Similarly, scholars have found a somewhat inconsistent relationship between women's level of participation in decision-making processes in household and her likelihood of experiencing IPV from husband or partner. A summary of these findings seem to indicate that participation in household decision-making could either increase or decrease the likelihood of a woman experiencing IPV. This assertion is suggested from Hindin, M. J., & Adair, L. S. (2002) study in the Philippines and Rahman et. al (2011) study in Bangladesh. In the study by Rahman et. al (2011) in Bangladesh, it was found that women's participation in household decision-making processes increases her risks of experiencing IPV from husband/partner. According to Rahman et. al (2011), a woman's likelihood of suffering from IPV “increases with increase of number of participation in household decision-making” (Rahman et. al, 2011). Although somewhat similar, the study by Hindin, M. J., & Adair, L. S. (2002) presents a more challenging finding. According to Hindin, M. J., & Adair, L. S. (2002), IPVAW seemed to increase with increase in the number of decision-making areas dominated by men. However, in households where women dominated the household decision-making processes, they were also equally more likely to suffer from IPV from husbands/partner to (Hindin & Adair, 2002). This double-edged result needs more focused investigation using similar methodology and tested across countries as sought in this paper.

2.7 The Consequences and Aftermaths of IPVAW

The consequences of wife beating are enormous. Esere, Idowu, Durosaro and Omotosho's (2009) study on why partners tend to rape in relationships involved twenty-two non-randomly selected married women, aged 22-40 years, who had been victims of IPR and VAW in Nigeria.. Their result shows that many of the respondents had experienced several levels of IPR: Nearly one-third had experienced body harm; more than two-fifth persistent headache; about two-fifth insomnia, one-tenth morbid fears; another one-tenth thoughts of suicide. In all, nearly half developed hatred for men (Esere et al., 2009).

Furthermore, victims of IPV are less productive and efficient at the workplace. In an survey conducted by Wathen, MacGregor and MacQurrie (2015) among a large sample of 8,429 Canadians, it was found that more than one-third of the respondents reported being victims of domestic violence, and about one-third of these reported that domestic violence often hamper their ability to get to work. Most of the respondents complained that domestic violence affected their performance on the job adversely; meanwhile, nearly all the respondents agreed that domestic violence has negative effects on victim's job performance and efficiency (Wathen, MacGregor and MacQurrie, 2015).

To this end, this study will attempt to add to previously existing knowledge in literature on IPV in Nigeria especially with a view to ameliorating the plight of women suffering from IPV and better preserve both the rights and potentials of women in Nigeria.

CHAPTER 3: THEORETICAL FRAMEWORK

3.1 INTRODUCTION TO RELEVANT THEORIES

There is no single universally accepted theory that can singlehandedly explain all aspects of IPV against women (Larsen, 2016). However, a synthesis of relevant theories employed in previous literature could help provide a more encompassing, robust and detailed understanding of the phenomenon of IPV against women. Such dimensional is intrinsically need in this study, especially considering its main objective - to investigate consistent factors exposing women to IPVAW in the three selected sub-Saharan Africa countries vis-à-vis the Gambia, Nigeria and Sierra Leone. To this end, five most relevant theories which seem to encapsulate most (if not all) possible factors of IPV used in previous literatures, including the three novel potential factors of IPVAW introduced in this paper are discussed in greater detail hereunder.

3.2 The Bio-Psychosocial Model

This theory attempted holistic model of IPV against women using a synthesis of physical (biological), mental (psychological) and social (or societal) factors potentially associated with instances and prevalence of IPV against women (See McKenry, Julian, & Gavazzi, 1995).

The model emphasizes the potential roles of biological factors (such as high levels of testosterone in males, etc.), psychological factors (such as spousal egocentrism, need for gratification, jealousy, hostility, lack of or declining satisfaction in the union, etc.) and social factors (such as levels of social stress e.g. unemployment, poor income, poor quality of marital relationship, lack of social help, and so on) in inflaming male perpetrated violence against women.

According to this theory, egocentric, jealous, poor, unemployed, and alcoholic husbands would be more likely to perpetrate instances of IPV against their female partner compared to their counterparts whose husbands have better psychological, social and economic living conditions

This theory can, among others, fuel the hypothesis that men who are unemployed or have no income will be more likely to manifest acts of violence against their intimate female partners than men who are employed or who earn steady income.

