore the
variations through which the predictors explain the outcome.
77
University of Ghana http://ugspace.ug.edu.gh
CHAPTER SIX
MULTIVARIATE ANALYSES
6.1 Introduction
This chapter explores a number of relationships. The previous chapter explored in pairs
the relationship between individual predictors, the individual intermediate variables and
the outcome variable. However, these predictors alone could not significantly explain
the outcome variable (perception of personal security), due to the effect of known and
unknown confounding variables. Multiple regression analyses have been used to help
provide explanations for some of the unexplained variations in the outcome variable. A
statistically significant association has been found between gender and perception of
personal security through the use of the Mann Whitney U-Test; in which men (median
= 100) were found to have higher perception of personal security than women (median
= 93.75) (U = 28650000, p < 0.01, z =-8.474, r = 0.068). Although significant, gender
explains perception of personal security only slightly, judging from the r value in the
Mann-Whitney U-Test. Four regression models were run in this chapter. Model 1
included only gender and perception of personal security. The difference in perception
of personal security between men and women was found to be significant in this model
(β = 3.204, p < 0.01). This has been explained further in subsequent sections.27
Since there is inconclusive evidence in the literature about the impact of personal
experience of violence on the expression of fear, which also affects the perception of
personal security, the fear facilitators were added to the regression model to create a
second model (Model 2). The second model explored the association between gender
and perception of personal security in the midst of the fear facilitators. The variation in
the perception of personal security for men increased slightly higher than that for
women (β = 3.280, p < 0.01), indicating that men’s perception of personal security
improved slightly in the midst of fear facilitators. This could also mean that women
Note that men were coded as 1 and women as 0, meaning that women are the reference category. The β score
available is the difference in score from the mean for men.
27
78
University of Ghana http://ugspace.ug.edu.gh
were slightly negatively affected by fear facilitators than men; leading to a drop in their
perception of personal security compared to men.
The third model introduced the control variables into the previous model. The control
variables are mainly socio-demographic characteristics of respondents. This third
model explored the association between gender and the perception of personal security
in the midst of the fear facilitators and socio-demographic characteristics. This reduced
the difference in perception of personal security for men, although this difference was
still significant (β = 2.954, p < 0.01). The individual demographic characteristics did
not, thus, improve the efficiency of the model in predicting perceptions of personal
security.
The fear inhibitors were introduced into the regression model to create a fourth model
for this study. These variables reduced the difference in perception to a large extent,
although the difference between men and women was still significant (β = 2.810, p <
0.01). The fear inhibitors, thus, improved the perception of personal security for
women, reducing that of men. This model is the overall model for the study. Before
proceeding to discuss the results in details, it is important to take a look at some of the
methods used and the assumptions that were satisfied for the running of the multiple
regression models.
6.1.1 Regression Method Used
The hierarchical or blockwise regression method was used for running the models. This
approach was used because the explanatory variables for the study were chosen based
on previous studies. Field (2009: 213) proposed that if there is sound theoretical
foundation for the choice of variables, ‘stepwise methods are best avoided’ in
preference for blockwise. Hierarchical or blockwise model allows the researcher to first
enter into the model those explanatory variables which are relevant to the outcome
variable, based on existing literature. The hierarchical or blockwise method was chosen
for this analysis based on previous studies, especially by McGarrell et al (1997), Schafer
et al (2006) and Scarborough et al (2010). Schafer et al (2006), for instance, found that
women are significantly more concerned about their personal security than men,
although men have some concern for the security of their properties. Scarborough et al
(2010) also found significant association between social disorder factors and fear of
79
University of Ghana http://ugspace.ug.edu.gh
crime. They also found no significant association between fear inhibitors and
expression of fear.
Therefore in Model 1, gender alone was first entered into the regression model; in
model 2, the fear facilitators were added; in model 3, the control variables were added;
and then in model 4, the overall model, the fear inhibitors were added.
6.2 Meeting Assumptions of Multiple Regression Analysis
Before multiple linear regression analysis could be generalized from the sample to the
population, certain assumptions have to be satisfied. The main ones are that of linearity,
normality of residuals, homoscedasticity, no multicolinearity, and no significant
outliers or high influence. These and some few others were considered at this stage of
the analysis.
The first assumption is that all the variables ought to be continuous or at least, the
dependent variable has to be continuous. The perception of personal security is
continuous, so this condition for the outcome variable was satisfied. Most of the other
variables in the framework are either continuous or dichotomous, except six of the
control variables – age, region of residence, household size, religion, ethnicity and
educational level. Dummy variables were created for each of these categorical
variables. The reference category for age was chosen as age 65 years and above. Natural
Resource endowed regions was chosen as reference category for region of residence;
single household membership was chosen as reference for household size; Islam was
chosen as reference for religion; the Akan ethnic group was chosen as reference for
ethnicity; and ‘no education’ was chosen as reference category for level of education.
The value of 1 was chosen for each of the categories and 0 for others.
Second, the assumptions of random error and homoscedasticity were checked with plots
of standardized residuals on y-axis (*ZRESID on y-axis) against standardized predicted
values of the dependent variable on the x-axis (*ZPRED on x-axis).28 The *ZRESID
are standardized differences between the observed data and the predicted values; whilst
28
These plots can be specified from the ‘plots’ option of the linear regression dialogue box.
80
University of Ghana http://ugspace.ug.edu.gh
the *ZPRED are standardized forms of the values predicted by the model. According
to Field (2009: 247), this graph
should look like a random array of dots evenly dispersed around zero. If this graph
funnels out, then the chances are that there is heteroscedasticity in the data. If there is
any sort of curve in this graph, then the chances are that the data has broken the
assumption of linearity.
A careful inspection of the standardized residual and the standardized predicted value
plots (see Figure 6 of Appendix B) found that the points indeed, converged around zero,
without spreading in any particular shape across the spectrum of the x-axis, although it
is not a very perfect scatter. The conclusion here, therefore, is that the data satisfies
both homoscedasticity (normality of error or constant variance) and linearity. The
assumption of homoscedasticity assumes that the error variances in the data are constant
for all values of x; whilst assumption of linearity assumes that the association between
the dependent and independent variables are collectively linear. Inspections of the PPplots of the individual variables against perception of personal security did not raise
any significant suspicion.
Third, the assumption of normality of residuals was checked with histogram plot of the
standardized residuals (Figure 7 of Appendix B) and the normal PP-Plot (Figure 8 of
Appendix B). Visual inspection of the both outputs gave assurance that the assumption
of normality of error has been met in the data.
Fourth, the assumption of no-multicolinearity was checked based on the criteria set up
by scholars like Field (2009), Bowerman and O’Connell (1990) and Myers (1990).
These scholars used the Variance Inflation Factor (VIF) and Tolerance value (T) to set
cut-off points for checking colinearity. There are three main rules governing the
interpretation of the VIF and T. First, Bowerman and O’Connell (1990) and Myers
(1990) proposed that the largest VIF must be less than 10 (Largest VIF < 10).
Otherwise, there should be cause for worry about colinearity among the predictors.
Second, the average VIF must be less than 1 or at least closely to 1 (Average VIF < 1
or closer to 1). Otherwise, the regression model may be biased. Field (2009) provided
a third rule about colinearity using tolerance value. Tolerance value must be greater
than 0.1 (T > 0.1). Otherwise, there must be cause for concern. Other scholars like
Menard (1995) purge the minimum acceptable T value at 0.2, meaning that T value
81
University of Ghana http://ugspace.ug.edu.gh
must be greater than 0.2 for the model to be classified as meeting the condition of noncolinearity. The average VIF is computed as follows:
Average VIF = (∑VIFi)/k
Where VIFI
= the VIF values for each explanatory variable;
K
= the number of predictors in the model
The maximum VIF for all the three models was 3.084 (for age group 20-44) in the
fourth model; and the average VIF for each of the models was found to be either one or
less than one. Moreover, none of the T values was less than 0.2. Therefore, the
assumption of no multicolinearity in the data has been met.
Fifth, outliers and influential cases were also checked. The standardized residuals
(Residuals from the model divided by their standard deviations) were used to check for
outliers. Since these standardized residuals are equivalent to z-scores, any value greater
than 3.29 standard deviations from the mean should be cause for concern because under
the normal curve, 99.9 percent of cases must fall within 3.29 standard deviations from
the mean. At 95 percent confidence interval, this value should range between +1.96 and
-1.96 deviations from the mean. All values beyond this range should be considered as
outliers. Similarly, at 99 percent confidence interval, the standardized residuals or zscores should range between +2.58 and -2.58; otherwise, they must be considered as
outliers.29 All the standardized residuals (Beta) had a value greater than 1.96 standard
deviations from the mean. The highest Beta (z=score) was -0.155, which was recorded
by Dry Savannah in Model 3. This was within the 95 percent confidence interval scores,
and calls for no outlier concerns.
The Adjusted Predicted value and the Cook’s Distance were used to assess variables
that have undue influence on the model. The Adjusted value is the value of the
dependent variable for each case when that case is excluded from the model. Therefore,
29
As a general norm, the unstandardized residuals should also not be too large. Larger residuals indicate substantial
deviation of a case from the best fit line. The unstandardized residuals, however, do not lend themselves for
comparison with other models. The standardized residuals are therefore preferred.
82
University of Ghana http://ugspace.ug.edu.gh
if a case were to be deleted from a model, its value should not be substantially different
from the originally predicted coefficient. Where there is vast difference between the
Adjusted predicted value and predicted value, it means that the particular case has
undue influence on the model, and could bias the results. The Cook’s value, on the other
hand, assesses the influence of cases on the overall model; and should not be greater
than 1. In other words, Cook’s Distance should be less than one if a case’s influence on
a model is to be tolerable (Cook’s Distance < 1). Otherwise, it means the case exerts
too much influence, and has to be investigated further. From the residual statistics of
the models, the mean Adjusted Predicted value is the same as the mean Predicted value
(82.49), which means that on the average, no case has undue influence on the models.
The maximum Cook’s Distance is 0.001. These statistics satisfy the assumption that
none of the cases have undue influence on the models.
Due to the assumption of linearity in the linear regression model, a linear relationship
is assumed to exist between all the explanatory variables and the outcome variable. The
best fit line that expresses the relationship between the explanatory variables and the
outcome variable is expressed as:
Outcomei = (Model) + Errori
Yi = b0 + b1x1 + b2x2 + b3x3 + …. + bnxn + εi
Where Yi = The outcome variable (Perception of personal security)
bi = the slope or gradient of the regression line, the strength of association with
the outcome variable
bo = the intercept which defines the outcome variable where all other predictors
are held constant
xi = individual predictors or explanatory variables
εi = the difference between the predicted and observed values of Y for the ith
respondent
This is an ordinary least square (OLS) where the best fit line with least deviations is
computed. For the overall model (Model 4), the relationship between the explanatory
83
University of Ghana http://ugspace.ug.edu.gh
variables and the perception of personal security is expressed as follows, in the order in
which they appear in the model in Table 6.3:
P4
= (gender) + (fear facilitators) + (control variables) + (fear inhibitors) + ε4
P4
= b0 + [b1(gender)] + [b2(tension) + b3(rob) + b4(attack)] +
[b5(spouse) + b6(age) + b7(HHS) + b8(rel) + b9(edu) + b10(emp) + b11(ethnic) +
b12(res) + b13(reg)] + [b13(conf) + b14(trust) + b15(gov) + b16(PSI)] + ε4
The coefficient of determination (R2) was 10.40 percent (R2 = 0.104). The next sections
explore variations in all the four models of the study.
6.3 Regression Diagnostics
The coefficient of determination (R2) of the models improved substantially from under
one percent in the first model to over ten percent in the overall model. The R2 in the
overall model is 0.104 (10.40 percent). In the first model, the F-Ratio changed from 0
to 81.50 percent, which is very significant. Gender, therefore, improved the model’s
explanatory power by 81.5 percent. The addition of the fear facilitators (Model 2)
improved the explanatory power of the model by 114.31 percent. This means that the
fear facilitators explain the perception of personal security much better than gender.
