Você está na página 1de 115

Correlates Of Wife Abuse Among Ever Married Males In Selected Regions In The Philippines

Maria Fidelis Manalo, MD, MSc.


*

Abstract
We describe the marital, social-psychological, social-structural and social location factors associated with wife abuse among married men in selected regions in the Philippines, using data from the Reproductive Health Survey 2000. About eight percent (8.4%) of married men in our sample of 5,507 reported wife abuse. Getting married young and having a history of parental abuse emerge as strong predictors of domestic violence. Lower levels of education in women are associated with a higher likelihood of wife abuse. Income, employment status of the men and their spouse, occupational status discrepancy, and mens educational attainment, do not predict intimate partner violence. Men who object to their partner wanting to improve her education or visit her friends without them and who get their way during disagreements, are more likely to use physical violence. Men who, jointly with their wives or by themselves, usually decide on food matters, are also more likely to give in to wife abuse. Men who predominantly handle the household budget are more likely as well to beat their wives, though the association is not significant. Keywords: Wife Abuse; Risk factors; Philippines; Domestic violence; Intimate partner violence

Department of Community & Family Medicine, Far Eastern University - Nicanor Reyes Medical Foundation, Fairview, Quezon City, 1118, Philippines. Email: mcmanalo@feu-nrmf.ph, lizamanalomd@gmail.com

1
Introduction

The United Nations (UN) Universal Declaration of Human Rights states that everyone has the right to life, liberty and security of person, and that no one shall be subject to torture or to cruelty, inhuman or degrading treatment. The right to privacy gives no one the right to hurt anybody, least of all members of his or her own family. Wife beating is a violation of the womans human rights. Domestic Violence, which affects mainly women, is contrary to the UN Convention on the Elimination of all forms of Discrimination Against Women, to which the Philippines is a signatory (Banatin, 1998). Violence against women (VAW) is a social problem that is widespread. It is more prevalent than is often recognized. It is a potentially fatal problem from which no woman is immune. Abuse of women has historically been a private violation that has been kept within the walls of the family home. This violence has also been historically denied. Crimes against women are events still perceived as anything but crime (KitchenerWaterloo Multicultural Center, 1998). Despite the limited data available, it should be considered as a public health problem because of the severity of effects on survivors and because of the alarming incidents of battering, assault and abuse as reported in media (Pineda-Mercado, 1997). Violence against women is a concept that intersects with the domain of domestic violence since research has shown that violence against women is more likely to happen in a private sphere (Population Report, 1999). Domestic violence can be defined narrowly, as the physical attack of women by their males partners,or more broadly, to encompass repeated verbal abuse, harassment, confinement, deprivation of financial resources, and other forms of psychological violence (UN, 1994). The United Nations also defines domestic violence as violence which occurs within the private sphere, mainly between individuals who are related through intimacy,

The

terminology domestic violence (DV),

wife abuse and intimate partner violence (IPV) are used

interchangeably throughout this study. While domestic violence can include wife abuse, child abuse and elder abuse, the current study focuses only on wife abuse, specifically, on physical violence. In this paper when discussing wife abuse, we used the terms wife, spouse and intimate partners synonymously. Abuse and violence are also used alternately.

blood, or law. (Johnson, 1997) According to the World Health Organization, studies show that an intimate male partner has beaten 20-50% of women at least once. Based on the Philippine National Police (PNP) reports, prevalence rates of physical violence against women aged 18 years old and over were at 28 per 100,000 women in 1999 and it increased to 35 and 39 per 100,000 women in 2000 and 2001 respectively. Highest prevalence rates of violence against women were in CAR, Western Mindanao, Northern Mindanao, and Southern Mindanao. According to the 1993 Safe Motherhood Survey which was conducted by the Department of Health (DOH), 10% of 8,500 respondents between the ages of 15 to 49 years reported that they were physically harmed by a friend or a family member and 3% claimed that they were hurt while pregnant. Central Visayas, Southern Mindanao, Eastern Visayas and Cagayan Valley have the highest prevalence rates of domestic violence. The large discrepancy in the prevalence estimates from the Philippine National Police (PNP, 2000) reports and the Safe Motherhood Survey seems to imply a gross underreporting of cases of violence against women to the police (Bascos-Deveza, 2001). In March to September of 1996, the DOH hospitals reported a total of 1,332 battered women. Cases handled by Lihok-Pilipina, the Womens Crisis Center and Kalakasan show that most victims are from 19 to over 50 years old, 80% have children. The women have been beaten for as short as 2 years to as long as 25 years. About 18% are professionals, 15% skilled workers, and 25% unemployed or unskilled workers. An NGO called Kalakasan that provides hotline services for women victims of domestic violence reported 158 cases of battering in a four-month period in 1992. From 1993 to 1995, the same NGO documented 600 cases that include physical injury from being boxed, slapped and kicked in various parts of the body (Pineda-Mercado, 1997). The WorldSAFE Study on domestic violence involving one thousand women aged 15-49 years among urban Filipino families in Paco, Manila reported a one-year prevalence of spousal abuse or maltreatment of 29%. The study also revealed that the leading types of physical maltreatment were slapping, kicking and beating repeatedly, even at the time of pregnancy (Ramiro et al, 1998).

The former Department of Health undersecretary reported that there is absence of consolidated data on violence against women and children. After assuming a conservative estimate of 10% of the population aged 15-49 years as being affected by the problem, the figure is said to be anywhere from 10-50% of the female population. Extrapolation of this would mean the possibility of 1.5 million women being victims of some form of domestic violence (Pineda-Mercado, 1997). Despite the unacceptably high prevalence of domestic violence, few studies have examined what happens in the man, in the wife and in the household to spark spousal violence. Theories on the causes of partner violence provide a framework for understanding and responding to this phenomenon. Additionally, different theories point to particular variables or risk markers that might alert clinicians to the potential for partner violence in an intimate relationship. Thus, the more integrated and encompassing the theoretical model, the more valid the model for the purpose of predicting intimate violence. The latter approach is not without its critics, however, and currently dispute rages over the relative importance of one framework over another. Early theoretical development on the causes of partner abuse (Gelles & Straus, 1978) identified several theories, organized into three broad categories ---intraindividual theory, sociocultural theory, and social-psychological theory ---that provide guidelines for the risk factors that need to be considered. These models and another broad category of theories (i.e. marital relationship factors) were used to develop a list of candidate variables that were analyzed by multiple logistic regression to determine their association with wife abuse. To attain this aim, secondary data from the Reproductive Health Survey 2000 was used. This research is notable because it is probably the first large-scale investigation to study wife abuse in selected regions in the Philippines by using married men as respondents. While previous studies provided valuable information on the nature and extent of wife abuse in the country, majority of the information was derived from women and from institutionalized and/or known cases of wife abuse. Very few researches provide community-based information, much more of a far-reaching scope, from which estimates of the association between wife abuse and marital, social-psychological, social structural and social location factors among married men in the general population were made. It is hoped that the results of this investigation would provide not only a broader theoretical understanding of the problem of domestic violence but also a resolute identification of

associations between wife abuse and specific risk factors that will serve as the basis for future preventive and rehabilitative actions. Knowledge of the profile of men who beat or hit their intimate partners will help in the development of preventive and therapeutic strategies---on the individual, family and societal levels--- that are relevant, not just for the well being of Filipino couples but that of the whole family as well.

2
REVIEW OF LITERATURE

There have been media emphasis of criminal violence committed by strangers with its resultant widespread fear of such. However, statistical evidence demonstrates that many violent acts occur between people who know one other. How many such acts, and which forms of criminal violence, are most likely to be committed by someone well known to the victim (i.e., a family member, former family member, or an individual in an otherwise close personal relationship or former relationship) is a matter of debate. Over the past two decades, there has been growing recognition of the multitude of health threats, including death, physical injury, and mental trauma, posed by domestic violence.

2.1.

CLINICAL FINDINGS

Family violence refers to the intentional intimidation, physical and/or sexual abuse, battering of children, adults, or elders by a family member, intimate partner, or caretaker.6 Abuse refers to a pattern of behaviors organized around the intentional use of power, including but nor requiring physical violence, by one person for the purpose of controlling another. From the Womens Desk Training Workshop on Violence Against Women developed by the University of the Philippines-Philippine General Hospital (UP-PGH, 1997), domestic violence is defined as any coercive behavior perpetrated by someone who is or was involved in an intimate relationship with the victim. This may include (1) physical abuse any physical harm inflicted on the woman, e.g., slapping, kicking, repeated beatings, boxing, hair-pulling, use of any object that can harm or kill; (2) sexual abuse any sensual act done against the wishes of the woman, e.g. kissing, touching, sexual molestation, rape; (3) economic abuse a refusal to give her basic needs like food, refusal to give money, leaving his debts on the woman, making the woman quit work, depriving her of inheritance; and

(4) emotional/psychological abuse calling her degrading names, threatening to kill her or leave her and the children, accusing her of having a paramour, refusing to talk to her, ignoring her, bringing his querida to the house, denial of sexual contact, etc. For the purpose of this study, wife abuse will refer specifically to physical violence.

2.1.1. The specific signs and symptoms for the diagnosis of wife abuse

Although it is most obvious when abusive men commit physical assault, domestic violence is best understood as an abusers pattern of coercive behavior that serves to establish and maintain power and control over intimate partners. An abusers tactics of control can progress very slowly, making domestic violence difficult to recognize in its early stages. The signs and symptoms of battered woman may range from gross evidence of physical injury to subtle signs and symptoms of chronic stress syndrome. The common types of physical injury that may be associated with battering includes: contusions, abrasions and minor laceration; fractures or sprains, injuries to head, neck, chests, breasts, and abdomen; injuries during pregnancies; multiple sites of injury and repeated or chronic injuries (UP-PGH, 1997). The physician should consider the possibility of abuse when the womans physical examination findings is incongruent with explanation for the injury; if there was delay in seeking medical attention; if pregnant women present with injuries; and, if there are chronic psychogenic complaints.

2.1.2. Other complaints and behavioral signs strongly suggestive of abuse

The stress of ongoing abuse may cause chronic pain, psychogenic pain, diffuse trauma without visible evidence, physical symptoms related to stress, chronic posttraumatic stress disorder, anxiety disorder, depression, feeling of isolation and inability to cope, suicidal attempts, panic attack, sleep and appetite disturbance, fatigue, decreased concentration, sexual dysfunction, chronic headache, abdominal and gastrointestinal complaints, palpitations, dizziness, paresthesia, dyspnea with clear breath sounds, atypical chest pain. The frequent use of prescribed minor tranquilizers or pain medications, frequent clinic visits with vague complaints and symptoms without evidence of physiologic abnormality may give clues to battering (UP-PGH, 1997).