THE BIOPSYCHOSOCIAL MODEL OF INTIMATE PARTNER VIOLENCE

Figure: Bio-psychosocial model of domestic violence

Source: Dr. George L. Engel, Bio-psychosocial Theory

However, this theory is not conclusive in determining such relationships, as argued in Atkinson et. al (2005), gendered relative status, rather than individual status, posited by the bio-psychosocial model, seem to be more important in predicting instances, prevalence and/or potentialities of perpetrating IPV against women by husband/partner (see Atkinson et. al, 2005 and Aizer, 2010,). Hence, the need to embark on this comparative analysis to further understand the relative gendered-power dimension powers imperative; this underscores the essence of the Marital Power theory as next identified and briefly expatiated.

3.3 The Marital Power Theory

Much relevant to the objective of this study, the Marital Power theory, as popularized in the works of Jewkes (2002) and Vyas & Watts (2009), posits that in every intimate relationship, spousal power relations is very important. It stresses three important stages of power relations:

Power bases:,

Power processes, and

Power outcomes.

Power bases considers the human capital of the individual spouse coming into the relationship such as their individual economic, educational and material power resources. Power processes refer to the interaction of the power bases such that, in the course of the relationship or interactions, the individual power status faces comparisons, one to the other. This could occur in the two partners' actions, “negotiations, assertiveness [as well as in their] problem-solving” skills. The result of this brings a “Power outcome” which actually reveals who is dominant in the power interactions (power processes). The partner who lacks power becomes dominated and more likely to be the victim of IPV.

This theory provides the basis for testing at least three important variables as correlates of IPVAW vis-à-vis (a) Spousal relative income (whether women earn more, less or about same as husband or partner), (b) Spousal relative age (also called “Age Disparity”- it tests whether woman is older, younger or as old as husband), and (c) Relative educational status (centers on who is more educated between woman or her husband/partner? Or, do they (woman and man) have the same level of education?). Understanding the relationship between women's experience of IPV and her relative personal socio-economic statuses (power bases) compared to that of their husband has been a subject of varied interests (Jewkes, 2002; Vyas & Watts, 2009; Aizer, 2010; Abramsky et. al, 2011; Dalal, 2011; Yoshikawa et. al, 2014; Kwagala et. al, 2013; Ola, 2017).

From the foregoing therefore, using spousal relative income, the main question it poses therefore include “Are women who earn more or about same as husband experience less likely to experience IPV from him compared to their counterparts who earn less than him?” This again is one of the main hypothesis in this paper. Since, women who are less economically empowered manifested higher likelihood of experiencing IPV from husband (Kwagala et. al, 2013) in Uganda, but lesser likelihood in Bangladesh (Dalal, 2011), it was logical to hypothesize that “Women who earn less income than husband would manifest higher likelihood of suffering from IPV compared to women who earn more or about same income level as husband”.

This investigation is very fundamental in the African setting especially with high rate of unemployment and increase in vulnerable employment up to 70% in Sub-Saharan Africa (ILO, 2013, p. 2) on the one hand, coupled with an increasing levels of female education (Diop, Coll-Seck & Duarte, 2016) and attendant increase in chances of securing white-collars jobs across the continent. This enables a fraction, thin though, of the women to earn about same, or more income, than husband in recent times.

Another relative factor worthy of note is relative age, also called age mixing. The question it poses is, “What is the relationship between woman's age relative to her husband and her likelihood of experiencing IPV from him” More precisely, “Do women who are older than their husbands more likely to experience IPV than those who are younger or about the same age as he?”

Finally, this theory also helps to assess the potential influence of spousal relative highest educational level, that is, whether women who are more educated than husband/partner are more or less likely to experience IPV than women who have lower or about same level of education as him.

While the strength of this theory lies in its comparability of spousal relative statuses as argued by (Atkinson, 2005; Aizer, 2010), it failed to reveal individual and community level factors of IPVAW. To complement the lapses, we employ the Socio-Ecological Model of Heise (1998) which is more robust and includes five important levels of analyzing IPVAW.