However, the third model’s efficacy in explaining the perception of personal security
reduced substantially to 29.94 percent. In other words, the control variables which were
added to the model (Model 3) did not improve its explanatory power better than the
second model. The control variables have the least effect in explaining perception of
personal security. The fear inhibitors, introduced in (Model 4), improved the
explanatory power of the model, the F-ratio, by 193.29 percent. The addition of the fear
inhibitors, therefore, improved the explanatory power of the perception of personal
security substantially. This brought the overall explanatory power of the predictors (R2)
to 10.40 percent. The changes in the F-ratio from Model 1 to Model 4 are each
significant at p < 0.01 (See Table 6.1).
84
University of Ghana http://ugspace.ug.edu.gh
Table 6.1: Model Summarye of the Regression Models
Model
R
R2
Adj R2
Std. Err
Change Statistics
Durbin-
of the
Estimate
Watson
R2
F
Change
Change
df1
df2
Sig. F
Change
(%)
e
1
0.072
0.005
0.005
22.23675
0.005
81.503
1
15754
0.000
2
0.162
0.026
0.026
22.00065
0.021
114.314
3
15751
0.000
3
0.246
0.060
0.059
21.62635
0..034
29.944
19
15732
0.000
4
0.323
0.104
0.103
21.11630
0.044
193.292
4
15728
0.000
1.366
Dependent variable: Perception of Personal Security
In terms of the generalisability of the model, the Adjusted R2 determines how well the
model generalizes the sample to the population. Ideally, Adjusted R2 should be the same
or closer to the R2. The difference between the R2 and Adjusted R2 tells the shrinkage
of the model, implying that if the model was transposed into the population, there would
be approximately (R2 – Adjusted R2) percent variance in the outcome variable. The
difference in the R2 and Adjusted R2 for each of the models is very slight; implying that
we can generalize the findings to the population, since there would be very little
variance in the scores of perception of personal security.
The next important indicator is the Durbin-Watson statistic. This helps check for the
independence of error of assumption in the regression. This assumption is satisfied if
the Durbin-Watson statistic is closer to 2 (but between 1 and 3). In the model, the
Durbin-Watson statistic is 1.366. This is indicative that the independence of error
assumption has been met in the data.
The ANOVA test also helps test whether the model predicts the outcome variable better
than simply using the mean to make conjectures. The predictors in each of the four
models made significant prediction of the outcome variable, the perception of personal
security (Table 6.2). For each of the models, the predictors provided significant
explanation of the perception of personal security at α-level of 0.01 (p<0.01). As has
already been alluded to in previous paragraphs, the second model explained the
85
University of Ghana http://ugspace.ug.edu.gh
outcome variable better. In other words, gender and the fear facilitators have the
greatest explanatory power (F(4, 15751) = 106.551, p<0.01).
Table 6.2: ANOVAe Test for the Model
Model
Sum of
df
Squares
1
2
3
4
Mean
F
Sig.
Square
Regression
40300.966
1
40300.966
Residual
7789925.585
15754
494.473
Total
7830226.551
15755
Regression
206294.673
4
51573.668
Residual
7623931.878
15751
484.028
Total
7830226.551
15755
Regression
472388.449
23
20538.628
Residual
7357838.102
15732
467.699
Total
7830226.551
15755
Regression
817142.711
27
30264.545
Residual
7013083.840
15728
445.898
Total
7830226.551
15755
81.503
0.000a
106.551
0.000b
43.914
0.000c
67.873
0.000d
e. Dependent Variable: Perception Scale for Personal Security
The third model provided the poorest explanation of the outcome variable (F(23, 15732)
= 43.914, p<0.01), although its impact is still significant. This model included gender,
fear facilitators and control variables. In other words, the introduction of the control
variables did not contribute much to the explanatory power of the model. In fact, it
reduced its efficacy from 114.31 percent to 29.944 percent as shown in Table 6.1.
The introduction of the fear inhibitors improved the model significantly (Model 4),
changing the F-ratio from 29.94 percent to 193.29 percent (F(27, 15728) = 67.87, p <
0.01). The explanatory power of all the variables in the model, therefore, provided
better explanation for the perception of personal security than could be done by the use
of means.
Overall, gender, the fear facilitators and the fear inhibitors provided the best
explanation for perception of personal security. The other socio-demographic variables
have the least explanatory power for the perception of personal security. The overall
model predicts the perception of personal security by 10.40 percent, whilst other
86
University of Ghana http://ugspace.ug.edu.gh
predictors that are not accounted for in the model predict the perception of personal
security by 89.60 percent.
6.4 The Results and Checking of Hypotheses
6.4.1 Gender
Gender was found to be a significant predictor of perception of personal security in all
the four models (see Table 6.3). Men had significantly higher perception of personal
security in the base model than women (β = 3.204, p < 0.01). In other words, men had
about three scores higher in perception of personal security than women. This
difference increased slightly to 3.280 scores in Model 2, after the addition of the fear
facilitators. The control variables, however, reduced this score in Model 3 (β = 2.954,
p<0.01); whilst the addition of the fear inhibitors in the overall model reduced this
further to β = 2.810, p<0.01. Although the difference in perception of personal security
between men and women remained significant in all four models at α-level of 0.05, the
addition of the fear inhibitors had the greatest impact, based on the wider difference
between the base model and the overall model, which is (3.204 – 2.810), 0.394. All
four models confirmed the first hypothesis, that women are less likely to feel secure
than men in Ghana; implying that the perception of personal security of women in
Ghana is lower than the perception of personal security of men. Women’s perception
of personal security is negatively affected by fear facilitators, but positively affected by
fear inhibitors. It is not clear from the results, whether the change in perception for
women leads to corresponding inverse effect on perceptions for men. This could be the
subject of further investigations.
6.4.2 The Fear Facilitators (Social Disorder Factors)
The fear facilitators were introduced into the second model. These variables
collectively had the greatest explanatory power on the model (F (4, 15751) = 106.551,
p<0.01). As was expected, all the three fear facilitators had significant inverse effect on
perception of personal security. Perception of tension or conflict in the community had
the greatest negative effect on perception of personal security (β = -7.532, p<0.01),
followed by experience of robbery in the five years preceding the interview (β = -2.929,
p<0.01), and then by experience of frightening attacks or threats in the one year
preceding the interview (β = -1.400, p<0.05). Proportionate variation in the fear
87
University of Ghana http://ugspace.ug.edu.gh
facilitators were found in the third and fourth models. The only difference in the
subsequent models was that variation in the experience of attack or threat was
significant at α-level of 0.05 in the second model, but shifted to be significant at α-level
of 0.01 in the third and overall models.
In the overall model, those who reported that there was conflict or tension in their
communities had 6.744 scores lower in perception of personal security than those who
did not report of such conflicts or tension. Those who reported experiencing robbery at
their homes in the last five years had 2.782 scores lower than those who did not report
of any such experiences. The score for those who were attacked or threatened in the 12
months preceding the interview was also lower than those who were never attacked (β
= -1.601, p<0.01), although obviously, this last variable had the least inverse effect on
perception of personal security compared to the other two fear facilitators.
88
University of Ghana http://ugspace.ug.edu.gh
Table 6.3: Hierarchical Regression Models Predicting Perception of Personal Security (β = Unstandardised Coeff)
MODEL 1
MODEL 2
2
VARIABLES
Constant
Main Independent Variable
Gender (Women)
Men
Fear Facilitators
Tension/conflict in neigh (No)
Yes
Robbery at home (No)
Yes
Attack or threats (No)
Yes
Control Variables
Spousal co-residence (No)
Yes
Age (65+ years)
<20 years
20-44 years
45-64 years
HH size (One member)
Two household members
Three household members
Four household members
Five household members
R = 0.005
β
S. E
80.984 0.243
3.204
0.355
MODEL 3
2
Sig.
0.000
R = 0.026
β
S. E.
83.385 0.284
0.000
3.280
MODEL 4
2
Sig.
0.000
R2 = 0.104
β
S. E
73.443 1.102
Sig.
0.000
0.371
0.000
2.810
0.363
0.000
-7.519
0.468
0.000
-6.744
0.459
0.000
0.000
-3.255
0.362
0.000
-2.782
0.355
0.000
0.024
-2.01
0.612
0.001
-1.601
0.598
0.007
1.107
0.420
0.008
0.784
0.410
0.056
-0.908
-1.644
-0.238
0.972
0.616
0.665
0.350
0.008
0.443
-0.222
-1.118
-0.272
0.95
0.603
0.650
0.815
0.064
0.676
-0.530
-0.543
-0.271
-0.917
0.691
0.679
0.680
0.593
0.443
0.424
0.690
0.122
-0.458
-0.654
-0.436
-1.123
0.675
0.663
0.664
0.580
0.498
0.324
0.324
0.053
Sig.
0.000
R = 0.060
β
S. E
89.257 0.906
0.351
0.000
2.954
-7.532
0.474
0.000
-2.929
0.367
-1.400
0.620
89
University of Ghana http://ugspace.ug.edu.gh
Table 6.3 (Cont.): Hierarchical Regression Models Predicting Perception of Personal Security (β = Unstandardised Coeff)
VARIABLES
Religion (Islam)
Catholics or protestants
Pentecostal or Charismatic
Other religions
Education (No education)
Complete primary
Complete secondary
Post-Secondary or Tertiary
Employment (No)
Yes
Residential status (Rural)
Urban
Region (Nat. Resource reg)
Dry Savannah regions
Greater Accra region
Other
Fear Inhibitors
Confidence in Pub Sec. scale
Trust in neighbours scale
Govt responsiveness (No)
Yes
Private Security Index (No)
Yes
MODEL 1
MODEL 2
MODEL 3
MODEL 4
R2 = 0.005
Β
S. E
R2 = 0.026
β
S. E.
R2 = 0.060
R2 = 0.104
Sig.
Sig.
β
S.E
Sig.
β
S. E
Sig.
0.620
0.129
0.297
0.522
0.541
0.612
0.235
0.811
0.627
1.418
1.095
1.164
0.511
0.529
0.598
0.006
0.039
0.052
1.977
1.313
-0.487
0.579
0.457
0.822
0.001
0.004
0.553
1.752
1.6170
0.651
0.566
0.447
0.805
0.002
0.000
0.419
1.072
0.435
0.014
0.950
0.426
0.026
-5.117
0.397
0.000
-4.406
0.389
0.000
-7.705
-5.962
-4.080
0.530
0.620
0.447
0.000
0.000
0.000
-6.315
-5.751
-4.590
0.521
0.605
0.438
0.000
0.000
0.000
0.122
0.117
0.008
0.007
0.000
0.000
2.998
0.350
0.000
-2.646
0.355
0.000
90
University of Ghana http://ugspace.ug.edu.gh
6.4.3 Control Variables
The control variables were introduced into the third model, and collectively had the
least explanatory power on perception of personal security (F(27, 15732) = 43.91,
p<0.01). Religion, household size and age did not have any significant explanatory
power on the model; except for the fact that persons in their active adult stages, 24 years
to 44 years had significantly lower perception of personal security than the elderly aged
65 years and above (β = -1.644, p<0.010. Those variables which had the most impact
were spousal residential status, education, type of place of residence and region of
residence. In Model 3, persons who shared residence with their spouses had
significantly higher perception of personal security than those who did not share
residence with spouses, married or not (β = 1.107, p<0.01). However, this variation was
not significant in the fourth model. The second hypothesis for the study was confirmed
at 95 percent confidence interval without the fear inhibitors. When the fear inhibitors
were introduced in Model 4, this difference in perception between those who cohabit
with spouses and those who do not cohabit was no more significant.
In terms of education, those who completed primary education and secondary education
had significantly higher perception of personal security (β = 1.752, p<0.01; β = 1.617,
p<0.01 respectively) than those who never completed any level of education. Although
persons who accessed post-secondary and tertiary education had higher perception of
personal security than those who never completed any level of education, this difference
was not significant. Difference in perception of personal security was slightly higher
for the employed than for the unemployed.