2.1.3. Overall Health Effects of Intimate Partner Violence

Although intimate partner violence results in significant morbidity and mortality, no large, prospective study has been conducted specifically on the health outcomes associated with intimate partner violence. Most studies of the health effects are retrospective, fairly small, and use different definitions of intimate partner violence and different methods of detection. Recent cross-sectional studies (Womens Health, 2002) done in the primary care setting have found intimate partner violence to be significantly associated with reports of more physical symptoms, worse overall mental health, and worse overall physical health (as measured by standardized health survey tools). Notably, studies that measured psychological as well as physical and sexual intimate partner violence found that poor health outcomes were just as significantly associated with psychological intimate partner violence as with physical and/or sexual intimate partner violence. A higher prevalence of substance abuse in victimized patients has been found in cross-sectional studies, but the relationship between substance use and intimate partner violence is complex.

2.2.

MAGNITUDE OF THE PROBLEM

2.2.1. Prevalence of Intimate Violence In a study supported by UNICEF entitled Breaching the Silence, 1,000 cases of domestic violence were documented. Some of the findings include the fact that in those 1,000 cases, women were the victims and that physical assault accompanied by verbal and economic abuse were the most frequent form of abuse done to these women. More than half of all these abuses was perpetuated by male spouses or the womens partners (Pineda-Mercado, 1997). Domestic violence is a recognized social problem at the global level. Population reports concluded an international comparison of prevalence of physical assaults on women by an intimate male partners with the statement that in nearly 50% of population based surveys from around the world, 10% to over 50% of women report being hit or other wise physically harmed by an intimate male partner at some point in their lives (Population Reports, 1999).

In the Philippines, according to the 1993 Safe Motherhood Survey of 8,500 respondents, friend or family member has physically abused 10 %. The survey likewise estimated prevalence of domestic violence at different age groups, educational level and regional location. Prevalence of abuse is higher among younger women, among less educated women and among certain regions in the country. Central Visayas, Southern Mindanao, Eastern Visayas and Cagayan Valley have the highest prevalence rates of domestic violence (Bascos-Deveza, 2001). The census in 20 Department of Health hospitals from March-September 1996 reported a total of 1332 battered women. The Department of Social Welfare and Development reported that from January to August 1996, there were 251 cases of physical abuse. A non-government organization, Kalakasan, Hotline reports that in 1992, there were 158 cases of battered women; in 1996, there were 600 cases. In Baguio Mining area, 50% of women residents were victims of battering (UP-PGH, 1997).

2.2.2. Consequences of wife abuse

On the Children

There is an expanding body of literature that describes the connection between intimate partner violence and child abuse. When cases of child abuse are identified, it has been estimated (by chart review of the mother's chart) that 59% of the mothers of abused children have also been victimized. Straus and Gelles (1990) who have studied violence in the American family for many years, found that in a national random telephone survey of 6002 adults, 50% of the fathers who frequently beat their wives also frequently abused their children. Edleson (1995) had estimated that in "30%-60% of families where either child maltreatment or woman battering is identified, it is likely that both forms of abuse exist." There is compelling evidence that childhood abuse results in long-lasting health problems. Even when children are not directly victimized, they may suffer greatly from witnessing violence. In a study of police calls for domestic assault, a child in 85% of cases witnessed the domestic assaults. It is extremely traumatic for a child to witness his or her parent being abused emotionally, physically, and/or sexually -- often by the other parent. Witnessing violence has been shown to result in emotional and developmental problems

in children, even those in infancy. Children who see their mothers being abused may blame themselves for the violence. They may experience physical complaints such as headaches, stomachaches and other illnesses. They can have nightmares or difficulty sleeping. They may act out their mixed emotions either by being aggressive or self-destructive or by trying very hard to be compliant or passive. These children could grow up believing that it is alright for men to hit women, that violence is a way to win arguments, that it is alright to hit someone if you feel angry or upset, that men are powerful while women are weak, that there are few, if any, negative consequences for abusive acts. These children might feel that they are responsible for the abuse and responsible for solutions.7

Mental health

Cross-sectional studies have found associations between intimate partner violence and an increased prevalence of overall worsened mental health status, depression, anxiety, somatization, and posttraumatic stress disorder. Most studies do not address the distinct contributions of childhood abuse and adulthood abuse nor the effect of the "cumulative dose" or severity of abuse on health outcomes. Because childhood victimization may have long-term psychiatric effect and put an affected person at increased risk for future victimization, this omission may cause confounding. One study did examine the contributions of both childhood and adulthood victimization and found that past childhood abuse and current adulthood abuse were equivalently associated with more physical symptoms, higher scores for depression, anxiety, somatization, low self-esteem, and higher rates of attempted suicide and substance abuse. Women who had experienced childhood abuse and who were also experiencing current adulthood abuse had the highest levels of these poor health outcomes (Womens Health, 2002).

Suicide and homicide

Given the well-documented levels of major depression among battered women, it is not surprising that they think about, attempt, or actually commit suicide. This outcome

10

is most likely to occur when women are overwhelmed with feelings of helplessness, have little support, and fear seeking help from police, medical, or other social services. Stark and Flitcrafts report (1988) indicates that 10% of a sample of battered women attempted suicide and that one half of that group tried more than once. This study also reported that 26% of all female suicide attempts coming to the attention of hospitals are associated with battering and that a large majority of those who attempt suicide had been seen before for battering injuries. A retrospective study done in Sweden documented an 8-fold increase in the risk of a suicide attempt serious enough to require hospitalization in abused women. Another study found that current adult women victims of intimate partner violence were 4 times more likely to have attempted suicide than nonabused women. In the United States, it is estimated that 30% of the homicides of women are committed by their intimate partner or ex-partner (Womens Health, 2002).

Healthcare utilization and cost

It is becoming evident that healthcare utilization is increased in patients who are current or past victims of intimate partner violence. For example, in a Swedish study involving 117 battered women and 117 controls studied retrospectively for 10 years and prospectively for 8 years, healthcare utilization as evidenced by hospitalization due to a broad variety of physical and mental health problems was significantly higher in the battered women at both the 5-year follow-up point and the 8-year follow-up point (Womens Health, 2002). There is also a growing body of data about the costs of victimization. Victimization results in healthcare costs for both mental healthcare and medical care, individual costs because of decreased quality of life for victims and their friends and families, and societal costs due to decreased productivity, increased use of social services and police services, and property damage. Analyses of National Crime Victimization Survey data reported an estimated $1.3 billion (in 1993 dollars) in annual medical costs because of "domestic crime against adults." (Miller et al, 1996). Costs of lost workdays tend to be assessed with medical costs, and estimates of medical costs and lost work productivity as a result of domestic violence run $5-10 billion (Meyer, 1992). The Bureau of National Affairs estimates costs of $3-5 billion from health care premiums and lost work.

11

2. 3. SERVICES FOR PROMOTION OF REPRODUCTIVE HEALTH AND VICTIMS OF WIFE ABUSE


2.3.1. Local Support Services There are numerous helping institutions in the country that provide different services to victims of domestic violence. These service include (1) medical and health services; (2) shelter for women and girls; (3) law enforcement agencies; (4) legal assistance; and (5) counseling hotlines/helpline (National Commission on the Role of Filipino Women, 2001). There is a hospital-based crisis center known as Project HAVEN (Hospital Assisted Crisis Intervention for Victims/Survivors of Violent Environments in 1995). This is government organization (GO) and non-government organization (NGO) undertaking of the Department of Health, National Commission on the Role of Filipino Women (NCRFW) and the Womens Crisis Center (WCC) as the lead agency for the NGOs involved in the project. Project HAVEN is situated at the East Avenue Medical Center in Quezon City. There are also crisis centers and shelters in different regions of the country notably: Ing Makababaying Aksyon (IMA) Foundation in Angeles, Bathaluman in Davao, Lihok Filipina in Cebu, Cordillera Womens Education & Resource Center (CWERC) in Baguio and Womens Center in Bacolod. The Department of Social Welfare and Development (DSWD) manages a shelter for abused women by the Congressional spouses Foundation known as The Haven. Likewise there are hotline services for battered women run by the Womens Crisis Center, Kalakasan at Arugaan and Department of Social Welfare and Development (DSWD). Police stations have set up womens desks initiated by the National Commission on the Role of Filipino Women (NCRFW) in coordination with the Philippine National Police and Department of Interior and Local Government (DILG). Aside from these, there is a COmmunity-Based Approach To Violence Against Women (COMBAT-VAW) being implemented by HArnessing Self-reliant Initiatives and Knowledge (HASIK), an urban poor NGO based in Quezon City and the Womens Legal Bureau (WLB), a legal resource NGO also based in Quezon City. This project adopts a four-pronged strategy of Education and Training, Support Services, Participatory Research, Documentation and Evaluation, and Replication that takes place within the framework of community organizing.

12

2.3.2. Promotion of Reproductive Health: The Role of UN Population Fund (UNFPA) and Department of Health At the 1994 International Conference on Population and Development (ICPD) held in Cairo, Egypt, an international consensus on an unprecedented global plan of action to reduce population growth was forged while promoting social and economic development. At the ICPD, the nations of the world agreed to a sweeping framework for action with goals for making voluntary family planning services universally available; providing those services in a broader context of reproductive health, including safe motherhood and child survival; affirming reproductive rights and other human rights; enhancing the education of girls; improving economic, political and social opportunities for women; ensuring that adolescents have access to appropriate education and services; and recognizing the responsibility men have to care for their families and children. The United Nations Population Fund (UNFPA) is the largest internationally funded source of population assistance to developing countries. Through the United Nations Population Fund, individuals and families receive access to information and services that promote reproductive and sexual health, including family planning and universal reproductive rights. UNFPA provides a broad range of services on a strictly voluntary basis that enable couples and individuals to make informed choices about family size and spacing. UNFPA recognizes that limited access to reproductive/sexual health information and services has a negative impact on general health and cripples individuals' ability to make informed choices about their family size, which in turn affects their social and economic welfare. UNFPA's approach to population focuses on people, not demographic targets. Its mission to stabilize population growth is realized by empowering couples and individuals, particularly women, through various means: increasing and prolonging girls education, providing access to basic and improved health care services, and expanding women's employment opportunities (United Nations Foundation (UN) Issues Women and Population,2002). The Department of Health is actively collaborating with the UNFPA to promote the Reproductive Health program of the government.

13

2.4.