3.4 The Socio-Ecological Model

Perhaps the most applied theory of IPV since the late 1990's as founded and most prominent in the work of Heise (1998), the socio-ecological model ties happenings within the family to the families' social environment (Heise, 1998). According to Okenwa-Emegwa et al., 2016), every family is situated within a community in which certain cultural traits prevail. People act therefore based on how their social environment influences them to. IPV against women will be high within communities where husbands believe they are the “lord of the household” and are expected to bring the wife into subjection by any means possible, usually involving violence or forms of marital control. On the other hand, IPV will be low in communities that believe in equality of partners.

Africa is dominated by patriarchal societies where the man socio-culturally believed to be the “head” of the home and thus saddled with the power to control his wife's (or wives') and children's behavior (Makama, 2013). Thus, community-level acceptance of IPV might reflect on the attitude of the individual towards IPV - a phenomenal nexus the French Sociologist and one of its founding fathers, Émile Durkheim (1857-1920), has termed “Social facts”.

This theory provides a need to measure the prevalence of attitudes towards IPV across regions per country as well as see the nexus between this attitude and actual experience of IPV. What is the current levels of acceptance of IPV among women by region in each country? Do women who justify IPV tend to experience IPV more than those women who refuse to justify it? What exactly is the relationship between women's attitude towards IPV and its actual experience? And, more importantly, how consistent is this relationship across the three countries?

In addition, the Social Ecological Model posits five levels of analyses of IPV and domestic violence: Individual Spouse/Personal level, Wife-Husband Relationship Level, Family or Community Level, Society/Regional, Larger Society Level (see Figure 2). These five levels of analyses are very important in the analyses of factors discussed in this paper.

Figure : The Social Ecological Model of Heise (1998)

3.4.1 Levels of Social Ecological Investigations and Relevance to Current Study

Individual Spouse/Personal level: Here four important individual factors are brought limelight and made subject of scientific speculations as potential factors of IPVAW. These four factors are tested in this study and their consistency evaluated across countries and beyond cultural and national boundaries. The factors are: spousal (individual) highest level of education; age; parental history/experience of violence (if father beat mother) and, husband's use of alcohol.

Using a woman and her partner's highest level of educational attainment, two important questions are fundamental: “Does a woman's level of education influence her probability of experiencing IPV from husband/partner?” And, on the other hand, “Does a man's level of education influence his probability of perpetrating IPV against his female partner?” And, “How consistent is the relationship between spouse's individual highest level of education and a woman's likelihood of experiencing IPV from husband/partner”.

Another important individual factor tested in this study is spousal individual age. Grouping the women into seven conventional demographic age brackets of 5-year interval, the central questions are “What age category of women is most, more or least likely to experience IPV from husband?” Likewise, using the same categories for the husband/partner, we ask, “What age category of men are most, more or least likely to perpetrate IPV against their female partner?” And, “how consistent is this relationship across the three countries?”

The third individual factor bothers on the parental history of violence. The ensuing question is, “what is the relationship between a women's parental experience of violence and her actual exposure to IPV from husband/partner?” Put succinctly, “Do women whose father beat their mother more likely to also experience IPV from their own husband/partner compared to women whose father did not beat her mother”. Finally, “How consistent is this relationship across the countries studied”? It is instructive to note that this is one of the main hypothesis of this paper because of its novelty among others.

Finally, we examined the relationship between women's experience of IPV from husband/partner and her husband/partner use of alcohol across the three countries. The main question is, “Are women whose husband/partner drinks alcohol more likely to experience IPV compared to their counterparts whose husband/partner does not drink alcohol? And, how consistent is this relationship across the countries?

2. Wife-Husband Relationship Level: At this level, Heise (1998) proposes family power structure between couple. “Who decides what?” Understanding the power dynamics could help extend our knowledge of relational factors of IPV. It should be noted that this level of analysis presents a theorizing somewhat similar to the “Marital Power Theory” described afore. However, the emphasis here is on family power structure, dominance and decision-making process, rather than the comparative or relative status as sought in the Marital Power Theory. Thus, using the Social Ecological Model of Heise (1998), at this level of analysis, this paper tests how woman's participation or non-participation in household decision-making process in three important family issues could be associated with her experience of IPV. The main questions are: “Do women who participates in decision over how the money she earns is spent have higher or lower likelihood of experiencing IPV than women who do not participate in such decision”? Likewise, “Do women who participates in decision over how the money her husband/partner earns is spent have higher or lower likelihood of experiencing IPV than women who do not participate in such a decision? Finally, women's participation in large household decision-making process is crucial.