Geographical location of residence had significant impact on perception of personal
security. Those who reside in urban areas had statistically significantly lower
perception of personal security than those who reside in rural areas (β = -4.406, p<0.01).
Similarly, as compared to those who reside in natural resource endowed regions,
persons from Savannah (β = -6.315, p<0.01), Greater Accra (β = -5.751, p<0.01) and
other (β = -4.590, p<0.01) regions had statistically significantly lower perception of
personal security. This finding rejects the third hypothesis, that residents in natural
resource-endowed regions are more likely to have lower perception of personal security
than those from regions without abundant natural resources. Residents from the other
91
University of Ghana http://ugspace.ug.edu.gh
regions were expected to have higher perception of personal security than those from
the Savannah and other regions. This hypothesis was not confirmed.
6.4.4 Fear Inhibitors
The fear inhibitors were introduced into the final model. This substantially increased
the effectiveness of the F-ratio from 29.94 percent to 193.29 percent, as shown in Table
6.1. According to the output in Table 6.3, each of the fear inhibitors increased
perception of personal security, apart from the private security index.
The confidence in public security services had a statistically significantly positive
association with perception of personal security (β = 0.122, p<0.01). The confidence in
public security services was measured as a continuous variable. Therefore, for any unit
change in confidence in public security services, perception of personal security
increased by a score of 0.122 if all other variables were held constant. Similarly, trust
in neighbours was measured as a continuous scale. The expression of trust in
neighbours had a statistically significantly positive association with perception of
personal security (β = 0.117, p<0.01). For any unit change in trust for neighbours,
perception of personal security increased by 0.118 scores if all other variables were
held constant.
Perception of government responsiveness to the concerns of people also had
statistically significant positive association to perception of personal security (β =
2.998, p<0.01). Respondents who thought that government took their concerns into
account in policy and law making had significantly higher perception of personal
security than those who thought otherwise.
The private security index, on the other hand, had a statistically significant negative
association with perception of personal security (β = -2.646, p<0.01). Those who had
at least one gadget, property or service available that provided some assurance of
protection of household had a lower perception of personal security. This is probably
because persons who live in communities with prevalence of violence are those who
are more concerned about protection for their homes.
92
University of Ghana http://ugspace.ug.edu.gh
6.5 Discussion
Overall, the results presented in the models are consistent with existing literature, with
a few exceptions. The study has confirmed that gender is an important demographic
predictor of perception of personal security. Women are generally thought to have
lower perception of personal security or higher fear of crime and violence than men
(Garafalo, 1981, McGarrell et al., 1997, Schafer et al., 2006, Scarborough et al., 2010),
in spite of some few variations (Hale, 1996). The persistence of gender as a dominant
predictor of perceptions of personal security has been attributable to systemic
patriarchal inequities that place women at universal disadvantage compared to men
(Holland, 2006; Walby, 1991; Kambarami, 2006; Sultana, 2011). These patriarchal
trends are sustained by the cultures of masculinity and femininity, which see to a
socialization which boosts the confidence and self esteem of men, and undermine the
independence and safety of women. Since women are socialized through femininity to
be very sensitive of their physiological and social vulnerabilities (Scott, 2003; Goodey,
1997; Madriz, 1997; Stanko, 1995); to feel that their security needs are appendages to
that of men (O’Neil and Hartmann, 2012; Kambrami, 2006; Kameri-Mbote, 2005;
Cammack, 2000; Riger et al., 1978); and that the wellbeing of their children is
exclusively theirs (Gilchrist et al., 1998); any increase in social disorder is likely to
affect their perceptions of personal insecurity.
This is attested by the fact that the addition of the fear facilitators to the model slightly
reduced the perception of personal security of women, slightly increasing that of men.
Social disorder factors which dwell on perception (tension in neighbourhood) had the
greatest negative association with perception of personal security. This association was
much stronger than the association of personal experiences of crime and violence. The
extant literature also supports this finding, that experience of violence does not
necessarily increase fear of crime. Although this study found significantly negative
association between experiences of robbery and threats, this association was not as
strong as that between socially perceived disorders of conflict and perception of
personal security. This is consistent with the literature, that women’s fear of violence
and crime increases with increasing social disorder (Schafer et al., 2006; Warr, 1984;
McGarrell et al., 1991). The fear of sexual assault, for instance, overshadows all their
93
University of Ghana http://ugspace.ug.edu.gh
perceptions of security and safety, even when there are discussions about non-personal
and non-violent issues (Ferraro, 1995; Warr, 1984; Fisher and Sloan, 2003).
The safety of women is subject to societal manipulation, and this is confirmed by the
fact that women’s score of perception of personal security improved with the addition
of fear inhibitors to the model. Men’s perception of personal security reduced after
introducing social fear inhibitors like trust for neighbours, confidence in public security
services and perception of government responsiveness to the needs of people. Women
feel safer than men when there is social order and assurance of justice in the community.
One finding which appears to be counter-intuitive at a glance is the effect of private
security at home on the perception of personal security. It would seem that once people
have assurance of private security at home, it would complement the wider public
security that the state provides to enhance the feeling of safety. However, the model did
not confirm this. Some scholars have explained that building high walls and defences
around homes, for instance, can self-perpetuate mistrust and potential violence (Robert,
2010). Therefore, people who are on high alert for protection may create barriers for
proper social integration that may enhance their perception of safety. It is also possible
that people who are on high alert for personal protection already live in places with
higher prevalence of crime and violence. In this case, it would not be the protection
they seek which diminishes their perception of personal security, but the reverse. Their
lower perception of personal security in the midst of social disorder may have motivated
them to seek private protection in order to boost their security, not vice versa.
Residential arrangement of spouses, type of place of residence and region of residence
significantly predicted perception of personal security. These variables narrowed the
difference in perception of personal security between men and women. Men’s
perception of personal security reduced, and by implication, women’s perception
increased with the addition of these socio-demographic variables.
The literature is inconclusive regarding the occurrence of insecurity in resourceendowed and resource-poor regions. The greed versus grievance literature proposes that
exploration of natural resources in commercial quantities increases the chance of
violence, crime and political instability (Keen, 2012; Collier and Binswager, 1999;
94
University of Ghana http://ugspace.ug.edu.gh
Collier, 2000; Berdal, 2005; Ballentine and Nitzschke, 2003; Ballentine, 2003). There
is another school of thought that proposes that scarcity of renewable natural resource
also increases the occurrence of violence due to potential struggles for control over
insufficient resources (Edward et al., 2003; Giddens, 2011). The finding in this study
supports the latter school. Residents in all the other regions had lower perception of
personal security than those in natural resource endowed regions. This finding clarifies
the resource curse literature in the context of Ghana. People who live in Accra,
resource-poor regions and other regions have lower perceptions of personal security
than those who live in resource endowed regions. The resource-scarcity debate where
struggles over scarce resources lead to deterioration in security is supported by these
findings. Many parts of the resource-poor northern regions of Ghana are endemic with
ethnic and chieftaincy conflicts; coupled with the fact that urban areas like Greater
Accra have their own challenges of youth unemployment and social vices. It would be
interesting in future studies to see how gender also plays a part in predicting perception
of personal security under conditions of natural resource availability and scarcity.
The support which spousal co-residence offers to perception of personal security is
another phenomenon that is worth exploring. It would be interesting to explore further
dynamics to ascertain how gender also influences the strong association between
spousal co-residence and perception of personal security. The questions to ask in such
a discourse is, who receives the most protection and safety in marriage and coresidence, is it men or women? Such an exploration would help understand dynamics
of spousal co-residence, especially since violence against spouses is common in many
societies. Education had a positive association with perception of personal security as
well, although it is difficult to explain why post-secondary and tertiary education had a
lower perception of personal security. This could be because those who attain higher
levels of education are extremely sensitive to their environment, and easily identify the
least threats to their safety.
People in urban areas have lower perception of personal security than those in rural
areas. This supports the literature that faster pace of unplanned urbanization in
developing countries come with increasing anti-social practices that affect the security
of urban dwellers. This is linked to the phenomenon of higher unemployment rates in
95
University of Ghana http://ugspace.ug.edu.gh
urban areas in Ghana (GSS, 2013). Unemployment rate is higher in urban areas than
rural areas.
Other socio-demographic characteristics like ethnicity, religion, age and household size
were not significant predictors of perception of personal security. This is
counterintuitive because many communities in Ghana dominated by adherents of Islam,
like the Zongo communities and Northern Region of Ghana have higher prevalence of
violence. The Dagbon Crisis of 2003 in Northern Region is a classical example
(Mahama, 2009).
It is important to note that lower perceptions of security or increased fear of
victimization does not necessarily correlate with actual occurrences of victimization.
Although women may report of higher perception of personal insecurity, this fear might
motivate them to move cautiously about in relatively safe places, thereby reducing any
actual occurrence of victimisation. Men, on the other hand, are more likely to be victims
of crime and violence due to their perceived sense of safety.
96
University of Ghana http://ugspace.ug.edu.gh
CHAPTER SEVEN
SUMMARY, CONCLUSIONS AND RECOMMENDATIONS
7.1 Summary
The dissertation contextualizes the discourse on gender inequality in Ghana within the
patriarchy and fear of crime literature. It problematised the system of patriarchy and its
attendant socializations of male and female into men and women, which inculcate into
them cultures of masculinity and femininity; and which undermine the place and role
of women in society. The study argued that the gendered socializations make it
imperative for men to have higher perceptions of personal security than women. Using
the fear of crime literature as basis to test this proposition, a conceptual framework was
developed with fear facilitators and fear inhibitors as mediating variables. The
Governance, Peace and Security module of the sixth round of the Ghana Living Survey
(2012/2013) was the main source of data for analyses. A four-item perception scale
(Alpha = 0.857) was developed from questions that enquired the degree of safety of
respondents in their daily activities at home and in their neighbourhoods. This
perception scale was the dependent variable to which the other predictor variables were
tested. A total of 15756 out of 18,000 respondents answered these set of questions,
making this number the valid N for all the analyses. Frequency tables, charts and crosstabulations were used to describe the characteristics of the respondents and to test the
strength of association between the variables.
Four multiple linear regression models were run to test the strength of association
between the predictor variables and the perception of personal security of the
respondents. Gender was the only predictor variable in the first model. Fear facilitators
were introduced in the second model, and the socio-demographic variables were
introduced in the third model. In the fourth model, the final set of variables, the fear
inhibitors, were introduced. This was the overall model which included all the
predictors. Efforts were made to ensure that all the assumptions for the bivariate and
multivariate tests were met. The next section points out some major conclusions drawn
from the various analyses.
97
University of Ghana http://ugspace.ug.edu.gh
7.2 Conclusions
The findings of the study have largely confirmed what exists in the literature. Gender
remains a significant predictor of fear of crime and perception of personal security.
Social disorder factors aggravate the gender difference in perception of personal
security. Perception of personal security of women reduces in the midst of fear
facilitators like crime and social tension. When there is the occurrence of tension or
conflict (perceived or real), women’s perception of personal security reduces more than
men’s perception of personal security. Personal experiences of crime or perceived
social tension aggravate the fears. Men, on the other hand, respond less negatively to
fear facilitators, except frightening threats. The literature on patriarchy, masculinity and
femininity support these findings, as men are generally socialized to be perpetrators of
violence with a few exceptions; and women to live under the shadow of women, even
if they have to live in fear.
Fear inhibitors like greater confidence in the public security services improve the
perception of personal security of women significantly. When there is confidence that
public security services are capable of efficiently addressing problems submitted to
them; and that households and properties are well protected from crime and violence,
women tend to have higher perception of personal security. Private security gadgets,
however, have negative association with perception of personal security; implying that
people who are fixated on relying on private security gadgets and services probably live
in places with higher social indicator factors.