UNDERSTANDING VIOLENCE AGAINST WOMEN

2.4.1. Power and control vs. Equality wheel

Violence towards Women is about power and control. Table 2.4.1. compares and contrasts what happens in an abusive versus nonabusive relationship (United Nations Foundation (UN) Issues Women and Population, 2002). Table 2.4.1. Power and control vs. Equality wheel Power and Control wheel Equality wheel Physical abuse: Non-Violence Using a weapon against her, Negotiation and Fairness: beating, throwing her down, Seeking mutually satisfying twisting arms, tripping, biting, resolutions to conflict; accepting pushing, shoving, hitting, slapping, change; being willing to choking, pulling hair, punching, compromise. kicking, grabbing. Non-Threatening Behavior: Isolation: Talking and acting so that she feels Controlling what she does, who she safe and comfortable expressing sees and talks to, where she goes. herself and doing things. Emotional Abuse: Respect : Putting her down or making her Listening to her non-judgmentally; feel bad about herself, calling her being emotionally affirming and names. Making her think she's understanding; valuing opinions. crazy. Mind games. Trust and Support: Economic Abuse: Supporting her goals in life; Trying to keep her from getting or respecting her right to her own keeping a job. Making her ask for feelings, friends, activities and money; giving her an allowance, opinions. taking her money. Honesty and Accountability: Sexual Abuse: Accepting responsibility for self; Making her do sexual things acknowledging past use of against her will. Physically violence; attacking the sexual parts of her admitting being wrong; body. Treating her like a sex object. communicating openly and Using Children: truthfully. Making her feel guilty about the Responsible Parenting: children, using the children to give Sharing parental responsibilities; messages, using visitation as a being a positive non-violent role way to harass her. model for the children. Threats: Shared Responsibility: Making and or/carrying out threats Mutually agreeing on a fair to do something to hurt her distribution of work; emotionally. Threaten to take the making family decisions together. children, commit suicide, and report Economic Partnership: her to Welfare. Making money decisions together; Using Male Privilege: making sure both partners benefit Treating her like a servant. Making from financial arrangements. all the "big" decisions. Acting like

14

the "master of the castle." Intimidation: Putting her in fear by: Using looks, actions, gestures, loud voice, smashing things, destroying her property.

2.4.2. Reasons why a woman stays in an abusive relationship

The University of the Philippines-Philippine General Hospital Womens Desk training manual (1997) emphasizes the importance of understanding that there are many reasons why women stay in battering relationships. Some reasons listed are: economic dependence, parenting concerns, religious beliefs, family pressure, fear, loyalty, pity, rescue complex, fear of partners suicide, denial, love, sense of duty, guilt, sense of responsibility, shame, embarrassment and humiliation, sense of security, identity crisis, low self-esteem and sex role conditioning.

2.4.3. Theories on the Causes of Partner Violence

Early theoretical development on the causes of partner abuse (Gelles & Straus, 1978) identified 15 theories, organized into three broad categories ---intraindividual theory, sociocultural theory, and social-psychological theory ---that provide guidelines for the risk factors that need to be considered. The intraindividual theory emphasizes the role of alcohol and other drugs and psychological traits such as self-esteem (Hamberger & Hastings, 1986) and antisocial personality disorder (Holtzworth-Munroe & Stuart, 1994) in partner violence. The contribution to later intrafamily aggression of individual biological and neurophysiological disorders such as childhood attention deficit disorders or head injuries as risk markers of relationship aggression are not often incorporated into sociological, feminist, or even psychological theories and research on intrafamily aggression. The sociocultural theories focus on the influence of social location (social class, education, income, employment status) on partner violence and attempt to integrate socio-cultural and family processes (Kaufman Kantor et al, 1995) In the feminist view, the central factors that foster partner assaults include the historically male-dominated social structure and socialization practices teaching men and women gender specific roles

15

(Smith, 1990). The social-psychological approaches, on the other hand, stress social learning through experience and exposure to violence in the family (Kalmuss, 1984) Table 2.4.3 shows the three broad categories of intimate violence theories and the imputed risk factors.

2.4.3. Theoretical Explanations and Risk Factors for Partner Violence Theoretical Explanations Intraindividual Risk Factors Factors within the individual: e.g., excessive drinking; personality disorders, biological or neurophysiological disorders Sociocultural Importance of social location: social class; education; income, employment status Social-structural and family processes: traditional gender roles in families Social-psychological Social learning: exposure to violence in the family one grows up in

2.5.

FACTORS ASSOCIATED WITH WIFE ABUSE

The analysis by Hotaling and Sugarman (1986) of risk markers for partner violence found that eight characteristics were consistent risk factors of the husbands violence toward the wife: (1) sexual aggression towards the wife/partner; (2) violence towards the children; (3) witnessing parental violence as a child or teen; (4) (working class) occupational status; (5) excessive alcohol usage; (6) (low) income; (7) (low) assertiveness; and (8) (low) educational level. Analyses from three national U.S. surveys also show that particular risk factors are enduring and generally consistent predictors of intimate violence across several years of survey research conducted among several thousand families in the general population. These risk factors include cross-generational violence (childhood abuse, witnessing parental violence), occupational status, excessive alcohol use, and socioeconomic status. Whereas research based on clinical populations, such as male batterers in treatment for relationship violence, has supported the significance of the latter sociological risk factors

16

from survey populations, the body of clinical, mainly psychological research has also emphasized the importance of personality variables as major risk factors. Kantor and Jasinski (1994) emphasized violence across generations (parental violence), socioeconomic risk factors (specifically income, education, occupational status), and socio-structural risk factors as risk markers for partner violence. Additional risk factors that have emerged from current research, such as marital factors (age at marriage, duration of marriage, number of children), alcohol and personality factors, were also considered as notable areas of inquiry.

2.5.1. Intraindividual Factors

The discussion that follows is focused mainly on the associations of intimate partner violence with intraindividual risk factors, such as substance abuse and abusers personality.

Substance abuse

Alcohol is the drug most commonly associated with violence and substance abuse is a consistent finding in many of the profiles of abusive men. Fagan (1990) found strong evidence of a psychopharmaceutical basis only for the combined effects of an alcoholcocaine combination. Aldarondo (in press) found that a lifetime history of any hard drug use (e.g., cocaine, amphetamines, heroin) by the husband was associated with a more persistent pattern of abuse in intimate relationships. Physical abuse, however, is not an inevitable consequence of any intoxicant usage. Conventional wisdom about the alcohol-violence linkage has emphasized alcohols powers as a disinhibitor or drunkenness as an excuse for violence (Kaufman Kantor & Straus, 1989). Attributing alcohols effects on marital aggression to disinhibition, however, understates the complexity of the relationship. Experimental laboratory studies examining alcohol effects on aggression have generally demonstrated both direct effects of alcohol on aggression and indirect effects mediated by beliefs about alcohols ability to change behavior. The consensus is that alcohol disrupts cognitive functioning by diminished ability to reason, reduced ability to perceive and calculate consequences of aggressive

17

behaviors, and enhanced perception of threat (Pihl, Smith, & Farrell, 1983). Aggression may be a consequence of alcohols impairment of perception, judgement, and memory. Distorted perceptions increase the likelihood of miscommunications, developing resentment and ability to take into account the consequences of aggressive actions (Kaufman Kantor & Asdigian, 1996). Reviews of research on alcohol and wife assaults, however, yield wide-ranging estimates suggesting that from 6%-85% of wife assaults involved alcohol. One or both of the parties is more likely to be intoxicated at the time of an assault when the husband is a heavy daily drinker or binge drinker. This was found to be the case for about one half of the violent families interviewed in the 1995 National Family Violence Survey in the U.S (Kaufman Kantor & Straus, 1987). The results of a local urban community-based study by Ramiro, Amarillo and Madrid involving one thousand women aged 15-49 years in Paco, Manila, showed that physical violence was more prevalent if the man is a heavy drinker or was under the influence of drugs (1998, in press). Alcohol-related family assaults have been explained in many ways pointing to the importance of social context (e.g., fights over a spouses drunkenness that lead to partner violence). Examination by Kaufman Kantor (1990, 1993) of the relationship between alcohol consumption patterns and intrafamily violence revealed significant effects of a family history of violence and current family alcohol use on the incidence of child abuse and wife abuse. The author concluded that aggressive beliefs about the effects of alcohol and the appropriateness of inflicting violence while under the influence may both be learned from the family. For example, an adult may recall being beaten by an intoxicated parent or recall violent encounters between parents when one or both were drinking. One result of such exposure is an increased likelihood that these experiences will be reenacted in the second generation because of modeling influences. People may choose to get drunk to forget their troubles. However, aggression does not necessarily follow from drunkenness as each drinking situation is composed of many situational variables (e.g., amount drunk, partners drunkenness, and issues raised by either partner when drinking).

18

Abusers personality

Violence across generations provides some understanding of how personality is shaped by the family-of-origin experience and contributes to future attributes and behaviors conducive to both victimization and violence. For example, the excessive controlling behaviors of maritally violent men may be caused by their feelings of helplessness and dependency stemming from abusive childhood (Shupe et al, 1987). Men who batter are often emotionally dependent, insecure, low in self-esteem and empathy, and exhibit low impulse control and poor communication and social skills (Holtzworth-Munroe, 1992) and are often irritable and belligerent. Aggressive and hostile personality styles are also reliably found in many studies of maritally violent men (Heyman, OLeary, & Jouriles, 1995). Use of the Minnesota Multiphasic Personality Inventory (MMPI) and Millon Clinical Multiaxial Inventory (MCMI) evaluations or other psychological inventories that are used with batterers reveals elevations on measures of borderline symptomatology, passive and aggressive tendencies, narcissistic and antisocial characteristics, anxiety, depression and somatic complaints (Dutton & Starzomski, 1993) Batterers blame partners for their abusive behavior and place enormous expectations on partners to feel good about themselves. They are very jealous and possessive of partners and try to isolate their partner. They tend not to trust other people, and therefore tend not to share inner world with others. They have limited or no social network. Hence, their partners are the closest persons they know. They tend to express all negative feelings as anger. They have an explosive temper and become enraged without provocation. They subject their partner to verbal assault (insults, put-downs, slanderous names) in addition to physical assault. They use control, in the form of violence and threats, to meet their needs. Batterers may threaten suicide should their partner leave them. They will do whatever it takes to drive their partner away. Similarly, they will do whatever it takes to get her back. They will grab the kids or apologize profusely, send her flowers, cry real tears, promise anything, and they

19

know exactly what she wants to hear: "I'll go to church with you", "I'll go to counseling", "I'll stop drinking", "I'll never hit you again." Once the partner returns, performance is repeated: whatever it takes to drive her away, followed by whatever it takes to get her back. Batterers can be very pleasant outside of the home and very unkind at home. They are frequently demanding and assaultive in sexual behavior. They hold very traditional, stereotyped views of male-female roles and relationships. They may be more violent when partner is pregnant or soon after giving birth. Batterers lack sympathy for partner's physical and emotional pain. They tend to minimize and deny the beatings or their severity and seem not to remember (Bowker, 1983) An important caution in summarizing or generalizing from the results of standardized assessment tools is that scale elevations vary across studies. No singular profile defines all abusive men. Furthermore, men without any abnormal findings in clinical assessment tools may engage in violence toward their partners.