3. Community or Macro Level: At this level, focus shifts to understanding the influence of structural factors such as unemployment, wealth index and residence type. Residence type here refers to the rural-urban residential dichotomy. Thus, we investigate the association between each of these and IPVAW experience: (a) household wealth index between poor, middle and rich households; (b) residence type - whether rural or urban residence, and (c) employment type.

4. Society Level: At this level also, more macro factors such as societal attitude towards IPVAW was tested. In societies where violence against women is granted and accepted as a norm, more women are likely to suffer from IPV. Several studies have reported this as a potential risk factor of IPV (Heise, 1998; Abramsky et. al, 2011; WHO, 2012; Garcia-Moreno et. al, 2006; Ola, 2017). Diop, Coll-Seck & Duarte (2016) argued that the intergenerational dimension of acceptance of IPV and actual occurrences in relationships. According to them, women whose mother manifest acceptance of IPVAW as justified are more likely too to be tolerant towards IPVAW compared to women whose mother did not justify IPV (Diop, Coll-Seck & Duarte, 2016). However, a study by Yoshikawa (2014) in Nepal has shown that rather than the woman's, it is the husband or male partner's attitude towards IPVAW that mattered. Women were beaten in Nepal whether or not they justified IPV for any reason. This needs verification. To this end, attitude towards IPV is employed as one of the main hypotheses of this paper. The question it raises thus is, “Are women who accept that IPV from husband or partner could be justified for some reason more likely to experience IPV compared to women who do not justify IPV for any reason?” How consistent is the relationship between woman's acceptance of IPVAW as somewhat or very justified and her likelihood to experience IPV from her husband or partner across all countries studied?

5. Larger Society Level: This emphasizes understanding the prevalence of IPVAW in society from the perspective of absence of legal frameworks to curtail the vice in the particular society. Likewise, the institutional frameworks that tend to support, justify, consolidate women subordination and justify men's perpetration of IPV also come into limelight at this level. A useful example is the Nigeria Penal Code system, enforced mainly in Northern part of the Country which tend to justify a man for beating his wife “in as much as he does not inflict heavy damage on her”.

3.5 The Social Learning Theory (SLT)

As developed by Albert Bandura (1971), SLT posits that human behavior is learnt by observing and replicating the behavior of others. This is what he called “Learning by direct experience” (Bandura, 1971, p.3). Applied to the study of IPV, we children tend to replicate their parents, their first agent of socialization and role models (Bandura, 1969). Studies have shown that women who experienced violence in childhood were likely to experience violence in relationships in adulthood. Although this nexus is not so clear, this study will test whether this position is consistent across all regions and countries.

3.6 CONCEPTUAL FRAMEWORK

FIGURE 5: COMPLEX THEORETICAL/CONCEPTUAL FRAMEWORK OF IPV AGAINST WOMEN

3.6.1 Explanation of Conceptual Model

The diagram in the Figure above attempts a summary conceptual description of the complex nature of Intimate Partner Violence Against Women featuring a heuristic conceptual synthesis of four most widely used theories of domestic violence:

a. Bio-social model as expounded by McKenry & Gavazzi, 1995

b. Social ecological model of Heise (1998)

c. Marital power theory of Jewkes (2002) and Vyas & Watts (2009)

d. Social learning theory of Albert Bandura (1971)

For sake of emphasis, Figure 5 reveals 5 factor levels (A - E) that could interact meaningfully and be subjected to scrutiny to understand the complex nature of IPVAW. These five levels are hereby discussed in this paper with a view to answering the research question, that is, which of these five-level factors are persistent beyond national and cultural boundaries? Each of these factors have been described in details earlier in this chapter. However, suffice it to mention that the synthesis of the theories described were heuristically considered possible by the author as diagrammatically presented above.