Demographic characteristics do not enhance the gender disparity in the perception of
personal security. Ethnicity, household size, and religious affiliations do not have any
significant impact on perception of personal security. However, residential
arrangements have a significant impact on perceptions of personal security. People who
stay with their spouses have enhanced perception of their security; and so do people
who live in rural areas and those who live in resource-rich areas. Minimal level of
education also improves people’s sense of security, although higher levels of education
probably improves people’s sensitivity to their environment, and respond to insecurity
stimuli faster than those with lower levels of education.
98
University of Ghana http://ugspace.ug.edu.gh
7.3 Recommendations
Having an optimum sense of personal security is crucial for reaping the full dividends
of social goods. Governments and policymakers must necessarily be interested in
ensuring that people feel safe in reaping the full benefits of governance.
Tackling the personal security question also tackles the gender question because
decreased perception of personal security of women also affects their ability to assert
their independence and to benefit from the constitutional provision of freedom of
movement. Women tend to depend on men for safety at odd times (at night and in
secluded areas) for their movements and access to certain social facilities. This trend
feeds on patriarchal systems and propagates cultures of femininity and masculinity.
Policymakers could incorporate demographic considerations in addressing human
security questions, since the two can hardly be separated. Socio-demographic and
geographic characteristics also influence perceptions of personal security.
Rural residence, for instance, induces higher perception of personal security; so some
characteristics of rural areas that lend themselves to enhancing people’s security should
be identified and highlighted for replication in urban areas. Higher unemployment rates
and social vices prevalent in urban areas should be treated as a security issue, although
the exact characteristics of the unemployed persons that reduce their sense of security
should be subject for further investigation.
The role of stable family arrangements in defining people’s sense of security also came
out strongly in this study. Spousal co-residence boosts perception of personal security.
It is unclear how policymakers can directly influence residential arrangements of
spouses. Affordable housing could be made available to young persons and those with
families in order to facilitate co-residence of those in sexual unions, especially for urban
dwellers.
Universal basic education should continue to be a priority on the national agenda, since
the perception for those with at least basic or secondary education is higher than those
with no education at all. Post-secondary education or tertiary education also helps to
improved perception of personal security through another pathway. People with postsecondary education become very alert of their environment, and could help identify
99
University of Ghana http://ugspace.ug.edu.gh
latent causes of disorder and insecurity. Legislation should be put in place to ensure
that education also becomes a national security tool.
Representative governance should be treated as a panacea for dealing with gender
differences in perception of personal security, and for tackling perception of personal
insecurity in general. Effective democratic governance is expected to achieve this, and
Ghana has demonstrated that she is on the right track to entrenching its democratic
credentials. The arm of the legislature which represents citizens in government should
be further empowered, and so should the local governments and assemblies. Effective
consultation of members of parliament (MPs) with their constituents should be made a
priority for all political parties and the house of parliament. Regular visitation to
constituencies and media outreach programmes that assure people that their concerns
are taken into account in national policies and legislations are important to boost
confidence of women especially. District, municipal and metropolitan assemblies
should be well resourced to have closer contact with residents in their respective
catchment areas.
Neighbourhood associations should be encouraged by opinion leaders, local assemblies
and traditional authorities. Efforts at building trust among community members should
be made through communal labour and community meetings. Such efforts could be
response to the broken windows hypothesis. It has dual benefit of boosting social
cohesion and sending signals to criminals that the community is united against antisocial behaviours. This can reduce crime rates and increase perceptions of personal
security.
The police should boost their efforts at building rapport between themselves and
citizens. Once there is confidence that the needs of households and communities could
be efficiently handled by the police and other security agencies, self confidence of
citizens would be boosted, especially for women. This could provide some
psychological empowerment for women who may otherwise be reluctant to report
incidence of domestic violence to the police for justice to be served. This should be
done with some moderation, though, since too much police presence could also instil
fear in people. Excessive police presence could be indicative of some social disorder in
a community.
100
University of Ghana http://ugspace.ug.edu.gh
There is no proof that the use of private security gadgets and services to protect
households improves perception of personal security. This is however quite tricky,
because the use of private security at home could be in response to high prevalence of
stimuli of fear, like armed robbery, unprovoked violence and crime. Further
investigations are needed in this regard.
101
University of Ghana http://ugspace.ug.edu.gh
APPENDIX A
Additional Tables
Table 1A: Item-Total Statistics for the Variables in the Perception of Personal Security Scale
(Alpha=0.857)
Scale
Scale
Correcte
Variables
Mean if Variance if d ItemItem
Item
Total
Deleted
Deleted
Correlati
on
Feel safe when walking down 251.9819
4178.309
0.698
your street alone at night
Feel safe when home alone after 243.6601
4921.606
0.714
dark
Feel safe in daily life
245.0334
4963.269
0.736
Household safe from crime
249.1439
4673.620
0.690
Cronbach's
Alpha
if
Item Deleted
0.828
0.815
0.809
0.823
Table 2A: Item-Total Statistics for Variables in the Public Security Service Trust Scale
(Alpha=0.764)
Scale Mean Scale
Corrected
if
Item Variance if Item-Total
Deleted
Item
Correlatio
Deleted
n
General confidence in public 141.9959
1935.258
0.659
security services
Public security services address 142.1965
2106.294
0.605
problem effectively
Level of confidence in the 150.9242
2566.400
0.536
Police
Cronbach's
Alpha
if
Item
Deleted
0.608
0.673
0.749
Table 3A: Item-Total Statistics for Variables in the Trust for Neighbours Scale
(Alpha=0.904)
Scale
Mean if
Item
Deleted
Trust in neighbours
215.7613
Trust in ethnic group members
216.2608
Trust in non-ethnic group memb 225.8047
Trust in members of other 222.6016
religions
Trust in same organization memb 219.2863
Trust in business owners
219.9401
Scale
Variance if
Item
Deleted
14157.505
14017.899
14591.150
14539.422
Corrected
Item-Total
Correlatio
n
0.705
0.780
0.750
0.739
Cronbach's
Alpha
if
Item
Deleted
0.892
0.880
0.885
0.886
14401.631
14655.374
0.760
0.690
0.883
0.894
102
University of Ghana http://ugspace.ug.edu.gh
Table 4A: Descriptive Statistics for Test of Association between Perception of Personal
Security and Region of Residence, Religion and Ethnicity
Perception Scale for Personal Security
Mean
Standard
Std Err Total N
Deviation of Mean
Region of Natural resource endowed 86.24
21.31
0.30
5040
Residence
Savannah, resource scarce 79.47
22.85
0.34
4419
Greater Accra Region
79.18
21.42
0.50
1806
Other regions
82.58
22.52
0.34
4491
Total
82.49
22.29
0.18
15756
Religion
Catholics/Protestants
83.43
21.53
0.31
4743
Islam
80.36
22.90
0.37
3901
Pentecostal/Charismatic
82.59
22.54
0.33
4668
Others
83.88
22.07
0.45
2444
Total
82.49
22.29
0.18
15756
Ethnicity of Akans
84.20
21.83
0.28
6281
respondent Ewe/Ga-Dangme
82.97
22.28
0.41
3005
s
Guan/Gurma/Grusi
81.91
21.49
0.46
2218
Mole Dagbani
79.65
23.31
0.39
3548
Others
81.35
22.18
0.84
704
Total
82.49
22.29
0.18
15756
Variables
Categories
103
University of Ghana http://ugspace.ug.edu.gh
Table 5A: Multiple Comparisons Table (Post-Hoc Test) for Region of Residence and Perception Scale
Perception Scale for Personal Security, Tukey
HSD
(I) Region of
Residence
(J) Region of Residence
Mean
Differenc Std.
e (I-J)
Error
95% Confidence
Interval
Sig.
Lower
Bound
Upper
Bound
Natural resource Savannah, resource scarce
6.76933* 0.45562 0.000
endowed
5.5987 7.9400
7.05611* 0.60632 0.000
5.4983 8.6139
Greater Accra Region
Other regions
0.45367 0.000
*
Savannah
Greater
Region
3.65825
*
Natural resource endowed -6.76933 0.45562 0.000
-7.9400 -5.5987
Greater Accra Region
.28678
-1.2997 1.8732
Other regions
-3.11108* 0.46845 0.000
0.61746 0.967
*
Accra Natural resource endowed -7.05611 0.60632 0.000
Savannah
-.28678
0.61746 0.967
*
Other regions
2.4926 4.8239
Other regions
-3.39786 0.61602 0.000
Natural resource endowed -3.65825* 0.45367 0.000
-4.3147 -1.9075
-8.6139 -5.4983
-1.8732 1.2997
-4.9806 -1.8151
-4.8239 -2.4926
3.11108* 0.46845 0.000
1.9075 4.3147
Greater Accra Region
3.39786* 0.61602 0.000
*. The mean difference is significant at the 0.05 level.
1.8151 4.9806
Savannah
104
University of Ghana http://ugspace.ug.edu.gh
APPENDIX B
Additional Charts and Graphs
Figure 1B: Histogram Representation of the Confidence in Public Security Scale
Figure 2B: A histogram Representing Trust in Neighbours and Community Members
105
University of Ghana http://ugspace.ug.edu.gh
Figure 3B: Modified Box Plot for Confidence in Public Security Services by Gender
Figure 4B: A Box Plot for Trust in Neighbours Scale
106
University of Ghana http://ugspace.ug.edu.gh
Figure 5B: Box Plot for Gender and Perception of Personal Security
Figure 6B: A Scatter plot of Standardized Residuals (*ZRESID) and Standardized
Predicted Value (*ZPRED)
107
University of Ghana http://ugspace.ug.edu.gh
Figure 7B: Normal Probability Plot of the Standardized Residuals
Figure 8B: Normal PP-Plot of Standardized Residuals
108
University of Ghana http://ugspace.ug.edu.gh
APPENDIX C
Summary of Variables Measurements
Unit of Analysis
Household Heads
Dependent variable
Perception of Personal Security (Continuous)
A four-item additive scale (Alpha=0.857)
scale = 0 to 100
Where 0 indicates lowest perception of personal security or highest perception
of personal insecurity; and 100 measures highest perception of personal
security.