2.5.2. Marital Factors

This portion of the discussion is focused mainly on the association between intimate partner violence and marital risk factors, i.e. age at marriage, early stage of marriage, and multiple children.

Age at Marriage

Youth is probably one of the strongest risk factor for aggression, and therefore we might expect that the prevalence of marital assaults is highest among youthful partners in the early stage of marriage. Others (McLaughlin et al., 1992; Straus et al, 1980) found that the highest rates of partner violence were among those 30 years old and younger, and in particular the group 18-24 years old. Martin, Tsui, Maitra, & Marinshaw (1999) reported that in each of the five districts in their study in India, abusive men were more likely than nonabusive men to have been young (<20 years) when they first married. Duration of Marriage

20

Since marriage duration may serve as a measure of a womans length of time at risk for wife abuse, Martin, Tsui, Maitra, & Marinshaw (1999) examined the association between wife abuse and the duration of the couples marriage. Between 78 and 85 percent of the couples in each of the five districts studied in India had been married for at least 5 years. Greater proportions of abusive men than nonabusive men were part of this longer-marriage group. Analysis of the bivariate associations between abuse and a longer marriage revealed statistically significant results for three of the five districts. The districtspecific logistic regression analyses however showed that although in each district, there was positive association between being in a marriage of a longer duration and abuse, only one of these associations reached statistical significance. In contrast, OLeary and associates (1989) found that 27% of men and 36% of women indicated that they had engaged in aggression one and a half (1 1/2) years after their marriage. After another year, the rates of partner violence decreased slightly to 25% and 32%, respectively. Elliot and co-workers (1986) found similar rates of partner violence among young adults. These rates were three to four times higher than those found by Straus and associates (1980) and Kaufman Kantor and associates (1994), neither of whom focused exclusively on the early marriage period. Other factors, however, may be associated with higher rates of violence in early marriage in addition to youth. Leonard and Senchak (1993) in a study among young couples (ages 18 to 29) in an urban setting applying for marriage licenses, found that higher levels of relationship aggression in early marriage were associated with the husbands heavy alcohol use, marital dissatisfaction, the husbands hostility, and beliefs that alcohol excuses aggression. This study illustrates that other risk factors may interact with early stage of marriage to affect the probability of intimate violence.

Multiple children

Sorenson, Upchurch, Shen (1996), based on data from the National Survey on Families and Households in the U.S., did not find consistent association between violence/injury in marital arguments and number and age of children at home. This is contrary to the findings of Vest, Catlin, Chen and Brownson (2002) in a large populationbased sample of U.S. women using surveillance data. They reported consistent association between intimate partner violence and presence of children in the home after

21

adjustment for covariates. Martin, Tsui, Maitra, & Marinshaw (1999) in their study in India found positive associations between the couple having multiple children as stress-related factors and wife abuse. Similarly, having four or more children were one for the factors found associated with physical abuse by Lown & Vaega (2001) from a large population-based cross-sectional survey of US residents of Mexican origin.

2.5.3. Social-psychological Factor

In this section of the discussion, we examine current knowledge on the social psychological factor of history of violence in the family of origin as a risk factor for marital assault.

History Of Violence In The Family Of Origin

A history of violence in the family of origin is probably the most widely accepted risk factor for the occurrence of partner violence. Based on their initial study of 2,000 U.S. families, Straus and associates (1980) presented strong empirical evidence for the social learning of violence within families. A basic theory underlying this is that the family is the training ground for violence and that even the use of ordinary physical punishment such as spanking is associated with child abuse and wife beating. The two National Family Violence Surveys found increased rates of physical punishment, child abuse, and wife abuse by male and female partners subjected to high rates of physical punishment as children. Thus, multiple forms of family violence relate not only across generations but also within generations. Additionally, witnessing interparental violence is significantly associated with adult marital assaults. Studies on battered women in shelters support the national survey data and report high proportions of cross-generational violence. In 42% to 81% of families of male batterers either these men were abused as children or they saw their mothers abused. The primary mechanism believed to be involved in intergenerational transmission of violence is that of modeling the parental behavior observed (Arias, 1984; OLeary, 1988). Analysis by Stith and Farley (1993) of male spousal violence in a treatment sample of male batterers and male alcoholics indicated that observation of parental violence in

22

childhood diminishes mens self-esteem. In turn, lowered self-esteem increases marital stress, alcoholism, and approval of marital violence. Normative approval of marital violence was also directly influenced by the process of social learning. Alexander, Moore and Alexander (1991) suggest that the development of attitudes towards women is also affected by having witnessed parental marital violence. Abuse in childhood or exposure to marital violence in the family of origin may similarly increase womens risks of victimization. Where intergenerational effects exist, the mechanism may be as follows: repeated attacks by parents can lead to damaged selfesteem and suppression of rage. These effects may be different for men and women and may be compounded by the current life situation. Women who were harshly punished in childhood or who witnessed parental violence are more likely to be victimized as adults because they have low self-esteem and have learned that assaults from a loved one are legitimate or because they are more likely to engage in mutual assaults with their spouses. For both men and women, recollections of maternal beatings predicted their perpetration of marital assaults, but recollection of paternal beatings predicted marital victimization. Langhinrichsen-Rohling et al (1987) speculate about the possibility that maternal violence carries with it a violent message about appropriate styles of conflict resolution but that paternal violence leads to a sense of helplessness. An important conclusion of research on the intergenerational transmission of partner violence is that many risk factors for current partner violence are interwoven in family-of-origin experience. For example, because alcoholism and family violence are so intertwined, both men and women exposed to violence at home may also have suffered the effects of parental alcoholism. Intergenerational transmission of parental rejection and depressed mood can also occur. Abuse and the proclivity for violence may be transmitted along with problems such as having a cold and unresponsive father. Simon and coinvestigators (1993) concluded that women subjected to abusive parenting may develop a hostile, rebellious orientation and consequently, be more likely to affiliate with and marry similar men. This mechanism could also be mediated, however, by the partners shared history of similar delinquent behavior (alcohol, other drug use) or current shared lifestyle factor.

23

2.5.4.

Social-structural Factors

In this portion of the discussion, we consider the social-structural factor of male dominance in family processes as risk factor for marital assault. We then examine household decision making patterns associated with intimate partner violence in the Philippines.

Male dominance in family processes

Some evidence suggests that the way the family unit is organized (e.g., male dominated vs. equality between partners) plays an important role in family functioning (Coleman & Straus, 1986; Straus et al, 1980). For example, the results from previous research suggest that wife beating is more common in households where power is concentrated in the hands of the husband or male partner. (Coleman & Straus, 1990; Levinson, 1989; Straus et al, 1980; Yllo & Straus, 1990). In these households, physical violence may be used to assert the dominant position of the male. (Babcock, Waltz, Jacobson, & Gottman, 1993).

Womens status and patterns of household decision-making in the Philippines According to Licuanan (1993), most Filipinos sincerely believe that the status of women is equal to that of the men. In fact, equality of men and women is provided for by law. It is no wonder that Filipino women get elected to public office and their educational attainment is quite similar to mens, unlike the situation in any developing nations. Based on the results of a five-country study, Mason (1996) concludes that women in the Philippines have more autonomy than most Asian women. Despite this apparent equality however, gender relations within households and families may vary. While the husband is considered the legal head of the family in the Philippines, a review of numerous studies by Medina (1991) and David (1994) showed that joint decision-making is the norm' in Filipino homes. Even within the norm of joint decisions, there is often a clear division or specialization of mens and womens roles. For example, both Alcantara (1994) and Medina (1991), concluded that women have the power to decide about financial resource allocation and are seen as treasurers of the household.

24

Additional evidence of the norm of joint decision-making with specialization is shown in the work of David (1994) and Alcantara (1994). David (1994) finds that in examining individual domains, decisions about the household budget allocation are consistently dominated by the wife and decisions about family finances and investments are dominated by husbands. The author suggests that these results reflect one member of the couple deferring to the other, who is supposed to have more skills in the area. David (1994) shows that women with more education have more influence in the financial and family planning realm, while education increases husband involvement in the household budget. As for family planning use, men more often have final say. However, David found that women do not necessarily accept their spouses authority and that generally, decisions on resource allocation and fertility are made jointly. Surprisingly, these results are not mediated by income or relative income and only moderately associated with womens education. The data also show that parity is a key factor in womens ability to contribute to fertility decisions. Based on gender relations in the Philippines, and using data from the Cebu Longitudinal Health and Nutrition Survey (CLHNS), Hindin and Adair (2002) demonstrated that patterns of household decision-making are strong predictors of violence. The greater the number of decision-making domain men dominate, the more likely they are to engage in intimate partner violence (IPV). However, they also found that when women dominate household decisions, they are also more likely to experience IPV. Only 6% of women reported IPV when all household decisions were made jointly compared to 25% when no decisions were made jointly.

2.5.5.

Social Location Factors

In the following discussion, we examine current knowledge on social location risk factors for marital assault, including educational attainment, occupation, wifes educational attainment, wifes occupation, occupational status discrepancy and per capita monthly income.

Educational attainment

The relationship between educational attainment and partner violence is complex and inconsistent. Review by Hotaling and Sugarman (1986) of more than 400 empirical

25

reports on husband-to-wife violence found that higher educational level was associated with less violence in more than half of the studies examined in their analyses. Other research has found a mixed relationship between educational attainment and partner violence. Straus and associates (1980), for example, found that partner violence was most common among individuals with high school diplomas or at least some high school education. More recent research suggests that the absence of a college education is associated with a high frequency of both moderate and severe violence. Martin and coworkers (1999), in their study of nearly 6,700 married men living in five districts in northern India in 1995-96, reported that greater proportions of abusive men than nonabusive men had low levels of education(<5 years). Research by Rollins and Oheneba-Sakyi (1990) on Utah households, however, found no differences between violent and nonviolent families based on education. Education may be most important as it relates to the likelihood of finding employment in a stable and well-paying job.

Occupation

Previous researches (Gelles, 1978; Gelles & Straus, 1978; Hornung, McCullough, & Sugimoto, 1981; Jasinski, 1996; Steinmetz & Straus, 1974) have suggested that households in which the husband is unemployed or employed only part-time have increased rates of violence, particularly severe violence. Unemployment or part-time work with few, if any, benefits is very stressful economically for families and may increase the likelihood of partner violence. Other researches (Hotaling & Sugarman, 1990; Rollins & Oheneba-Sakyi, 1990) however, has found no relationship between employment status and violence among intimates. Evidence also suggests that, in addition to unemployment, occupational class or status is related to partner violence. For example, rates of violence between husbands and wives have been found to be twice as high in families of blue-collar workers than in families of white-collar workers (Kaufman Kantor & Straus, 1987) . McLaughlin et al (1992) also found that rates of moderate aggression were twice as high among workingand middle-class husbands, compared with upper-class husbands.