Whereas the Bio-psychosocial model of McKenry, Julian and Gavazzi (1995) emphasized the need to consider biological and psychological factors (including social skills) of individual partners, we could argue that this is in tandem with the first (and somehow second) level of factors later introduced by Lori Heise in 1998. This is represented in the diagram at two levels - A, B. Meanwhile, the last factor mentioned in McKenry, Julian and Gavazzi (1995), the social factors are somewhat mostly related to the 4th and 5th level of Heise (Society and larger society). And, whereas the Bio-psychosocial model equally emphasized the importance of social skills, these are less defined in comparison with Heise (1998) second level - Relationship level. The theory well devoted to this would be the Marital power theory as expounded by Jewkes (2002) and Vyas and Watts (2009). Hence, the Marital power theory presents a detailed explication of the “Relationship” (the second) level of Heise and advances it further. The patterns are well described and further advanced in this paper by considering 5 relational factors - spousal relative income, relative educational level, relative age, negotiating powers in terms of decision over how to spend individual or family income (of husband/partner, of the woman herself), decision over woman's healthcare, and decision over large household purchases (see Figure 5).

The last theory applied in this paper, the Social learning theory of Bandura (1971), which when applied emphasizes the examining the import of childhood experiences of individual spouse especially spousal parental history of violence - “if father beat mother” is captured under Heise (1998) first level - individual factors.

Beyond the three, Heise theory is mostly significant for its explication on the potential import of the societal and large society stage - levels four and five.

All the main concepts and propositions of each theory have been synthesized and diagrammatically presented in Figure 5 above. They are all tested in this paper in each of the selected African Countries - the Gambia, Nigeria and Sierra Leone for their consistencies.

CHAPTER FOUR: METHODOLOGY: DATA ANALYSIS AND RESULTS

Research Design

4.1 The Bio-Psychosocial Model

4.1.1 Database and Sample Size

Given the research interest of this study, to examine factors consistent in exposing women to IPV across three selected African countries, I employed nationally representative Demographic and Health Survey (DHS) data of the Gambia (2013), Nigeria (2013) and Sierra Leone (2013). The DHS is a Cross Sectional quantitative survey conducted usually every five years and is supported with funds from the United Stated Agency for International Development. The DHS covers issues related to “population, health and nutrition” and domestic violence (DHS, 2018). The DHS uses an internationally standardized questionnaire thus making comparison among countries possible.

4.2 Sampling Procedures and Sample Size

The DHS usually involved different sampling procedures based on the country where the survey is conducted. However, in all, each country has developed its sampling frame which are usually borrows its sampling frame from the latest population and housing census in the Country with or without necessary modification where deemed necessary. This general statement holds true in all the countries studied in this paper. Women aged 15-49 years are selected in households where interviews are conducted with the aid of well-trained research assistants. However, for inclusion in the domestic violence module, only one woman per households were selected and this is done after all interviews have been conducted. The domestic violence questionnaire is usually treated last in the household in most DHS. Women who participate in the module are the only ones who get to know about such an interview. This is meant to provide high level confidentiality and privacy. Interviews are usually either suspended till another time or stopped abruptly if privacy could no longer be ascertained. Such incomplete responses are excluded from the data but counts of such occurrences are usually reported. This is the general principle in most DHS and they hold true in these countries studied in this paper. Note that women in some “collective housing units” e.g. prison, hotels, hospitals, camps or boarding houses were usually not included in the DHS samples (GBOS & ICF International, 2014; NPC & ICF International, 2014; SSL & ICF International, 2014).

In the Gambia, the 2013 GDHS borrowed from the already existing sampling frame used in the last 2003 general population and housing census (GBOS & ICF International, 2014). More precisely, the 2013 GDHS employed a stratified random sampling which involved two stages: First, each of the LGAs were stratified into rural and urban. This does not include Banjul and Kanifin which are entirely urban centres (as shown later). Each residence was then subdivided into enumeration areas (EAs) in proportion to size. In all, 218 EAs were identified. Each EA consisted of Primary Sampling Units which in turn are independently selected. Households in each of the EAs were listed. The second stage involved the selection of 25 households per EA through systematic random selection. In each household, women aged 15-49 who spent a night before in the household were considered eligible for interview in the general DHS module. In all, the 2013 GDHS involved a total of 10,233 women among whom 4,525 (44%) were selected for inclusion in the domestic violence study. Finally, 1,287 (28.45%) were never-partnered, leaving 3,237 (71.55%) ever-partnered women. These are included in this study.