Independent variable
Gender (Dichotomous)
1 = Men
0 = Women
Intermediate variables 1: Fear facilitators
Tension or conflict in community (Dichotomous)
1 = Yes, there is tension or conflict
0 = No, there is no tension or conflict
Experience of robbery (Dichotomous)
1 = Yes, myself or a member of my household
0 = No, neither myself nor a member of my household
Experience of attack or threat (Dichotomous)
1 = Yes, myself or a member of my household
0 = No, neither myself nor a member of my household
Intermediate variables 2: Fear inhibitors
Confidence in public security service (Continuous)
A three-item additive scale (Alpha=0.772)
Scale = 0 to 100
Where 0 indicates no confidence at all in public security service; and 100
indicates maximum trust in public security services to provide protection
Trust for neighbours (Continuous)
A six-item additive scale (Alpha=0.904)
Scale = 0 to 100
Where 0 indicates no trust at all for neighbours; and 100 indicates unreserved
trust for neighbours in the community
Perception of Government responsiveness (Dichotomous)
1 = Yes
0 = No
109
University of Ghana http://ugspace.ug.edu.gh
Private Security Index
1 = Rely on at least one private security gadget or service
0 = Do not rely on any private security gadget or service
Control variables
Spousal co-residence for those married (Dichotomous)
1 = Yes, stay together with spouse in same household
0 = No, do not stay together in same household
Age (Ordinal)
1 = < 20 years
2 = 20 – 44 years
3 = 45 – 64 years
4 = 65+ years [Reference]
Residence (Dichotomous)
1 = Urban
0 = Rural
Region of Residence (Categorical)
1 = Natural resource endowed regions [Reference]
2 = Dry savannah, resource scarce regions
3 = Greater Accra Metropolitan Area
4 = Other regions
Household size (Ordinal)
1 = 1 member [Reference]
2 = 2 members
3 = 3 members
4 = 4 members
5 = 5+ members
Religion (categorical)
1 = Catholics or protestants
2 = Islam [Reference]
3 = Pentecostal/Charismatic
4 = Other
Educational level (Ordinal)
1 = No education [Reference]
2 = Primary education
3 = Secondary school
4 = Post-secondary and tertiary
Employment status (Dichotomous)
1 = Yes, engaged in economic activities in last 7 days
0 = No, not engaged in economic activities in last 7 days
Methods of Analysis
110
University of Ghana http://ugspace.ug.edu.gh
Univariate Analysis: Frequency tables, bar charts, pie charts, histograms
Bivariate Analysis: Chi-square, ANOVA, Correlations
Multivariate Analysis: Multiple Linear Regression
111
University of Ghana http://ugspace.ug.edu.gh
REFERENCES
Abama, E. and Kwaja, C. M. (2009), ‘Violence against women in Nigeria: How the
Millennium Development Goals addresses the challenge’, Journal of PanAfrican Studies, vol. 3, no. 3, pp. 23-34
Abrechtsen,
Eirik
(2003),
‘Security
vs
Safety’,
http://www.iot.ntnu.no/users/albrecht/rapporter/notat%20safety%20v%20secu
rity.pdf (February 10, 2015)
Acharya, Amitav (2001), ‘Human Security: East versus West’, International Journal,
442-460
Acierno, Ron, Alyssa A. Rheingold, Heidi S. Resnick, and Dean G. Kilpatrick (2004),
‘Predictors of fear of crime in older adults’, Anxiety Disorders, vol. 18, pp. 385396
Agarwal, Bina (1997), “‘Bargaining’ and Gender Relations: Within and beyond the
household’, Feminist Economics, vol. 3, no. 1, pp. 1-51
Alaga, Ecoma (2011), ‘Gender and security policy in West Africa’, Working Paper
(Abuja: Friedrich-Ebert-Stiftung)
Alkire, Sabina (2003), ‘A Conceptual Framework for Human Security’, Working Paper
2 for the Centre for Research on Inequality, Human Security and Ethnicity,
CRISE
Amoakohene, M. I. (2004) Violence against women in Ghana: A look at women’s
perceptions and review of policy and social responses, Social Science and
Medicine, 59: 2373-2385
Amankrah, John Yaw (2008), ‘Youth Unemployment in Ghana: Prospects and
challenges’,
Working
Paper
for
Ghana
Statistical
Service,
http://www.cepa.org.gh/researchpapers/Youth73.pdf (January 24, 2015)
Ampofo, Akosua Adomako (1993), ‘Controlling and punishing women: Violence
against Ghanaian women’, Review of African Political Economy, vol. 20, no.
56, pp. 102-111
Anderson, David M. (2002), ‘Vigilantes, violence and the politics of public order in
Kenya’, African Affairs, Vol. 101, No. 405, pp. 531-555
Ardayfio-Schandorf, Elizabeth (2005), ‘Violence against women: The case of Ghana’,
United Nations Expert Group Meeting organized by the UN Division for the
Advancement
of
Women,
Geneva,
April,
http://www.un.org/womenwatch/daw/egm/vaw-stat-2005/docs/expertpapers/Ardayfio.pdf (January 26, 2015)
Armstrong, Elizabeth A and Mary Bernstein (2008), ‘Culture, Power and Institutions:
A Multi-Institutional politics approach to social movement’, Sociological
Theory, vol. 26, no. 1, pp. 74-99
Arriagada, I. and L. Godoy (2000), ‘Prevention or repression? The false dilemma of
Citizen Security’, CEPAL Review, 70: 111-36
Auty, Richard M. (1993), Sustaining Development in Mineral Economies: The resource
curse thesis (London: Routledge)
Ayoob, Mohammed (1986), Regional Security and Third World (London: Croom
Helm)
Ayres, R. L. (1998), ‘Crime and violence as development issues in Latin America and
the Caribbean’, Latin America and Caribbean Studies, viewpoint studies series
(Washington D. C: World Bank)
112
University of Ghana http://ugspace.ug.edu.gh
Ballantine, K. (2003) ‘Beyond Greed and Grievance: Reconsidering the Economic
Dynamics of Armed Conflict’, in K. Ballentine and J. Sherman (eds), The
Political Economy of Armed Conflict: Beyond Greed and Grievance (London:
Lynne Riener), pp. 259–83
Ballentine, Karen and HeikoNitzschke (2003), Beyond Greed and Grievance: Policy
Lessons from Studies in the Political Economy of Armed Conflicts, International
Peace Academy Policy Report (New York: International Peace Academy)
Barrett, Michelle (1980), Women’s Oppression Today: Problems in Marxist feminist
analysis (London: Verso)
Bawah, Ayaga Agula, Patricial Akweongo, Ruth Simmons and James F. Phillips
(1999), ‘Women’s Fears and Men’s Anxieties: The impact of family planning
on gender relations in Northern Ghana’, Studies in Family Planning, vol. 30,
no. 1, pp. 54-66
Bawah, Ayaga Agula (2002), ‘Spousal communication and Family Planning behaviour
in Navrongo: A longitudinal assessment’, Studies in Family Planning, vol. 33,
no. 2, pp. 185-194.
Berdal, Mats (2005), Beyond Greed and grievance – and not too soon...: A Review
Essay, in Review of International Studies, Vol. 31, pp. 687-698
Black, B.M., Weisz, A.N. and Bennett, L.W. (2010), Graduating social work students’
perspectives on domestic violence, Journal of Women and Social Work, vol. 25
no. 2, pp. 173-184
Bhasin Kamla (2004), Exploring Masculinity (New Delhi: Women Unlimited
Publishers)
Black et al. (1982), ‘A study of battered women in a psychiatric setting’, Women and
Therapy, research paper 13.
Blanchard, Eric M. (2003), ‘Gender, International Relations, and the Development of
feminist security theory’, Signs, vol. 28, no. 3, pp. 1289-1312
Booth, Ken (1991), ‘Security and Emmancipation’, Review of International Studies,
Vol. 17, 313-26
Borwankar, R., R. Diallo and A. E. Sommerfelt (2008), Gender-based violence in subSaharan Africa: A review of Demographic and Health Survey findings and their
use in National Planning (Washington D.C: USAID)
Bower, B. L. and O’Connell, R. T. (1990), Linear Statistical Models: An applied
approach (2nd ed.) (Boston, MA: Duxbury)
Bowman, C. G. (2003), ‘Domestic violence: Does the African context demand a
different approach?’ International Journal of Law and Psychiatry, vol. 26, pp.
473-491.
Bradshaw, Sarah (2004), ‘Socio-economic impacts of natural disasters: a gender
analysis’ Serie Manuales 32, United Nations Economic Commission for Latin
America (ECLA)
Bradley, C. (2001), ‘Family and Sexual Violence in PNG: An Integrated Long-Term
Strategy’, Report to the Family Violence Action Committee of the Consultative
Implementation and Monitoring Council, Paper 84 (Port Moresby: Institute of
National Affairs)
Brown, G. (2004), ‘Gender as a factor in the response of the law-enforcement system
to violence against partners,’ Sexuality and Culture, vol. 8, pp. 1–87.
113
University of Ghana http://ugspace.ug.edu.gh
Brunton-Smith, Ian and Jonathan Jackson (2012), ‘Urban Fear and its roots in place’,
in Vania Ceccato (ed.) Urban Fabric of Crime and Fear (London: Springer),
pp. 55-82
Buchmann, C. and E. Hannum (2001), ‘Education and stratification in developing
countries: A review of theories and research, Annual review of sociology, Vol.
27, pp. 77-102
Bursik, R. J., Jr., and H. G. Grasmick (1993), Neighborhoods and crime: The
dimensions of effective community control (New York: Lexington).
Byres, Eric and John Cusimano (2010), ‘Safety and Security: Two sides of the same
coin’, Control: Promoting Excellence in process automation,
http://www.controlglobal.com/articles/2010/safetysecurity1004/ (February 10,
2015)
Cain, Mead, Syeda Rokeya Khanam and Shamsun Nahar (1979), ‘Class, Patriarchy and
Women’s Work in Bangladesh’, Population and Development Review, Vol. 5,
No. 3, 403-438
Cammack D. (2000), Promoting Gender Sensitive Operations (Nairobi: WFP)
Cantalupo, Nancy Chi, Lisa Vollendorf Martin, Kay Pak and Sue Shin (2006),
‘Domestic Violence in Ghana: The open secret’ Georgetown Public Law and
Legal
Theory
Research
Paper
no.
10-56,
http://scholarship.law.georgetown.edu/cgi/viewcontent.cgi?article=1432&cont
ext=facpub (January 26, 2015)
Caprioli, M. (2005), ‘Primed for Violence: The role of gender inequality in predicting
internal conflict’, International Studies Quarterly, vol. 49, pp. 161-178
Caldwell, John C. and Pat Caldwell (1987), “The Cultural Context of High Fertility in
Sub-Saharan Africa”, Population Development Review, Vol. 13, pp. 409–437.
Carrigan, T. R. W. Connell and J. Lee (1985), ‘Towards a New Sociology of
Masculinity’, Theory and Society, vol. 14, 5, pp. 551-600
Castells, Manuel (2007), ‘Communication, Power and Counter-power in the Network
Society’, International Journal of Communication, vol. 1, pp. 238-266
Connell, R. W. (2005), Masculinities, 2nd Ed (Cambridge: Polity)
Connell, R. W. and J. Messerschmidt (2005), ‘Hegemonic Masculinity: Rethinking the
Concept’, Gender and Society, vol. 19, no. 6, pp. 829-859
Convention on the Elimination of All Forms of Discrimination against Women
(CEDAW) (1992), ‘Compilation of General Comments and General
Recommendations Adopted by Human Rights Treaty Bodies’, General
Recommendation 19, Violence against Women, UN Doc. HRI/GEN/1/Rev.1,
http://www.un.org/womenwatch/daw/cedaw/recommendations/recomm.htm#r
ecom19
Cornwall, A. and N. Lindisfarne (eds) (1994), Dislocating Masculinity: Comparative
Ethnographies (London: Routledge)
Charvet, J. (1982), Modern Ideologies: Feminism (London: J. M. Dent and Sons
Limited)
Chesler, P. (1991), ‘Mothers on Trial: The Custodial Vulnerability of Women’,
Feminism and Psychology, vol. 1, No. 3, 409-425
Clare, Anthony (2000), On Men: Masculinity in Crisis (London: Chatto and Windus).
Coker-Appiah, D. and Cusack, K. (eds). (1999), ‘Breaking the silence and challenging
the myths of violence against women and children in Ghana’, A report on a
114
University of Ghana http://ugspace.ug.edu.gh
national study of violence, (Accra: Gender Studies & Human Rights
Documentation Centre)
Collier, P., L. Elliot, H. Hegre, A. Hoeffler, M. Reynal-Querol and N. Sambanis (2003),
Breaking the Conflict Trap: Civil war and development policy (Washington, D.
C.: World Bank)
Collier, Paul, Anke Hoeffler, and Nicholas Sambanis (2005), ‘The Collier-Hoeffler
Model of Civil War Onset and the Case Study Project Research Design, in Paul
Collier and Nicholas Sambanis (eds.), Understanding Civil War: Evidence and
Analysis, Volume 1: Africa (Washington, D. C.: The World Bank), pp. 1-34
Collier, Paul (2000), ‘Doing Well out of War’, in Mats Berdal and David Malone (eds.)
Greed and Grievance: Economic Agendas in Civil Wars (Boulder: Lynne
Rienner)
Collier, Paul and Hans Binswanger (1999), ‘Ethnic loyalties, State formation and
Conflict’, A background paper prepared for the Africa in the 21st Century
Project (Washington, D. C.: World Bank)
Dodoo, F. Nii-Amoo (1998), “Men Matter: Additive and interactive gendered
preferences and reproductive behaviour in Kenya”, Demography vol. 35, pp.