Wifes education

26

Having some form of post-school (college) education was found by Jewkes and coinvestigators (2002) to be protective against having ever been abused. Straus, Gelles & Steinmetz (1980) have likewise observed this finding that women who had higher education were less likely to be abused. It seems that the mechanism of protection is not mediated solely through economic independence, since many women with less education were economically independent. Rather, it could also be mediated through greater social empowerment (i.e. social networks, self-confidence, or an ability to utilize sources of information and resources available in society). subservience scores of more educated women. Women with more than a high-school education appeared to be at higher risk for injury from domestic violence than women who were high-school graduates (RRa=2.7, 95% confidence interval: 1.4 to 5.0) (Kyriacou et al, 1999). However, the relative risk varied significantly depending on which variables were included in or excluded from the multiple logistic-regression models. Hence, their results need to be interpreted with caution. Results of the 1993 Safe Motherhood Survey, which involved 8,481 married women showed that 1 for every 10 married Filipina, experienced domestic violence in the form of physical abuse. Chi square test of independence (at 1% level of significance) showed that the prevalence of abuse is higher among less educated women. One indicator of that social empowerment could be a consequence of higher education is the lower personal

Wifes occupation

According to the National Commission on the Role of Filipino Women (NCRFW), Filipino women have been joining the labor force partly out of economic necessity and partly in response to economic opportunities. However, there remain wide gaps in female and male labor force participation rates. Womens labor force participation rate (52.8%) still lagged behind mens labor force participation rate (82.3%). Women's employment rate was slightly lower (89.7%) than mens (90.6%). Women's unemployment rate was 10.4% in October 2001. Based on the new major occupation categories of the National Statistics Office (NSO), among women who were employed, 68.5% were professionals, 66.7% were

27

clerks, 58.1% were officials of government and special-interest organizations, corporate executives, managers, managing proprietors and supervisors, 54.3% were service workers and shop and market sales workers, 49.5% were technicians, 46.2% were laborers, 28.3% were in trades, 16.6% were in farming, forestry and fishing occupations 8% were plant and machine operators and assemblers and 12% were in so called special occupations. In this country, cases handled by Lihok-Pilipina, the Womens Crisis Center and Kalakasan showed that among the women who were victims of domestic violence, 25% were unemployed or unskilled workers, 18% were professionals while 15% were skilled workers. A recent study (David, Chin, & Herradura, 1998) in the Western Visayas region of the Philippines shows that intimate partner violence (IPV) is related to womens employment status, with one in three of the working or socially active women surveyed reporting abuse (David et al., 1998).

Primary breadwinner in the family Smith (1998) found that households in which the wife works, or works more than her male partner, might be at greater risk for spousal violence. One explanation of this finding suggests that men may use physical violence to compensate for their inability to be the primary breadwinner in the family (Straus et al, 1980). Hornung, McCullough, and Sugimoto (1981) provide some evidence that relative occupational status may also be an important risk factor for partner aggression. In particular, they find that in couples in which both partners are working, the risk is increased if the husbands occupational attainment is higher than his wife. Yllo and Straus (1990) have suggested that, on the one hand, in a couple in which the wife has the higher status, the husband may feel threatened by this and use violence to restore his authority. On the other hand, the husband may use violence as a mechanism of control and dominance when the wife has a much lower status than he has. The results of these studies provide strong evidence that inequalities of status and power are undesirable in marriages and are significant predictors of partner physical aggression.

28

Income

Evidence from researches (Dibble & Straus, 1980; Hotaling & Sugarman, 1986; Kaufman Kantor, 1990; Straus & Smith, 1990) suggests that low income, as indicated by poverty, is an important risk factor for partner violence. The study of Martin and associates (1999) in India reported that in each of the five districts, greater proportions of abusive men compared with nonabusive men lived in extreme poverty (those who had fewer that two of the possessions used to assess socioeconomic status). There are many theories linking domestic violence to resources. Goode (1970) suggested that men beat women when they lack other means to control them. Gelles (1978) alternatively proposed that the relationship between poverty and intimate partner violence is mediated through stress. Families living in poverty may suffer from stress because of their inability to meet their needs with the resources available to them. This stress may lead to frustration and possibly aggression. However, a large study by Hoffman et al (1994) of men in Thailand failed to confirm the stress hypothesis. More recently, Bourgois (1996) has argued that poverty and unemployment reduce the ability of men to attain certain ideals of successful manhood, particularly those based on ideas that men should be the providers for the family. Men lash out at women whom they can no longer control or economically support. In contrast, a local study (Bascos-Deveza, 2001) found no correlation between domestic violence prevalence rates and the 1994 poverty incidence using Spearmans rank correlation across regions (rs=0.01).

29

3
CONCEPTUAL FRAMEWORK, HYPOTHESES AND OBJECTIVES

3.1.

CONCEPTUAL FRAMEWORK

The association of several factors to wife abuse was examined in this study. The factors that were considered for possible inclusion in the statistical model were the marital factors, social-psychological factors, social-structural factors and social-location factors. The hypothesized interrelationships of these factors are shown in the subsequent sections and are presented schematically in Figure 3.1. The outcome variable of interest is wife abuse while the exposure variables being considered are those referable to the broad categories of intimate violence theory. First, based on the intraindividual explanation, that abuse is more common in families with risk factors such as excessive drinking and personality disorders. Second, based on the sociocultural explanation, that abuse is more common among families with low social class, low educational level, low income, working class occupational status and also among those families with unequal family power structures or traditional gender roles. Third, based on the social-psychological explanation, that abuse is more common among those exposed to violence in the family one grows up in. And fourth, based on the results of the study of Martin and co-workers (1999), wife abuse is more prevalent when the husband was young and inexperienced when he first lived with his wife, when he has been married longer, and when the couple had multiple children. These variables are purported to be stress-related factors. Youth is probably one of the strongest risk markers for aggression. At the same time, it was hypothesized that age at marriage would also be affecting ones level of educational attainment. If a man got married at a young age, he may not have finished his schooling yet and thus he may not get a stable and well paying job, resulting in lesser per

30

capita monthly income for his family. Conversely, education appears to influence age at marriage. Local data show that for people with high school education, the median age at first marriage was 21.5 years, compared to the 19.8 and 18.7 years of those with elementary and no education, respectively. For the college educated, the median age was about 25.2 years. It was likewise hypothesized that age at marriage would also affect social structural and family processes. If a man got married at a young age, he may not be ready for responsibilities yet and leave his wife to take care of the household chore and caring for the children and elderly. As a consequence of being young, he may be emotionally dependent, insecure and exhibit low impulse control such that he may object to his wife improving her education and visiting friends without him or he may insist in getting his way when there are disagreements. He may also not have the necessary maturity yet to share decision making with his wife and discuss matters with her frequently. A history of violence in the family of origin is probably the most widely accepted risk factor for the occurrence of partner violence according to some studies. This may in turn affect the age at which a man will enter marriage. It was hypothesized that in order to get away from his dysfunctional family-of-origin, a man may get married young. On the other hand, getting married at a young age, which lessens the chance of finishing his schooling and getting a stable and well-paying job may enhance the stress levels of families. This may subsequently enhance the probability not just of wife abuse but child abuse as well, thus perpetuating the intergenerational cycle of violence. As with the other risk factors considered thus far, economic, educational and occupational deprivation are strong predictors for intimate violence. These social location factors in their turn may affect social-structural and family processes factors. For example, occupational status discrepancy appears to influence a mans attitude towards better education and social well-being of spouse. In a couple in which the wife works, or works more than her male partner, the husband may feel threatened and may try to restore his authority by objecting to his spouses desire to improve her education or to her visiting her friends without him. In some cases, the combination of low self-esteem, feelings of lost interpersonal control from low level of education and unemployment, and not enough money to support the familys needs might lead a man to make his authority

31

felt by dominating the decision-making process, by objecting to his spouses better education and social well-being, by hardly discussing matters with his wife and by insisting in getting his way when there are disagreements. Having multiple children was postulated to be a risk factor for wife abuse and may have relationships with the other variable under study as well. However, due to unavailability of such data, such factor was not included in this study. Similarly, it is beyond the scope of this study to look into alcohol use, substance abuse, husbands personality and biological or neurophysiological disorders as potential causes of wife abuse. In summary, the outcome variable is wife abuse, specifically, wife beating, while the independent variables are: (1) marital factors (age at marriage and duration of marriage) (2) social location factors (educational attainment of respondent, occupation of respondent, educational attainment of spouse, occupation of spouse, occupational status discrepancy between the couple and per capita monthly income) (3) social-structural and family processes factors and (4) social-psychological factor of history of abuse in the family-of-origin.

32

Figure 3.1 CONCEPTUAL FRAMEWORK Correlates of Wife Abuse

Marital Factors

Sociocultural Factors
Social Location: Socio-Structural And Family Processes: Responsibility: Household Tasks Responsibility: Handling Household Budget Decision Making On Food Matters Decision Making on -Purchase/Sale Of Assets -Family Planning Attitude To Social Well-Being & Better Education Of Spouse Frequency of Communication On -What To Spend Money For -Political-Social Issues -Family Planning Matters Conflict Resolution

Age At Marriage Duration Of Marriage Number Of Children

Husbands Educational Attainment Husbands Occupation Wifes Educational Attainment Wifes Occupation Primary Breadwinner In Family Per Capita Household Income

WIFE ABUSE

Intraindividual Factors
Substance Abuse Alcohol Use Abusers Personality Biological /Neurophysiological Disorders

Socio-Psychological Factors
History Of Violence In The Family Of Origin - Parental Abuse In Childhood - Exposure to Marital Violence

33

3.2. VARIABLES AND OPERATIONAL DEFINITIONS

The outcome variable of interest in this study is wife beating in the past year. Operationally, wife beating is based on the respondents own admission of having beaten or hit his wife (pinagbuhatan ng kamay) in the past year when the questionnaire was administered to him via face to face interview. Reponses are dichotomized as to the presence or absence of wife beating in the past year. Independent variables under study include the following:

3.2.1.

Marital Factors

Age at marriage was defined as the respondents age when he had first lived with his civil or common law wife. Setting the cutoff at 24 years of age, based on the cut off set by the Department of Health and textbooks of Medicine to delineate adolescence (youth) from adulthood, this variable was dichotomized as: less than or equal to 24 years old (exposed group) versus 25 years old or older. Duration of marriage was defined as the length of time the respondent has been married/living together with his wife/partner. Setting the cutoff at 5 years, based on Martin et al s study in India, this variable was dichotomized as: more than 5 yrs (exposed group) versus less than or equal to 5 years.