In Nigeria, the sampling procedure consists of a three-stage sampling technique based on the last 2006 general population and housing census. Beginning with each of the first from States level to the Local Government Areas, second to localities (communities), then to Census Enumeration Areas or Primary Survey Units in which households are then selected (45 household per cluster). In all, the Stratified 3 stage cluster design helped to identify 904 PSU's (372 - Urban; 532 were Rural), 40,680 households (involving 943 interviews per State). Total number of all women respondents was 38,948 (15-49 years). Among these, 27,634 (constituting 71%) were successfully interviewed in the domestic violence module after they had given their informed consent in tandem with the WHO (1990) recommendation for research. However, about 6,438 of the women (23.3%) were never union. This further brings us to a total sub-sample of 21,196 ever-partnered women who were examined in this study.

In Sierra Leone, the 2013 SLDHS employed a two-stage stratified random sampling procedure much similar to the 2013 GDHS. The 2013 SLDHS borrowed from the 2004 population and housing census. First, Primary sampling units (PSUs), or clusters, were selected from the existing EAs. There were 435 clusters consisting of 277 rural and 158 urban selected proportionally by size. In all, a total sample of 16,658 women were included in the 2013 SDHS. Among them, only 5, 185 were included in the domestic violence module. Furthermore, 1,046 (20.19%) of these were never in union. Thus, we have a total subsample of 4,138 women were successfully interviewed.

Domestic violence data are gathered with the aid of internationally standardized questionnaires with core questions based on the Conflict Tactic Scale of Strauss (1990).

Table 4.1: Showing Sample Size in all and respectively to ever-partnered women.

4.3 Measurement, Coding Variables and regression Models

The dependent variable in this study is “Woman's experience of Intimate Partner Violence from husband/partner” (IPVAW). However, the dependent variable is derived from a combination of woman's responses of ever experiencing any physical, sexual or emotional violence from husband or partner. Hence, women who have experienced any of these are classified as having experienced IPV. It should be noted that there are two forms of coding of this dependent variable based on the time or period the form (s) of IPVAW occurred: this could be since age 15 (measuring lifetime prevalence) or within 12 months preceding the survey (12 months prevalence).

All women who reported to have experienced IPV in the last 12 months preceding the survey were also automatically considered to have witnessed IPV since age 15; however, not all women who have witnessed IPV since age 15 can be rightly said to have witnessed IPV within the last 12 months. Thus, this important differentiation gave rise to two possible outcomes (two dependent variables): “Ever IPV” - representing “women's experience of IPV since age 15” and “IPV12Months” - which represents “women's experience of IPV in 12 months preceding survey”. This uniform measurement was applied in all the three countries thus making results very comparable.

Measuring each component of the Dependent Variable:

Each of the components of measuring IPVAW: physical, sexual and emotional IPV, are measured as follows:

1. Physical, emotional and sexual abuse from husband or partner. The questions were limited to woman's experience with the current or most recent husband or partner (NPC, 2014). An affirmative response to any of the questions in any of the three components implies an IPV experience by the respondent:

Physical Violence: The women were asked: “Does (did) your (last) husband/partner ever do any of the following things to you? -

(i) Push you, shake you, or throw something at you? (ii) Slap you? (iii) Twist your arm or pull your hair? (iv.) Punch you with his fist or with something that could hurt you? (v) Kick you, drag you, or beat you up? (vi.) Try to choke you or burn you on purpose? (vii) Threaten or attack you with a knife, gun, or any other weapon?

(b) Emotional Violence: The women were asked: Does (did) your (last) husband/partner ever? - (i) Say or do something to humiliate you in front of others? (ii) Threaten to hurt or harm you or someone close to you? (iii) Insult you or make you feel bad about yourself?

(c) Sexual Violence: The women were asked: “Does (did) your (last) husband/partner ever do any of the following things to you? - (i.) Physically force you to have sexual intercourse with him even when you did not want to? (ii) Physically force you to perform any other sexual acts you did not want to? (iii) Force you with threats or in any other way to perform sexual acts you did not want to?