229–42.
Dodoo, Francis Nii-Amoo, Melissa Sloan and Eliya Msiyaphazi Zulu (2003), ‘Space,
context and hardship: Socializing children into sexual activity in Kenyan
slums’, in S. Agyei-Mensah and J. B. Casterine (eds), Reproduction and Social
Context in Sub-Saharan Africa (Westport, CT.: Greenwood Press), pp. 146-160
Dodoo, F. Nii-Amoo and Ashley E. Frost (2008), “Gender in African Population
Research”, Annual Review of Sociology, vol. 34, pp. 431–52.
DuBow, F., F. McCabe and G. Kaplan (1979), Reactions to Crime: A critical review of
the literature (Washington, D. C.: U. S. Government Printing Office).
Dutton, Donald G. and Tonia L. Nicholls (2005), ‘The gender paradigm in domestic
violence research and theory’, Aggression and Violent Behaviour, Vol. 10, pp.
680-714
Eckman, Anne, Aparna Jain, Sarah Degnan Kambou, Doris Bartel and John Crownover
(2007), ‘Exploring dimensions of masculinity and violence’, a research paper
for the Western Balkan Gender-Based Violence Prevention Initiative,
http://www.icrw.org/publications/exploring-dimensions-masculinity-andviolence (Accesses on April 8 2015)
Edstrom, Jerker, Abhijit Das and Christ Dolan (2014), ‘Introduction: Undressing
patriarchy and masculinities to re-politicise Gender’, IDS Bulletin, Vol. 45, No.
1 (Oxford: John Wiley and Sons Ltd), pp. 1-10
Edward, Miguel, Satyanath Shankar, and Ernest Sergenti (2003) Economic Shocks and
Civil Conflict: An instrumental variables approach, Working Paper 40 for the
Bureau for Research in Economic Analysis of Development (BREAD)
Eisler, Riane (1990), The Chalice and the Blade: Our History, Our Future (London:
Unwin Paperbacks).
Ejibunu, Hassan Tai (2007), Nigeria’s Niger Delta Crisis: Root causes of
Peacelessness,
EPU
Research
Paper,
Issue
07/07,
http://epu.ac.at/fileadmin/downloads/research/rp_0707.pdf (August 15, 2013)
115
University of Ghana http://ugspace.ug.edu.gh
Engels, Friedrich (1972), Origin of the Family, Private Property, and the State, Eleanor
Burke Leacock (Trans.), Original publication in 1884 (New York: International
Publishers)
Emslie, Carol and Kate Hunt (2009), ‘Men, Masculinities and Heart Disease: A
systematic review of the qualitative literature’, Current Sociology, Vol. 57, No.
2, pp. 155-191
Emslie, Carol (2005), ‘Women, men and coronary heart disease: Review of the
qualitative literature’, Journal of Advanced Nursing, Vol. 51, No. 4, pp. 382395
Estes, Carroll, Terry O’Neill and Heidi Hartmann (2012), ‘Breaking the Social Security
Glass Ceiling: A Proposal to modernize women’s benefits’,
http://www.ncpssmfoundation.org/Portals/0/breaking_ss_glass_ceiling.pdf
(February 26, 2015)
Eves, Richard (2010), ‘Masculinity Matters: Men, Gender-Based violence and the
AIDS Epidemic in Papua New Guinea’, in Vicki Luker and Sinclair Dinnen
(eds) Civic Insecurity: Law, order and HIV in Papua New Guinea (Canberra:
The Australian National University), pp. 47-80
Farr, Vanessa and Kiflemariam Gebre-Wold (eds.) (2002) Gender Perspectives on
Small Arms and Light Weapons, Regional and International Concern ( Bonn:
Bonn International Center for Conversion)
Farrall, S. and D. Gadd (2004), ‘Evaluating Crime Fears: A research note on a pilot
study to improve the measurement of the “Fear of crime” as a performance
indicator, Evaluation, Vol. 10, No. 4, pp. 493-502
Ferraro, Kenneth F. (1995), Fear of Crime: Interpreting victimization risk (Albany, N.
Y.: State University of New York Press)
Ferraro, Kenneth F. (1996), ‘Women's fear of victimization: Shadow of sexual assault?’
Social Forces, vol. 75, pp. 667–690.
Ferraro, Kenneth, and R. LaGrange (1987), ‘The measurement of fear of crime,’
Sociological Inquiry, vol. 57, pp. 70−101.
Field, Andy (2009), Discovering Statistics Using SPSS, Third Edition (London: Sage
Publications Ltd)
Fisher, B. S., and J. J. Sloan (2003), ‘Unraveling the fear of victimization among college
women: Is the “shadow of sexual assault hypothesis” supported?’ Justice
Quarterly, vol. 20, pp. 633–659.
Ferraro, K. F. and La Grange R. L. (1998), ‘Are older people more afraid of crime?’,
Journal of Aging Studies, Vol. 10, pp. 277-287
Fodor, J. George and Rayka Tzerovska (2004), ‘Coronary Heart Disease: Is gender
important?, Heart and Lung, Vol. 30, No. 3, pp. 167-176
Francis, Diana (2004), ‘Gender, Power Asymmetries and Gender in Conflict
Transformation’, a research paper for the Bergof Research Centre for
Constructive
Conflict
Management,
available
at
http://edoc.vifapol.de/opus/volltexte/2011/2570/pdf/francis_handbook.pdf
(February 24, 2015)
Freud, Sigmund (1965), A General Introduction into Psychoanalysis (Alexandria:
Library of Alexandria)
Frith, Uta (2013), ‘Are there innate mechanisms that make us social beings?’, Pontifical
Academy of Social Sciences, Scripta Varia 121, available at
116
University of Ghana http://ugspace.ug.edu.gh
http://www.casinapioiv.va/content/dam/accademia/pdf/sv121/sv121-frithu.pdf
(February 24, 2015)
Galtung, J. (1969), ‘Violence, Peace and Peace Research’, Journal of Peace Research,
vol. 6 no. 3, 167-91
Galtung, J. (1975) ‘Peace: Research, Education, Action: Essays in Peace Research
Volume One’, Bucuresti, Romania: CIPEXIM.
Galtung, J. (1985), ‘Twenty-five years of peace research: Ten challenges and some
responses’, Journal of Peace Research, 22(2), 145-46
Galtung, J. (1990) ‘Cultural Violence’, Journal of Peace Research, vol. 27, no. 3, pp.
291–305
Garofalo, J. (1981), ‘The fear of crime: Causes and consequences,’ Journal of Criminal
Law and Criminology, vol. 72, pp. 839−857.
Garofalo, J. and J. Laub (1978), ‘The fear of crime: Broadening our perspectives’,
Victimology, vol. 3, pp. 242-253.
Gelles, Richard (1972), The Violent Home (Beverly Hills, CA: Sage)
Gender Policy for the YMCA (2011), ‘Empowering the female and male youth of
Ghana for the African Renaissance through the promotion of gender equality’,
http://www.ymcaghana.org/home/Brochure/Final_GENDER%20POLICY.pdf
(March 11, 2015)
Ghana Statistical Service (GSS) (2014), Ghana Living Standards Survey Round 6:
Main Report (Accra: Ghana Statistical Service)
Ghana Statistical Service (GSS) (2013), 2010 Population and Housing Census:
National Analytical Report (Accra: Ghana Statistical Service)
Ghana Statistical Service (GSS) (2009), Ghana Time Use Survey 2009: How Ghanaian
women and men spend their Time (Accra: Ghana Statistical Service)
Ghanaweb. (2010), ‘109,784 cases of domestic violence recorded’,
http://www.ghanaweb.com/GhanaHomePage/NewsArchive/artikel.php?ID=18
3316
Giddens, A. (2011), ‘The geopolitics of climate change’, Politics of Climate Change
(Polity, Cambridge), pp. 203-228
Gilchrist, E., J. Bannister, J. Ditton, and S. Farrall (1998), ‘Women and the “fear of
crime”: Challenging the accepted stereotype’, British Journal of Criminology,
Vol. 38, No. 2, pp. 283-298
Gill, Rosalind (n.d.), Seminar on ‘Masculinity in Crisis?’ in Rethinking Masculinity:
Men
and
their
Bodies.
Available
at:
http://fathom.lse.ac.uk/seminars/21701720/21701720_session5.html (March
20, 2015)
Glass, G. V. (1966), ‘Testing homogeneity of variances’, American Educational
Research Journal, vol. 3, no. 3, pp. 187 – 190.
Goldberg, H. (1979), The new male: From self-destruction to self care (New York:
William Morrow).
Goodey, J. (1997). ‘Boys don't cry: Masculinities, fear of crime and fearlessness,’
British Journal of Criminology,vol. 37, pp. 401–418.
Grazebrook, Trish (2011), ‘Women and Climate Change: A case study from Northern
Ghana’, Hypatia, vol. 26, no. 4, pp. 762-782
Grossman, Anna and Cate Owren (eds.) (2008), ‘Gender, Climate Change and Human
Security: Lessons from Bangladesh, Ghana and Senegal’, a report prepared for
the Hellenic Foundation for European and Foreign Policy (ELIAMEP) in May
117
University of Ghana http://ugspace.ug.edu.gh
GURR, T. R. (1970) Why Men Rebel, (Princeton, NJ: Princeton University Press).
Gyimah, S. O. (2009), ‘Cohort differences in women’s educational attainment and the
transition to first marriage in Ghana’, Population Research and Policy Review,
Vol. 28, No. 4, pp. 455-471
Hagan, John (2007), ‘Death in Darfur: The Science of Human Rights, War Crimes, and
Humanitarian Emergencies’, Researching Law, 18, 3 (2007), pp. 1-8
Hale, Chris (1996), ‘Fear of Crime: A review of literature’, International Review of
Victimology, Vol. 4, pp. 79-150
Harcourt, Wendy (2009), ‘Literature Review on Gender and Fragility’,
http://erd.eui.eu/media/review-by-w-harcourt.pdf (January 25, 2014)
Harders, Cilja (2011), ‘Gender relations, violence and conflict transformation’, in B.
Austin and M. Fischer, H. J. Geissmann (eds.), Advancing Conflict
Transformation: The Berghoff Handbook II (Opladen/Farmington Hills:
Budrich Publishers), pp. 132-151
Harley, N. (2005), PNG Armed Violence Assessment (Port Moresby: UNDP)
Heilman, B.P. (2010), ‘A life-course perspective of Intimate Partner Violence in India:
What picture does the NFHS-3 paint?’ A master’s thesis submitted to MALD,
Tufts University.
Henningsten, Erik and Peris Jones (2013), ‘‘what kind of hell is this!’ Understanding
the Mungiki movement’s power of mobilization’, Journal of East African
Studies, vol. 7, no. 3, pp. 371-388
Hinkle, J., and D. Weisburd (2008), ‘The irony of broken windows policing: A microplace study of the relationship between disorder, focused police crackdowns and
fear of crime,’ Journal of Criminal Justice, vol. 36, pp. 503−512.
Horne, Christine, F. Nii-Amoo Dodoo and Naa Doduaa Dodoo (2013), ‘The shadow of
indebtedness: Bridewealth and norms constraining female reproductive
autonomy’, American Sociological Review, Vol. 78, no. 3, pp. 503-520
Holland, Jack (2006), Misogyny: The World’s Oldest Prejudice (London: Robinson)
Humphreys, Macartan (2005), ‘Natural Resources and Armed Conflicts: Issues and
Options’, in Karen Ballentime and Heiko Nitzschke (eds.) Profiting from
Peace: Managing the Resource Dimensions of Civil War (Boulder, Colorado:
Lynne Reinner Publishers inc. For International Peace Academy)
Hunnicutt, Gwen (2009), ‘Varieties of Patriarchy and Violence against Women:
Resurrecting “Patriarchy” as a theoretical tool’, Violence Against Women, vol.