3.2.2.

Social Location Factors

Husbands educational attainment was defined as the highest level of schooling completed by the respondent. This variable was classified into: No education/Primary school (exposed group): those who did not attend school or those reached elementary level High school/vocational-tech: those who reached high school or those took vocational/technical courses College: those who reached college or did post-graduate studies Husbands occupation was defined as to whether or not the respondent was engaged in a job in which he is paid in cash or in kind. This was categorized as: Unemployed (exposed group) without occupation Blue collar worker - those who do manual work that do not require any academic training, e.g. farmer/fisherman/animal producer/forestry worker, transportation or

34

communication worker, service worker, quack doctor, craftsman, laborer White collar worker - those who do mental work that require special training, e.g., government official, professional or executive or administrative or technical personnel, clerk, sales worker, service worker, sports or entertainment worker, clerk, military/police/firemen, seaman, barangay health worker, midwife, government employee, private employee or owner. Wifes educational attainment was classified similarly as the respondents. Wifes occupation was classified similarly as the respondents. Primary breadwinner in the family was defined as to who is primarily responsible for earning income in their household. This was dichotomized as: predominantly husband/both spouses were main breadwinners, versus predominantly wife/others were main breadwinners (exposed group). Per capita monthly income was defined by total monthly income of the household in cash and in noncash divided by the number of family members. This was dichotomized as: less than or equal to Php1,000.00 versus more than Php1,000.00.

3.2.3.

Social-Psychological Factor

History of parental abuse was defined as the respondents own admission of having been hit or beaten by one or both of his parents in his lifetime. This was dichotomized as: without history of parental abuse versus, with history of parental abuse (exposed group).

3.2.4.

Social Structural Factors

Person primarily responsible for doing certain household tasks was defined as to who is primarily responsible for doing household chores such as cooking meals, cleaning up after meals, going to market/buying the groceries, cleaning the house, washing or ironing the clothes. Based on cluster analysis, this variable was dichotomized as: Wife predominantly (exposed group) versus Both/predominantly husband. Person primarily responsible for handling the household budget was defined as to who is primarily responsible for handling the household budget. Based on cluster analysis, this variable was dichotomized as: predominantly husband (exposed group) versus Wife predominantly/Both.

35

Usual decision-maker on food matters was defined as to who is primarily responsible for deciding on what food to buy or prepare for family meals. This variable was dichotomized as: Wife predominantly versus Husband predominantly/both (exposed group). Usual decision-maker on purchase or sale of assets or consumer goods was defined as to who is primarily responsible for deciding on whether to purchase consumer durable goods for the household and on whether to purchase or sell family assets. This variable was dichotomized as: Husband predominantly (exposed group) versus Both/predominantly wife. Usual decision-maker on family planning matters was defined as to who is primarily responsible for deciding on the use of family planning methods and how many children to have. This variable was dichotomized as: Husband predominantly (exposed group) versus Both/predominantly wife. Attitude towards better education and social well being of spouse was defined as to whether the respondent objects to his partners wish to improve her education and to visit her friends alone. This variable was dichotomized as: No objection to these issues versus objection to these issues (exposed group). Frequency of communication of expenses was defined as to the frequency of discussion with spouse regarding what to spend money on. This variable was dichotomized as: Often versus sometimes/never (exposed group). Frequency of communication on political and social issues was defined as to the frequency of discussion with spouse regarding what is happening in the community (i.e., socio-political issues). This variable was dichotomized as: Often versus sometimes/never (exposed group). Frequency of communication on family planning matters was defined as to the frequency of discussion with spouse regarding how many children to have and whether to use a family planning method. This variable was dichotomized as: Often versus sometimes/never (exposed group). Conflict resolution was defined as to who gets ones way when there are disagreements. This variable was dichotomized as: husband gets his way (exposed group) versus couple compromise. 3.3 HYPOTHESES

1. Men who married at a young age, 24 years old or younger, are more likely to beat/hit their wife than men who married at an older age, 25 years old or older.

36

2. Men who have been married longer (more than 5 years) are more likely to beat/hit their wife than men who had been married/living together for a shorter duration (less than or equal to 5 yrs). 3. Men who had a history of violence in the family of origin, either from his father, his mother or both, are more likely to beat/hit their wife than men who had no such history. 4. Men who predominantly or jointly with their wives do household tasks (cook meals, clean up after meals, go to market/buy the groceries, clean the house, wash or iron the clothes) are less likely to beat/hit their wives than men who do not do such tasks and leave their wives to do them or than men who do such tasks jointly with their wives. 5. Men who predominantly handle the household budget are more likely to beat/hit their wives than men who do not handle the budget and let their wives to do so or than men who handle the budget jointly with their wives. 6. Men who predominantly or jointly with their wives decide on what food to buy or prepare for family meals are more likely to beat/hit their wives than men who do not usually decide on food matters and allow their wives to do so. 7. Men who predominantly decide on purchase or sale of family assets or consumer goods are more likely to beat/hit their wives than men who do not usually decide on such matters and concede that their wives decide on such matters or than men who decide on such matters jointly with their wives. 8. Men who predominantly decide on use of family planning methods and how many children to have are more likely to beat/hit their wives than men who do not usually decide on such matters and accede to their wives deciding on such matters or than men who decide on such matters jointly with their wives. 9. Men who object to their spouse/partners improving their education or visiting their friends without him more likely to beat/hit their wives than men who do not object to such matters. 10. Men who only sometimes or never discuss what to spend money for with their spouse are more likely to beat/hit their wives than men who often times discuss this matter with their spouse/partners. 11. Men who only sometimes or never discuss what is happening in the community (political, social issues) with their spouse are more likely to beat/hit their wives than men who only often times discuss this matter with their spouse/partners. 12. Men who only sometimes or never discuss sometimes or never discuss with his

37

spouse how many children to have and whether to use a family planning method are more likely to beat/hit their wives than men who often times discuss these matters with their spouse/partners. 13. Men who usually get their way when there are disagreements are more likely to beat/hit their wives than men who usually resolve such disagreements by couple compromise. 14. Men with no education or who reached grade school level of education only are more likely to beat/hit their spouse than those who had college education or higher. 15. Men who reached high school level of education only are more likely to beat/hit their spouse than those who had college education or higher. 16. Men who are unemployed are more likely to beat/hit their spouse than men who employed as white collar workers. 17. Men who are employed as blue collar workers are more likely to beat/hit their spouse than men who employed as white collar workers. 18. Men are more likely to beat/hit their wives when the wives had no education or had reached primary level of education only than when the wives had college education or higher. 19. Men are more likely to beat/hit their wives when the wives had high school education only than when the wives had college education or higher. 20. Men are more likely to beat/hit their wives when the wives are unemployed than when the wives work as white collar workers. 21. Men are more likely to beat/hit their wives when the wives work as blue collar workers than when the wives work as white collar workers. 22. Men who are the primary breadwinner in the family are less likely to hit/beat their wives than when men are not the primary breadwinner and their wives are the primary breadwinners. 23. Men who earn a per capita monthly income of P1,000.00 are more likely to beat/hit their wives than men who earn a per capita monthly income of more than P1,000.00.

3.4.

OBJECTIVES OF THE STUDY

General Objective: To determine the correlates of wife abuse among ever married males in selected regions in the Philippines. Specific Objectives:

38

To determine whether wife abuse among ever married males in selected regions is associated with: A. Marital factors 1) Age at marriage 2) Duration of marriage B. Social-psychological factor such as history of abuse from parents C. Social-structural factors such as: 1) Person primarily responsible for doing household tasks 2) Person primarily responsible for handling the household budget 3) Usual decision-maker on food matters 4) Usual decision-maker on purchase and sale of assets or consumer goods 5) Usual decision-maker on family planning matters 6) Attitude towards better education and social well-being of spouse 7) Frequency of discussion on what to spend money for 8) Frequency of discussion on political-social issues 9) Frequency of discussion on family planning matters 10) Conflict resolution D. Social location factors such as: 1) Husbands Education 2) Husbands Occupation 3) Wifes Education 4) Wifes Occupation 5) Primary breadwinner in the family 6) Per Capita Monthly Income

39

4
SIGNIFICANCE OF THE STUDY

Family violence has been increasingly recognized in the Philippines as a compelling social problem with deleterious physical as well as mental health consequences. Various studies provide a wide range of estimates of abuse of Filipino women depending on the sample studied. Results of the 1993 Safe Motherhood survey showed that 1 in every 10 married women 15-49 years experienced domestic violence in the form of physical abuse. Prevalence of abuse is higher among younger women, among less educated women and among certain regions in the country. Although several case studies and research reports have focused on wife abuse in other countries and in the Philippines, few community-based surveys have documented the perceptions and attitudes of men towards women and the role that the man and his wife play in their marital relationship which perhaps could explain the mans abusive behaviors. This study is unique because it is probably the first large-scale investigation to study wife abuse in the Philippines by using representative samples of men in selected regions in the Philippines, where prevalence estimates are a little higher compared to other parts of the country. Intimate partner violence is a complex problem and most prevention and intervention efforts must involve multifaceted programs to be implemented by the individual, the family and the community with the support of government and nongovernment organizations. This would require effort, time and resources. Since there are many more pressing problems afflicting the Filipino people and the country--- the economic crisis and widespread poverty and its effects, the unstable political situation, the state of peace and order, to name a few,---domestic violence has not received much attention from government and non-government organizations. As such it remains a hidden domestic problem with many women and their children continuing to suffer its consequences, which at times could be fatal in its severest form. This study aims to identify the relevant risk factors in partner violence focusing on

40

the abuser. The ability to identify the abuser who poses an ongoing risk of future severe assaults is critical to establishing effective prevention and interventions. Being able to characterize a current abuser has implications for policy and practice, at the individual, family and societal level. The results of this work also hopes to serve as a baseline for more stringent efforts to identify and prevent partner violence in the 5th Country Program of the UN-FPA as well as to encourage more participation of government and non-government organizations in finding better solutions to the problem of partner violence.

41

5
METHODOLOGY

5.1.