Note: Any affirmative “Yes” to any of the questions = 1, and having “No” in all = 0, per respondent.

Measuring 12 Months Prevalence of IPVAW (1 Year prevalence)

When a woman responded in the affirmative (a “Yes”) to any of the questions, she is further asked about the frequency from a list of four options: (a) Never (b) Often (c) Sometimes (d) Yes but not in the last 12 months. Options (b) and (c) indicate 12 months prevalence and are coded = 1, otherwise “0”

The Independent Variables

The research objective of this paper has ideally necessitated that a number of variable be tested to ascertain their relationship to woman's experience of IPV in each country. However, with information from theories and empirical literatures reviewed, five main variables have been identified and these formed the research hypotheses. Four of these are commonly reported, one is novel; The four commonly reported and tested are: Parental history of violence (that is, if woman's father beat her mother), Attitude towards IPV (if woman has internalized usually patriarchal societal norms allowing wife beating), and husband's alcohol consumption, marital power (measured using woman's Relative income compared to husband's). The fifth is husband controlling behavior.

(1) Parental history of violence: If father beat mother = 1, if no = 0.

(2) Attitude towards wife beating: This measures women's attitude towards IPV and is a good index for measuring tolerance. It should be clarified that acceptance of IPV does not connote that women wanted to be abused. Instead, measuring this could reveal woman's perception of her self-esteem, rights and privileges as a member of a society or conjugal rights if married. Responses were coded in two different ways. First, out of five hypothetical scenarios. women were asked if it was right for a man to beat his wife if she - “goes out without telling him”, “neglects the children”, “argues with him”, “refuses to have sex with him” or “burns the food”? Women who responded a “Yes” to any of these questions were recoded as “1”: only those who rejected all were coded “0”. All responses were measured in two ways: (i) first, into dummy variable - acceptance of at least one = 1, rejection of all = 0. Responses were further decongested among those who agreed to any of the reasons based on the frequencies of acceptance. Women who agreed with any two were regarded as seeing IPVAW as “somewhat justified”. Acceptance of three or more reasons were labeled the “very justified” group. Thus three categorical variables coded as: “Not justified”=0, “Somewhat justified”=1, “Very justified”=2

(3) Woman's relative income to the husband's/partner's: Four categorical variables: “less than husband”=0, “about same” =1, “more than husband”=2, “husband brings in no money” = 3 (see Aizer, 2010).

(4) Husband control - This is derived from a synthesis of five variables. Husbands are c=said to control wife is he - (a) is jealous if she talks to other men (b) always accuses her of unfaithfulness (c) does not allow her to meet her female friends (d) always willing to know where she is (e) tries to limit her contact with her family. A “Yes” (coded as “1”) response to any of these shows that the man exercise forms of controlling behavior thus limiting the woman's dignity of human person. When the response is “No” in all scenarios, then there is no husband control and it is coded “0”.

Other important variables controlled for:

A huge number of other possible variables were controlled for, especially those found to have been tested in previous literatures I reviewed, were also tested in this paper to provide a more robust understanding and explanation in the dimensions of IPV against women in these countries and to prevent the omitted variable bias. Such factors include:

Personal factors: (a). Woman's highest educational attainment and (b) Husband's highest educational attainment - both involved 4 categories: No Education=0, Primary=1, Secondary=2, or Higher=3). Woman and husband's age group: 15-19=1; 20-24=2; 25-29=3; 30-34=4; 35-39=5; 40-44=6; 45-49=7. (c) religious affiliation of woman

Relational factors (1): Personal factors compared such as Relative age: woman is younger=0; about same age (if they are in the same age group=1, woman older=2; Relative education: woman is less educated =0; about same level of educational attainment (if they are in the same level=1, woman is more educated=2.

Relational factors (2): social or relational skills - decision making participation: (a) who decides how to spend the money woman earns - woman alone=1; woman and husband=2; husband/partner alone=3; someone else =4. (b) who decides how to spend the money the husband makes - woman alone=1; woman and husband=2; husband/partner alone=3; someone else =4; husband earns no money =5; (c) who decides woman's healthcare - woman alone=1; woman and husband=2; husband/partner alone=3; someone else =4; (d) who decides large household purchases - woman alone=1; woman and husband=2; husband/partner alone=3; someone else =4. This will show the importance of relationship in planning.