15, no. 5, pp. 553-573
Ibrahim Index of African Governance 2013
Jackson, J. (2006) ‘Introducing fear of crime to risk research’, Risk Analysis vol. 26,
no. 1, pp. 253–264
Jackson, Jonathan (2005), ‘Validating new measures of the fear of crime’, International
Journal of Social Research Methodology, Vol. 8, No. 4, pp. 297-315
Jackson, Jonathan and Mai Stafford (2009), ‘Public Health and fear of crime: A
prospective cohort study, British Journal of Criminology, Vol. 49, No. 6, pp.
832-847
Kambarami, Maureen (2006), ‘Femininity, sexuality and culture: Patriarchy and female
subordination in Zimbabwe’, Understanding Human Sexuality Lecture Series,
by the Africa Regional Sexuality Resource Centre, Harare,
http://www.arsrc.org/downloads/uhsss/kmabarami.pdf (February 2, 2015)
118
University of Ghana http://ugspace.ug.edu.gh
Kameri-Mbote, Patricia (2004) ‘Gender, Conflict and Regional Security’, in Makumi
Mwagiru (ed.), African Regional Security in the age of Globalisation (Nairobi:
Heinrich Boll Foundation), pp. 83-94
Kaplinsky, R. (2000), ‘Globalisation and economic insecurity’, IDS Bulletin 32(2), 1324
Keen, David (2012), Greed and Grievance in Civil War, in International Affairs, Vol.
88, No. 4, pp. 757-777
Kerr, Sergeant Lindy (1993), ‘Crime and Older People: Reducing fear of crime among
the elderly: The Victorian experience through the Confidence Living Program
for older Victorians’ a paper presented at a conference in February (Canberra:
Australian Institute of Criminology)
Kishor, S. and Johnson, K. (2004), Profiling domestic violence – A multi country study
(Calverton, Maryland: ORC Macro)
Knight, David and Maria Tullberg (2011), ‘Managing masculinity/mismanaging the
corporation’, Organisation, vol. 19, no. 4, pp. 385-404
Kolo, Kenneth (2010), ‘Youth Perceptions of Human Security in Africa’,
http://www.nepad.org/nepad/blog/2010/09/28/1695/youth-perceptions-humansecurity-africa (December 1, 2014)
Kruijt, D. and K. Koonings (1999), ‘Introduction: Violence and fear in Latin America’,
in K. Koonings and D. Kruijt (eds), Societies of fear: The legacy of civil war,
violence and terror in Latin America (London: Zed books)
Kuenzi, M. (2005), ‘The role of non-formal education in promoting democratic
attitudes: Findings from Senegal’, Democratisation, vol. 12, no. 2, pp. 223-243
Lannert, Brittany K. Antonio M. Garcia, Kathryn E. Smagur, Matthew M. Yalch, Alytia
A. Levendosky, G. Anne Bogat and Joseph S. Lonstein (2014), ‘Relational
trauma in the context of intimate partner violence’, Child Abuse and Neglect,
Vol. 38, pp. 1966-1977
Le Billon, Philippe (2001), The Political Ecology of War: Natural Resources and
Armed Conflicts, Political Geography, Vol. 20, pp. 561-584
Lewis, Dan A. and Greta Salem (1986), Fear of Crime: Incivility and the production of
a social problem (New Brunswick, N. J.: Transaction Books)
Lo, Kwai-Cheung (2010), Excess Masculinity in Asian Cultural Production (Beijing:
Sunny Press)
Madriz, E. (1997), Nothing bad happens to good girls: Fear of crime in women's lives,
(Berkeley: University of California Press).
Mahama, Yakubu (2009), Murder of an African King: Ya-Na Yakubu II, 2nd edition
(New York: Vintage Press Inc)
Mama, Amina (2014), ‘Beyond Survival: Militarism, equity and women’s security’,
Dick Foeken, Ton Dietz, Leo de Haan and Linda Johnson (eds.), Development
and Equity: An interdisciplinary exploration by Ten Scholars from Africa, Asia
and Latin America (Leiden: BRILL)
Mama, Amina and Margo Okazawa-Ray (2012), ‘Militarism, conflict and women’s
activism in the global era: challenges and prospects for women in three West
African contexts’, Feminist Review, vol. 101, pp. 97-123
Maas, A. H. E. M. and Y. E. A. Appelman (2010), ‘Gender differences in coronary
heart disease’, Netherlands Heart Journal, Vol. 18, No. 12, pp. 598-602
Matthews, J. Tuchman (1989), ‘Redefining Security’, Foreign Affairs Vol. 68, 162-77
119
University of Ghana http://ugspace.ug.edu.gh
MacKay, Judith and George Mensah (2004), Atlas of Heart Disease and Stroke
(Geneva: World Health Organisation)
Makoloo, Maurice Odhiambo (2005), Kenya: Minorities, indigenous peoples and
ethnic diversity, A report for the Minority Rights Group International
McAdam, D. (1982), Political Process and the Development of Black Insurgency,
1930-1970 (Chicago: University of Chicago Press)
McDowell, L. and R. Pringle (1992), Defining women: Social institutions and gender
division (Oxford: Polity Press)
McGarrell, Edmund F., Andrew L. Giacomazzi and Quint C. Thurman (1997),
‘Neighbourhood disorder, integration and the fear of crime,’, Justice Quarterly,
vol. 14, no. 3, pp. 479-500
Meintjes, S., A. Pillay and M. Turshen (2001), ‘There is no aftermath for women’, in
Sa Meintjies, A. Pillay and M. Tursheen (eds.), The Aftermath: Women in postconflict transformation (London: Zed)
Miller, M. (2001), Poverty as a Cause of War?, Interdisciplinary Science Review, Vol.
25, No. 4, pp. 273-297
Muggah, Robert (2012)” Why personal security should be part of the post-2015
development
agenda’,
IPI
Global
Observatory,
available
at
http://theglobalobservatory.org/2012/11/why-personal-security-should-bepart-of-the-post-2015-development-agenda/ (November 22, 2014)
Møller, V. (2005) ‘Resilient or resigned? Criminal victimisation and quality of life in
South Africa’. Social Indicators Research vol. 72, pp. 263–317
Moser, Caroline O. N. and Dennis Rodgers (2005), ‘Change, violence and insecurity in
non-conflict situations’, Working paper 245 for Overseas Development
Institute, London
Moser, C. O. N. and C. McIlwaine (2004), Encounters with Violence in Latin America:
Urban poor perceptions from Columbia and Guatemala (London: Routledge)
Moser, C (1993) Gender Planning and Development: theory, practice and training
(Routledge, London)
Mudege, Netsayi, Eliya M. Zulu and Chimaraoke Izugbara (2008), ‘How insecurity
impacts on school attendance and school dropout among urban slum children in
Nairobi’, International Journal of Conflict and Violence, 2(1), 98-112
Murshed, S. M. (2002) ‘Civil War, Conflict and Underdevelopment’, Journal of Peace
Res 39(4):387–393
Murunga, Godwin Rapando (2006), ‘Mungiki and the Re-traditionalisation of Society
Argument: A quest for recognisiton?’, CODESRIA Bulletin, Nos. 3 and 4, pp.
28-32
Myers, R. (1990), Clasical and Modern regression with applications (2nd ed.) (Boston,
MA: Duxbury)
New Partnership for African Development (NEPAD) (2001), NEPAD founding
document,
Abuja,
October,
http://www.nepad.org/system/files/framework_0.pdf (January 13, 2015)
Neubert, Mitchell J. and Simon Taggar (2004), ‘Pathways to informal leadership: The
moderating role of gender on the relationship of individual differences and team
member network centrality to informal leadership emergence’ The Leadership
Quarterly, vol. 15, no. 2, pp. 175-195
Odotei, Irene and Albert K. Awedoba (2006), Chieftaincy in Ghana: Culture,
Governance and Development (Accra: Sub-Saharan Publishers)
120
University of Ghana http://ugspace.ug.edu.gh
Odumosu, Olakunle F. (1999), ‘Social Costs of Poverty: The case of crime in Nigeria’,
Journal of Social Development in Africa, vol. 14, no. 2, pp. 71-85
Ofei-Aboagye, R.O. (1994). Altering the strands of the fabric: A preliminary look at
domestic violence in Ghana. Feminism and the Law, 19(4): 924-938
Ogundari, K. and A. B. Aromolaran (2014), ‘Impact of education on Household welfare
in Nigeria’, International Economic Journal, Vol. 28, No. 2, pp. 345-364
Opoku, Boateng Eric (2012), ‘Coping with Urban Poverty: An examination of
household and individual livelihood strategies in Ghana; A Case study of
Sunyani
Municipality’,
http://ir.knust.edu.gh/bitstream/123456789/4513/1/BOATENG%20ERIC%20
OPOKU%20THESIS%202011.pdf (January 24, 2015)
Osam, Susanna (2004), Violence against Women in Ghana: An analysis of cases
presented in the print media (Accra: ABANTU for Development)
Page, Ella (2010), ‘Men, masculinity and guns: Can we break the link?’ an International
Action Network on Small Arms (IANSA) Women Network paper,
www.iansa.org/women/
Paris, Roland (2004), ‘Still an inscrutable concept’, Security Dialogue 35 (3) pp. 370371
Paris, Roland (2001), ‘Human Security: Paradigm shift or hot air?’, International
Security, 26 (2), pp. 87-102
Parmar, P. Kaur, Pooja Agrawal, Ravi Goyal, Jennifer Scott and P. Gregg Greenough
(2014), ‘Need for a gender-sensitive human security framework: results of a
quantitative study of human security and sexual violence in Djohong District,
Cameroun’, Conflict and Health, vol. 8, no. 6, pp. 1-12
Pateman, C. (1988) The Sexual Contract (Stanford: Stanford University Press)
Pearl, Rebecca and Irene Dankelman (2008), ‘Gender, Climate Change and Human
Security in Ghana’, in Anna Grossman and Cate Owren (eds.), Gender, Climate
Change and Human Security: Lessons from Bangladesh, Ghana and Senegal, a
report prepared by The Women’s Environment and Development Organisation
(WEDO), May , pp. 33-46
Philaretou, Andreas G. (2001), ‘An analysis of masculine socialization and male sexual
anxiety’, a dissertation submitted to the faculty of Virginia Polytechnic and
State University for a PhD in Human Development, Blacksburg, Virginia,
December.
Piazza, James A. (2008), ‘Do Democracy and Free Markets Protect us from
Terrorism?’, International Politics, Vol. 45, pp. 72-91.
Pickens, Jeffrey (2005), ‘Attitudes and Perceptions’, N. Borkowski (ed.),
Organisational Behaviour in Health Care (Boston, NY: Jones and Bartlett
Publishers), pp. 43-76.
Rand, M. (2008), Criminal Victimization, 2007 (Washington, D. C.: United States
Department of Justice)
Riger, S., M. Gordon, and R. Le Bailley (1978), ‘Women's fear of crime: From blaming
to restricting the victim,’ Victimology, vol. 3, pp. 274–284.
Roberts, Benjamin (2010), ‘Fear factor: Perceptions of safety in South Africa’, in B.
Robert, Mwa Kivilu and YD David (eds) South African Sicial Attitudes: 2nd
Report: Reflections on the age of hope (HSRC Press: Cape Town), pp. 250-257
Ross, Michael L. (2012), The Oil Curse: How Petroleum Wealth Shapes the
Development of Nations (Princeton, N. J.: Princeton University Press)
121
University of Ghana http://ugspace.ug.edu.gh
Ross, Michael L. (2004), ‘How do natural resources influence civil wars? Evidence
from thirteen cases’, International Organisation, 58, pp. 35-67
Rotberg, Robert I. (2004), ‘The failure and collapse of Nation-States: Breakdown,
Prevention, and Repair’, in Robert I. Rotberg, ed., When States fail: Causes and
Consequences (New Jersey: Princeton University Press), pp. 1-50.
Rountree, P., and K. Land (1996), ‘Perceived risk versus fear of crime: Empirical
evidence of conceptually distinct reactions in survey data,’ Social Forces, vol.