STUDY DESIGN

This study utilized secondary data obtained from a cross-sectional study on reproductive health (RH) commissioned by the UNFPA-DOH to the College of Public Health. The main aim of the survey was to generate baseline data for the Fifth Cycle Project of the Country Development Programme. The cross-sectional study was conducted in several regions in the country including Region II (Cagayan, Nueva Vizcaya and Quirino), Region VI (Aklan, Antique and Capiz), the Cordillera Autonomous Region (Kalinga, Apayao, Ifugao), Region IV (Palawan), Region XIII (Surigao del Sur, Surigao del Norte and Agusan del Sur), Region XI (Sultan Kudarat), Region XII (North Cotabato), and the Autonomous Region of Muslim Mindanao (Maguindanao). The survey covered various elements of the Family Planning/Reproductive Health Programme such as Family Planning (FP), Maternal and Child Health (MCH), Fertility and Fertility Preferences, Mortality, Reproductive Morbidities, Gender Issues, Migration and Adolescent Reproductive Health. The study design of this particular investigative research was cross-sectional analytical design.

5.1.1. Study population

The mother study employed various methods of data collection, both quantitative and qualitative, covering different target populations to answer its various objectives. The study population of the mother study consisted of never married females (NMF) aged 2549 years, never married males (NMM) 25 years or older, ever married males (EMM) and females (EMF) in the same age groups and adolescents, 15-24 years old. Meanwhile, the study population of this study consisted only of ever married men, including only those who were currently married or living-in and excluding the divorced and the separated.

42

5.1.2. Sampling design

The sampling design used in the study was stratified two-stage systematic cluster sampling. The barangays, which served as the primary sampling units (PSU), were first stratified according to urban-rural location and according to presence or absence of community-based interventions (CBI). However, the sampling frame for PSUs excluded the barangays that were very inaccessible, had peace and order problems. The number of barangays drawn from each stratum was proportional to the distribution in the province. In each province, twelve PSUs were randomly selected by sampling with replacement. Because the probability of selection of the PSUs was proportional to size, the bigger barangays had higher chances of inclusion into the study. The sample households, the secondary sampling units were selected using systematic random sampling. All of the eligibles in the sample household were included in the original survey. Thus, the household can be thought of as a cluster of elementary units. The study population of the mother study (Reproductive Health Survey 2000) consisted of evermarried males, including those who were married or living-in, the separated /divorced and widowed. However, only ever-married males (EMM) who were currently married or livingin were included as subjects of this study.

5.1.3. Statistical Power of the Study

The statistical power (1- ) is the ability of the study to demonstrate an association if one exists (Last, 1995). Using the EPITABLE feature of the EPI-INFO 6 software and setting the alpha () at 0.05, power was determined for a total sample size of 5507 evermarried males. Table 5.1.3 shows the prevalence of wife abuse among the unexposed (Po), the magnitude of odds ratio that is considered of public health importance such that it should be detected as statistically significant, and the corresponding power of the test for detecting an association between wife abuse and selected study variables. For example, when the smallest odds ratio to be detected is specified at 1.5, a statistical power (1- ) of 92% for the variable person primarily responsible for doing household chores means that there is a 92% chance of rejecting a null hypothesis of no association

43

if the true odds ratio is at least 1.5. For an odds ratio of 1.5, the statistical power of the study is relatively low for the following study variables: handling the household budget (78.7%), occupation (77.7%), and primary breadwinner in the family (62.8%). The statistical power of the study could not be computed for the variable history of parental abuse due to the very little number of those exposed to child abuse.

Table 5.1.3. Power of the test for detecting the association between the selected exposure variables and wife abuse
Variable No. of Exposed Ratio of Nonexpose d /Exposed Po (%) Power (%) (1- ) OR=1.5 OR=2.0

1. Age at marriage 2. Duration of marriage 3. History of parental abuse 4. Do household tasks 5. Handle household budget 6. Decide on food matters 7. Attitude towards better education & social well-being of spouse 8.Freq of discussion on expenses 9. Freq of discussion on political & social issues 10. Freq of discussion on family planning 11. Conflict resolution 12. Educational attainment 13. Occupation 14. Wifes educational attainment 15. Per capita household income 16. Wifes occupation 17.Primary breadwinner in the family

2430 4405 43 941 526 2704 4211 923 923 1195 4649 2545 493 2199 1687 3585 322

1.28 4.09 128.21 4.29 10.57 1.05 0.20 5.02 5.02 3.62 0.20 1.18 9.37 1.52 2.10 0.55 16.03

9.8 7.9 8.3 8.0 8.2 9.3 13.3 8.2 8.2 8.1 20.1 7.6 8.5 7.7 8.8 8.5 8.4

100 100 --91.9 78.7 100 100 92.5 92.5 96.1 100 99.7 77.7 99.4 99.2 100 62.8

100 100 --100 99.7 100 100 100 100 100 100 100 99.7 100 100 100 97.1

5.2.

DATA COLLECTION The mother study had 9 data collection tools, and this study utilized only

one. The subjects were interviewed using pre-tested and structured interview schedules. The data of interest in this stud was limited to the section of the Household Survey on household income (cash plus non-cash income) and number of persons in household. From the Ever-Married Male Interview Schedule, data included were the

from Sections B (Personal Characteristics) and F (Gender Issues/Violence Against Women). (refer to Appendix 1. for the abridged questionnaire).

5.3. DATA PROCESSING

44

5.3.1. Creation of variables

Data on ever-married males, excluding those of the separated/divorced and widowed, was obtained from the encoded data set in EPIINFO Version 6 and exported to Statistical Package for Social Sciences (SPSS) version 10.0.The variables analyzed in this study were extracted from this data set. The following variables were created from the extracted variables: Wife abuse (outcome) - obtained from the question asking if the respondent has ever beaten/hit his spouse or partner (pinagbuhatan ng kamay) in the past year (Q. F7). Absence of wife abuse was coded as 0, presence of wife abuse was coded as 1. Age at marriage - obtained from the respondents answer in Q. B2b asking for his age at marriage. If the current partner is not the respondents first wife, then this variable was extracted by subtracting from his age on his last birthday (Q.B1b) the number of years he has been married or living together with his current partner (Q.B3b). Getting married at 25 years or older was coded as 0, while getting married at 24 years or younger was coded as 1. Duration of marriage - this variable was extracted by subtracting from the respondents age on his last birthday (Q.B1b) his age at marriage (Q.B2b). Having been married for 5 years or less was coded as 0 while having been married for more than 5 years was coded as 1. History of parental abuse - obtained from the question asking if his father (Q. F6b) or mother (Q. F6d) has ever hit or beaten the respondent. Absence of history of parental abuse was coded as 0 while presence of history of parental abuse was coded as 1. Education - obtained from QB5b on the highest level of schooling completed by the respondent. The categories that corresponded to no schooling, kinder, Grades 1 to 7, and vocational after grade school were recoded as 2; high school years 1 to 4, and vocational after high school as 1; and college years, college vocational and postgraduate were recoded as 0. Occupation - obtained from QB7b asking for the respondents current occupation. The answers were recoded as follows: White collar worker (0) - if the respondent was employed as a government official, professional or executive or administrative or technical personnel, clerk, sales worker, service worker, sports or entertainment worker, clerk, military/police/firemen, seaman, barangay health worker, midwife, government employee, private employee or owner.

45

Blue

collar

worker

(1)

if

the

respondent

was

employed transportation

as

a or

farmer/fisherman/animal

producer/forestry

worker,

communication worker, service worker, quack doctor, craftsman, laborer Unemployed (2) if the respondent did not have any job Wifes education - obtained from QB6b on the highest level of schooling completed by the respondents spouse or partner and recoded in the same way as educational attainment of respondent. Wifes occupation - obtained from QB8b asking for the current occupation of the respondents spouse/partner and recoded in the same way as occupation of respondent. Occupational status discrepancy - obtained from Q.F2a asking who is primarily responsible for earning income in their household. If the husband/both spouses are the primary breadwinner in the household, then this was coded as 0 while if the wife/other members of the household are the primary breadwinner in the household, then this was coded as 1. Per capita monthly income - household income in cash and in noncash/in kind, divided by the number of persons in the household. If the income was more than Php P1,000.00, this was coded as 0 while if the income was Php P1,000.00 or less, this was coded as 1.

5.3.2. Cluster Analysis

Cluster

analysis,

an empirical and objective method for performing

classifications, was conducted for data reduction. Cluster analysis is the name for a group of multivariate techniques whose primary purpose is to partition a set of objects (e.g., respondents) into two or more groups based on the similarity of the objects for a set of specified characteristics (cluster variate). The resulting clusters of objects should then exhibit high internal (within-cluster) homogeneity

and high external (between-cluster) heterogeneity. With the clusters defined and the underlying structure of the data represented in the clusters, the researcher has a means of revealing relationships among the observations that was perhaps not possible with the individual observations. Thus, if the classification is successful, the objects within clusters will be close together when plotted geometrically, and the different clusters will be far apart. But with this technique comes the responsibility on the part of the researcher to

46

apply the underlying principles appropriately (Hair Jr. et al, 1998). In any application, the objectives of cluster analysis cannot be separated from the selection of variables used to characterize the objects to be clustered. Whether the objective is exploratory or confirmatory, the researcher has effectively constrained the possible results by the variables selected for use. The derived clusters reflect the inherent structure of the data only as defined by the variables. The inclusion of an irrelevant variable increases the chance that outliers will be created on these variables, which can have a substantive effect on the results. Thus, one should never include variables indiscriminately but instead carefully choose the variables with the research objective as the criterion for selection. In this study for example, since it used secondary data collected by means of a questionnaire, there were a large number of observations that needed to be classified into manageable groups. As an objective data reduction procedure, cluster analysis was done in section F of the questionnaire dealing with Gender Issues/Violence Against Women (VAW). Thus, combining a priori knowledge, experts opinion and the results of the cluster analysis, the social-structural variables were created. To illustrate, 5 questions dealt with who was primarily responsible for each of the following tasks: cook meals, clean up after meals, go to the market or buy the groceries, cleans the chouse and washes and iron the clothes. Specifying the number of clusters as being two, respondents were categorized into: one cluster who answered it was the wife predominantly who was responsible for these household tasks and another cluster who answered that it was both spouses or the husband predominantly who was primarily responsible for the said tasks. Hence, with cluster analysis, data from 5 questions initially were reduced to one variable (person primarily responsible for household tasks) categorized into predominantly wife and both/predominantly husband. On the other hand, there were initially 4 questions which dealt with how often the respondent would discuss with his spouse the following matters: what to spend money for, what is happening in the community, how many children to have and whether to use a family planning method. Specifying the number of clusters as being two, the resulting two clusters of respondents were not homogeneous in their responses to each of the four questions. This pointed to the inclusion of irrelevant variables that increased the chance