Household/Community factors: (a) desire for more children - wants within two years =1; wants after 24 months=2; wants but not sure of timing=3; Not decided=4; wants no more=5; sterilized/infecund=6; (b) woman's number of living children and current pregnancy - None=0; 1 child=1; 2 children=2; 3 children=3; 4 or more children=4). (c) Household wealth index: Five categories (Poorest=1; Poorer=2; Middle=3; Richer=4; Richest=5) (d) Forms of union/presence of other wives = monogamy=1; polygyny =2. This is further decongested based on the number of wives a man has including respondent - Only respondent =1; two wives=2, 3 wives=3; 4 or more wives=4.

Societal factors: (a) Region - administrative or demographic delineation where household is located; (b) Residence type - urban=1; rural=2.

Interaction terms: (a) Education and residence are interacted to give the import of highest level of education and place residence on IPVAW.

Thus, a minimum of twenty variables were tested in all in each and across all countries. The model fitted are described below at 95% Confidence interval.

4.4 Formal Logistic Models

The following seven (7) binary logistic models were fitted for this study. The variables are:

Hypotheses models: First, independent tests.

Ln(P/1-P) = B0 + B1.Somewhat_justified + B2.Very_justified + E

Ln(P/1-P) = B0 + B1.Earns_Same + B2.Earns_More + E

Ln(P/1-P) = B0 + B1.alcohol + Error Term

Ln(P/1-P) = B0 + B1.husbandcontrol + Error Term

Ln(P/1-P) = B0 + B1.fatherbeatmother + Error Term

2. All variables in the hypothesis tested:

Ln(P/1-P) = B0 + B1.Somewhat_justified + B2.Very_justified + B3.Earns_Same + B4.Earns_More + B5.alcohol+ B6.husbandcontrol + B7.fatherbeatmother + Error Term

4.5 Results and Findings

Table 4.2: Age Distribution of Respondents

Total number of women across all countries is 28,571

The mean age of the women across countries shows that they are in the age range of 30-34, compared to their husband or partner who are in the age range of 40-44. One striking revelation here is the 10-year age gap as distributed on the average comparing the mean of the women and that of their husbands or partners. One reason may be age-long fact that older men marry younger women. Considering the minimum and maximum, while strictly, only women of the reproductive age group (15-49 years) were included into the survey, many guessed the age of their husband or partner to be up to 95years. For instance in the Gambia, about 4% of the partners are in the older than 64 years old. In all, the age standard deviation of women is less than 9 years in all compared to men's up to 13 years.

Table 4.3. Summary statistics: Demographic characteristics of respondents

Note: (a) is also included in the Christian category; (b) the Gambia data reported only two religious affiliations - Christian and Islam.; (n/a) Not available

There is a somewhat fair share of the women by residence in the Gambia except that more than 60% of the women are from the rural areas from Nigeria and Sierra Leone. Majority of the women profess to belong to the Islamic religion in the Gambia and Sierra Leone. In Nigeria likewise, more than half (58.5%) belong to the Islamic religion. This large difference in religious affiliation has a root in the history of the three countries as already discussed under the section providing a detailed information in the daily lifestyle and gender relations in these countries.

The three countries provides a relatively similar situation in the household wealth index as about 40% of the women are in poor or poorer homes, about 20% in the middle wealth index, and about another 40% in the rich homes. It is striking to note that regional variations exist. For instance, in the Gambia, two regions - Banjul (the country's capital) and Kanifin are both on the rich to richest wealth index, and have no rural residence. However, surprisingly, Banjul has the highest prevalence of violence against women in the Gambia when decongested by regions.

It is also striking to note that more than one-third of all women respondents in the Gambia (62.9%), and Sierra Leone (72%) have no formal education. However, majority of these women are married (92% in the Gambia, 91.3% in Nigeria and 87.3% in Sierra Leone). One important factor for this could be early marriage (Gender report, 2014).

Table 4.4. Prevalence of key variables in the study

Table 4.4 shows the prevalen...