74, pp. 1353−1376.
Rowbotham, Sheila (1981), ‘The trouble with Patriarchy’, in No Turning Back:
Writings from the women’s liberation movement, 1975 to 1980 (ed.), A Feminist
Anthropology Collections (London: Women’s Press)
Russell, Diana E. H. (1982), Rape in Marriage (New York: Macmillan)
Rutten, Marcel and Sam Owuor (2009), Weapons of Mass Destruction: Land ethnicity
and the 2007 elections in Kenya, Journal of Contemporary African Studies, Vol.
27, No. 3, pp. 305-324
Sampson, R. J., and S. W. Raudenbush (2004), ‘Seeing disorder: Neighborhood stigma
and the social construction of “Broken Windows,”’ Social Psychology
Quarterly, vol. 67, pp. 319−342.
Sampson, R. J., S. W. Raudenbush, and F. Earls (1997), ‘Neighborhoods and violent
crime: A multilevel study of collective efficacy,’ Science, vol. 277, pp.
918−924.
Saunders, D. (1988). ‘Wife abuse, husband abuse or mutual combat: A feminist
perspective on the empirical findings’, In K. Yllo and M. Bograd (Eds.)
Feminist perspectives on wife assault, (Newbury Park: Sage) pp. 90–113
Schafer, Joseph A., Beth M. Huebner and Timothy S. Bynum (2006), ‘Fear of crime
and criminal victimization: Gender-Based contrasts’, Journal of Criminal
Justice, Vol. 34, pp. 285-301
Scarboroudh, Brittney K., Toya Z. Like-Haislip, Kenneth J. Novak, Wayne L. Lucas
and Leanne F. Alarid (2010), ‘Assessing the relationship between individual
characteristics, neighbourhood context and fear of crime’, Journal of Criminal
Justice, Vol. 38, pp. 819-826
Scott, H. (2003), ‘Stranger danger: Explaining women's fear of crime,’ Western
Criminology Review, vol. 4, pp. 203–214.
Seidler, Victor Jeleniewskic (2005), Transforming Masculinities: Men, Cultures,
Bodies, Power, Sex and Love (London: Routledge).
Sharaky, Abbas M. (2005), ‘Oil, water, minerals, and the crisis in Darfur, Sudan’, (A
Research Paper, Institute of African Research and Studies, Cairo)
Sharp, Imogen (ed.) (1994), Coronary Heart Disease: Are women special? (London:
National Forum for CHD Prevention)
Sideris, T. (2001) ‘‘Rape in War and Peace: Social Context, Gender, Power and
Identity.’’ In S. Meintjes, A. Pillay and M. Tursher (eds.), The Aftermath:
Women in Post-Conflict Transformation (London, UK: Zed Books Ltd), pp.
142–158.
Skogan, Wesley G. (1990), Disorder and Decline: Crime and the Spiral of Decay in
American cities (New York: Oxford University Press)
Skogan, W. G. and M. G. Maxfield (1981), Coping with Crime (Beverly Hills: Sage)
122
University of Ghana http://ugspace.ug.edu.gh
Stafford, Mai., Tarani Chandola and Michael Marmot (2007), ‘Association between
fear of crime and mental health and physical functioning’, American Journal of
Public Health, Vol. 97, No. 11, pp. 2076-2081
Stanko, E. A. (1990), Everyday violence: How men and women experience sexual and
physical danger, (London: Pandora).
Stanko, E. A. (1995), ‘Women, crime, and fear,’ Annals of the American Academy of
Political and Social Science, vol. 539, pp. 46–58.
Statistics Canada (2003), ‘Family violence in Canada: A statistical profile,’ Ottawa7
Canadian Centre for Justice Statistics.
Stern, Maria and Joakim Ojendal (2010), ‘Mapping the Security Development Nexus:
Conflict, Complexity, Cocophony, Convergence?’ in Security Dialogue,
Special Issue on the Security-Development Nexus Revisited, 41 (1), pp 5-30
Stets, J., and M. Straus (1992), ‘Gender differences in reporting marital violence,’ in
Physical violence in American families (New Brunswick, NJ.: Transaction
Publishers), pp. 151–166.
Stewart, Frances (2010), Horizontal Inequalities as a Cause of Conflict: A review of
crisis findings, Background Paper, World Bank Development Report 2011.
Available
at
https://openknowledge.worldbank.org/bitstream/handle/10986/9126/WDR201
1_0029.pdf?sequence=1 (May 19, 2014)
Stiglitz, Joseph E., Amartya Sen, and Jean Paul Fitoussi (2010), ‘Report by the
Commission on the Measurement of Economic Performance and Social
Progress’,
http://www.stiglitz-sen-fitoussi.fr/documents/rapport_anglais.pdf
(Accessed on April 10, 2015)
Steiner, Richard (2010), Double Standard: Shell practices in Nigeria compared with
international standards to prevent and control pipeline oil spills and the
deepwater horizon oil spill (Amsterdam: Friends of the Earth Netherlands)
Sultana, Abeda (2011), ‘Patriarchy and women’s subordination: A theoretical analysis’,
The Arts Faculty Journal, July 2010-June 2011
Swatay, Ahmed Zeb (2012), ‘Analysis of the pressures of masculinity: The case of
Pakhtum men/boys in district Swat’, A research report submitted to SANAM,
Pakistan, April
Sylvester, Christine (1994), Feminist Theory and International Relations in a
Postmodern Era (Cambridge: Cambridge University Press)
Tenkorang, Eric Y., Eric H. Yeboah and Yaa Adobea Owusu (2013), ‘Factors
influencing domestic and marital violence against women in Ghana’,
http://paa2013.princeton.edu/papers/130186 (January 26, 2015)
Throup, David W. (2011), ‘Ghana: Assessing the risk to stability’, A report of the CSIS
Africa Programme, (Washington, D. C.: Centre for Strategic International
Studies)
Tomasello M, Vaish A. (2012), ‘Origins of Human Cooperation and Morality’, Annual
Review of Psychology, vol. 64, pp. 231-255
Tulloch, Marian (2003), ‘Combining classificatory and discursive methods:
Consistency and variability in responses to the threat of crime’, British Journal
of Social Psychology, Vol. 42, No. 3, pp. 461-476
UNPFA (2003) ‘‘Ending Widespread Violence Against Women.’’ In Promoting
Gender Equality http://www.unfpa.org/gender/violence.html
123
University of Ghana http://ugspace.ug.edu.gh
United Nations (2000), ‘United Nations Millennium Declaration’, Resolution adopted
by the General Assembly on September 8, 2000, available at,
http://www.un.org/millennium/declaration/ares552e.htm (November 22, 2014)
United Nations (1989), ‘Violence against women in the family’, ST/CSDHA/2 (New
York: United Nations)
UNICEF (2000), ‘Domestic violence against women and girls’, Innocenti Digest No.
6, Report prepared by Innocenti Research Centre, Florence, Italy
United Nations Commission on Human Security (UNCHS) (2003), ‘Final Report of the
Commission
on
Human
Security’,
http://www.unocha.org/humansecurity/chs/finalreport/ (January 23, 2015)
United Nations Development Programme (UNDP) (1994), Human Development
Report 1994 (New York: UNDP)
United Nations Development Programme (UNDP) (2002), Gender Approaches in
Conflict and Post-conflict Situations (New York, NY.: Bureau for Crisis
Prevention and Recovery)
United Nations Development Programme (UNDP) (2012), Strengthening the Rule of
Law in Crisis-Affected and Fragile Situations, Global Programme Annual
Report (New York, N.Y.: UNDP)
United Nations Development Programme (2013), ‘Citizen insecurity thwarts Latin
America’s
development,
says
UNDP’,
http://www.undp.org/content/undp/en/home/presscenter/pressreleases/2013/11
/12/citizen-insecurity-thwarts-latin-america-s-development-says-undp.html
(December 1, 2014)
United Nations Office on Drugs and Crime (UNODC) (2011), 2011 Global Study on
Homicide: Trends, Context, Data (New York: UNODC)
United Nations Security Sector Reform Task Force (2012), Security Sector Reform:
Integrated technical guidance notes (New York: United Nations)
United Nations Women (2011), 2011-2012 Progress of the World’s Women: In pursuit
of justice (New York: UN Women)
Vaccarino, Viola, Lori Parsons, Nathan R. Every, Hal V. Barron and Harlan M.
Krumholz (1999), ‘Sex-Based differences in early mortality after Myocardial
infarction’, The New England Journal of Medicine, Vol. 341: pp. 217-225
Valji, Nahla, Bronwyn Harris and Graeme Simpson (2004), ‘Crime, Security and Fear
of the other’, in SA Reconciliation Barometer, Col. 2, Issue 1,
http://www.csvr.org.za/old/wits/articles/artvhgs.htm (March 23, 2015)
Vidal, John (2010), Nigeria’s Agony Dwarfs the Gulf Oil Spill. The US and Europe
Ignore
it,
The
Guardian
online,
http://www.theguardian.com/world/2010/may/30/oil-spills-nigeria-niger-deltashell/print (August 15, 2013)
Walby, Sylvia (1991), Theorizing Patriarchy (Cambridge, MA: Basil Blackwell)
Walker, L. (1984), The battered woman syndrome (New York: Springer).
Warr, M. (1984), ‘Fear of victimization: Why are women and the elderly more afraid?’
Sociological Science Quarterly, vol. 6, pp. 681−702.
Warr, M. (2000), ‘Fear of crime in the United States: Avenues for research and policy,’
Journal of Criminal Justice, vol. 4, pp. 451−483.
Warr, M. and C. Ellison (2000), ‘Rethinking Social Reactions to Crime: Personal and
Altruistic Fear in family households’, The American Journal of Sociology, vol.
106, no. 3, pp. 551-573.
124
University of Ghana http://ugspace.ug.edu.gh
Weber, Max (1978), Economy and Society (Berkeley, CA: University of California
Press)
Werthes, Sascha, Corinne Heaven and Sven Vollnhals (2011), ‘Assessing Human
Insecurity Worldwide: The way to a Human (in)security index’, Institute for
Development and Peace Report 102/2011, Universitat Duisburg Essen
West, D. J., C. Roy and F. L. Nichols (1978), Understanding Sexual Attacks (London:
Heinemann)
Wilcox, R. R. (2005), Introduction to robust estimation and hypothesis testing (2nd ed.)
(Burlington, M. A.: Elsevier)
Will, J. A. and J. H. McGrath (1995), ‘Crime, Neighborhood Perceptions, and the
Underclass: The relationship between fear of crime and class position’, Journal
of Criminal Justice, vol. 23, pp. 163-176.
Wilson, Elizabeth (1983), What is to be done about Violence against Women?
(Harmondsworth: Pengiun)
Wilson, J., and G. Kelling (1982), ‘Broken windows, Atlantic Monthly, vol. 211, pp.
29−38
World Bank (2001a), World Development Report 2000/2001: Attacking Poverty (New
York: Oxford University Press)
World Bank (2002), ‘Ghana: Strategic Country Gender Assessment’, Briefing note for
the
Ghana
Poverty
Reduction
Strategy,
http://siteresources.worldbank.org/EXTAFRREGTOPGENDER/Resources/G
hanaCGA-R.pdf (March 11, 2015)
Wrigley-Asante, Charlotte (2008), ‘Men are poor but women are poorer: Gendered
poverty and survival strategies in Dangme West District of Ghana’, Norwegian
Journal of Geography, vol. 62, pp. 161-171
Wyant, B. (2008), ‘Multi-level impacts of perceived incivilities and perceptions of
crime risk on fear of crime,’ Journal of Research in Crime and Delinquency,
vol. 45, pp. 39−64.
ONLINE RESOURCES
‘Independent-Sample
T-Test
Using
SPSS’,
Laerd
Statistics
Online,
https://statistics.laerd.com/spss-tutorials/independent-t-test-using-spssstatistics.php
125
Clement Sefa-Nyarko