47

that outliers were created on these variables, which then had a substantive effect on the results. Thus, combining a priori knowledge, experts opinion and this initial result of the cluster analysis, it was decided to create from the four questions not one, but three variables: frequency of discussion on expenses; frequency of discussion on politicalsocial issues; and, frequency of discussion on family planning matters (which combined two questions dealing with how many children to have and whether to use a family planning method). Hence, with cluster analysis, data from 4 questions initially were reduced to three variables, each categorized into two: one cluster of respondents who answered often and another cluster of respondents who answered sometimes/never. In summary, the following social-structural and family process variables were created as a result of the cluster analysis: Person primarily responsible for doing household tasks obtained from Q.F2b, F2c, F2d, F2f and F2g dealing with household tasks. Person primarily responsible for handling the household budget obtained from Q.F2e dealing with who primarily handles the household budget. Usual decision-maker on food matters - obtained from Q.F3aa dealing with who usually decides on what food to buy or prepare for family meals. Usual decision-maker on purchase or sale of assets or consumer goods - obtained from Q.F3ab and F3ae dealing with who usually decides on whether to purchase consumer durable goods for the household, such as radio or TV and on whether to purchase or sell family assets (e.g., car, house and lot, family heirlooms, jewelry). Usual decision-maker on family planning matters obtained from Q.F3ac and Q. F3ad dealing with who usually decides on the use of family planning methods and on how many children to have. Attitude towards better education and social well being of spouseobtained from Q. F9 and Q.F10 dealing with whether the husband would object or not if the partner wanted to improve her education and if the partner wanted to visit her friends alone. Frequency of communication on expenses - obtained from Q. F11a dealing with how often the respondent discussed with his spouse what to spend money on. Frequency of communication on political and social issues - obtained from Q. F11b dealing with how often the respondent discussed with his spouse what is happening in the community (i.e., socio-political issues). Frequency of communication on family planning matters - obtained from Q. F11c and Q. F11d dealing with how often the respondent discussed with his spouse how many

48

children to have and whether to use a family planning method. Conflict resolution - obtained from Q.F15 dealing with how the respondent usually resolved his disagreements with his spouse.

5.4 .

DATA ANALYSIS Analysis of the data was done in several stages: (1) Descriptive analysis; (2)

Crude analysis; (3) Stratified analysis; and (4) Multivariate logistic regression analysis.

5.4.1. Descriptive analysis

Descriptive statistics was used to present the profile of the subjects included in this study. Qualitative variables like respondents educational attainment and occupation; educational attainment and occupation of partner, and history of abuse in family of origin were described in frequencies and percentage. The median, mean, and standard error of quantitative variables such as age at marriage, duration of marriage and per capita monthly income were computed. The prevalence of wife beating/hitting was also computed. Missing observations were automatically dropped from the analysis by the statistical software used.

5.4.2. Inferential analysis

Four different logistic regression models, one for each major group of independent variables , were derived, viz., Model on Marital Factors Model on Social Location Factors Model on Social Structural and Family Processes Factors Model on Social-Psychological Factor

The logistic regression model can be transformed into the logit of p as follows:

Logit (P) = 0 + i1E + i2Xi2 +

3 E*Xi2 +4Xi3++ijXij

(Equation 1)

49

where Logit(P): E 0 i E*Xi : : : :

logit of the model exposure variable of interest the intercept the logistic regression coefficient interaction between E and Xi

Note that in the absence of statistical interaction, the odds ratio, which measures the magnitude of association between a particular exposure variable and wife abuse, can be derived by taking the antilog of ij; i.e.

ij
ORi j = e In a model with statistical interaction, containing a main effect variable E, a covariate X, and their interaction E*X, the odds ratio can be derived by taking the antilog of 1 +3; i.e.

1+3
OR = e

5.4.2.1.

Crude analysis

Crude analysis was performed to assess the relationship between the independent variables and wife abuse. It was also conducted to screen for potential confounders.Only variables with p-value less than or equal to 0.25 in the crude analysis were considered for inclusion in the multiple logistic regression. This criterion was used to avoid exclusion of variables that,on crude analysis,may have a weak association with wife abuse simply because of the uncontrolled confounding effect of other variables.

5.4.2.2.

Multiple Logistic Regression Analysis

Multiple logistic regression analysis, using STATA version 6 (Startcorp, 1999), was performed to estimate the magnitude of association between particular exposure variables and the outcome of wife abuse while controlling for confounding and accounting for effect measure modification (i.e. interaction). Below were the steps in modeling.

Exposure Variable Selection

50

For models that have multiple exposure variables, only those with p-value less than or equal to 0.25 on crude analysis were included in the first main effects logistic model. For example, for social location factors, six variables were subjected to crude analysis. Of these, only three, namely husbands educational attainment, wifes educational attainment and per capita monthly income, were significant in the crude analysis and were thus subjected to logistic regression. The importance of exposure variables of interest was verified by backward elimination process utilizing Wald s statistic. Variables with p-values greater than 0.05 in the logistic regression were subsequently deleted from the main effects model starting with the one with the highest pvalue. Of the six social location factors, only wifes educational attainment was retained in the main effects model. Screening for effect measure modifiers using stratified analysis

Stratified analysis was used to screen for probable effect measure modifiers and their corresponding interaction terms. Interaction means that the odds ratio (OR) for the exposure variable varies with the value of a covariate. The presence of statistical interaction was suspected when the p-value of the test for homogeneity of the odds ratios was less than or equal to 0.05. The variables that satisfied this criterion as well as their corresponding interaction with the exposure variable were thus included in the logistic model used for identifying effect measure modifiers. Prior knowledge should be the basis for identification of probable effect measure modifiers. Unfortunately, literature that identified effect measure modifiers to the association between wife abuse and the exposure variables was so scanty. Therefore, variables that could logically modify the effect of the exposure variables from a biological perspective were screened for statistical interaction. For instance, interaction terms explored were (1) age at marriage with history of abuse; (2) history of abuse with respondents educational attainment; (3) history of abuse with per capita household income; (4) person primarily responsible for handling the household budget with respondents educational attainment (5) frequency of discussion on family planning with educational attainment of spouse; and, (6) frequency of discussion on expenses with educational attainment of spouse. Literature that identified effect measure modifiers was so limited but it made sense

51

from a biologic perspective to think that if the outcome is wife abuse, the effect (as estimated by the OR) of history of parental abuse would differ greatly with age at marriage. It is feasible to argue that the odds of marital assaults due to history of parental abuse would be increased much more among men who married at a younger age than among married men who married later in life when they were more mature. Similarly, it is also biologically plausible to consider that the effect of history of parental abuse might vary across levels of the respondents educational attainment. It could be expected that the odds of marital assaults would be increased much more among married men with history of parental abuse who have low educational attainment than among married men with history of parental abuse but have high educational attainment. In the process of acquiring higher education which boost self-esteem, men who have reached or finished college may have learned how to cope with having been abused by their parents and thus may have better control of possible violent tendencies emulated form their parents. It likewise makes sense from a biologic perspective to take into account that the relationship between history of parental abuse and wife abuse might not be the same for all levels of per capita monthly income. It could be presumed that the odds of marital assaults would be increased much more among married men with history of parental abuse who have low per capita monthly income than among married men with history of parental abuse but have high income. Men who have low per capita income may have lower self-esteem and might experience a lot of stress that might trigger or aggravate possible violent tendencies. It is also feasible to consider from a biologic perspective that the relationship between person primarily responsible for handling the household budget and wife abuse might not be the same for all levels of respondents educational attainment. Husbands who keep some of their income and do not give it so their wives manage the household budgets could be controlling and domineering and have increased likelihood of giving in to intimate partner violence. It could be presumed that the odds of marital assaults would be higher among those with low educational attainment. However, the effect of men primarily handling the household budget on wife abuse could disappear with high educational attainment as this relates to the likelihood of finding employment in a stable and well-paying job, thereby allowing husbands to attain certain ideals of successful

52

manhood, particularly those based on ideas of men as providers for the family. It is biologically meaningful as well to argue that the effect of the exposure variable wifes educational attainment on the outcome of wife abuse might vary across levels of frequency of discussion on how many children to have & whether to use a family planning method as well as frequency of discussion on expenses. The effect on wife abuse of wifes low educational attainment (related to lack of economic independence and lack of social empowerment) is expected to be neutralized by open communication lines, manifested by frequent communication between spouses.

Identification of effect measure modifiers by multiple logistic regression Presence of effect measure modification or statistical interaction meant that the effect measure (in this case the odds ratio) of one of the variables was not constant over levels of another. Identification of effect measure modifiers was based on Likelihood Ratio Test (LRT) that compared the main effects model, i.e., the model

that included the exposure variables of interest (E) and the probable effect measure modifier, EM (determined from the stratified analysis), with the model that adds the corresponding interaction term to the main effects model (E + EM + E*EM). The interaction terms were considered to contribute to the model if their addition to the main effects model yielded a likelihood ratio test with a p-value less than 0.10. Following the algorithm of Hosmer and Lemeshow, identification of the effect measure modifiers was done one at a time.

Identification of Confounders

Assessment of confounding started with a full model that included the exposure variables, the potential confounders (determined from the crude analysis), and the identified effect measure modifiers as well as their corresponding interaction terms. The confounding effect of the potential confounders was verified through the change in estimate criterion (Equation 2). Change-in-estimate = ORc ORa ORa Where: (Equation 2)

53

ORa ORc

= =

odds ratio derived from the full model odds ratio derived from the model without the potential confounder

Basically, it compared the odds ratios that were adjusted (ORa) and unadjusted (ORc), for the potential confounder. The adjusted OR (ORa) was derived from the full model. The change-in-the estimate criterion has been arbitrarily set by many at 10%. A change equal to or greater than 10% was considered to be sufficient distortion of the measure of association implying that the variable was a confounder and must be retained in the model. Meanwhile, a change less than 10% would imply that the variable was not a confounder and would be deleted from the model. The process stated with the potential confounder that had the highest Wald pvalue based on the full model and was repeated until none of the potential confounders could be deleted from the model. Figure 5.4 analyzing the data. shows the steps that were taken in

54

6
RESULTS

6.1.

PROFILE OF THE RESPONDENTS

For this study, data from a total of five thousand five hundred seven (5,507) ever married men from Regions II (Cagayan, Nueva Vizcaya and Quirino), VI (Aklan, Antique and Capiz), Cordillera Autonomous Region (Kalinga, Apayao, Ifugao), IV (Palawan), XIII (Surigao del Sur, Surigao del Norte and Agusan del Sur), XI (Sultan Kudarat), XII (North Cotabato), and the Autonomous Region of Muslim Mindanao (Maguindanao) of the Philippines were processed. The subjects, with ages ranging from 16 to 87 years, had a mean, median and modal ages of 41, 39 and 35 years respectively. The majority was married, either in the Church or civilly. More than half live in the rural areas. (Table 6.1.1) Table 6.1.1. Distribution of respondents as to age, marital status and residence Frequency Percent <20 24 0.4 Age 20-29 903 16.4 (Years) 30-39 1765 32.0 40-49 1405 25.5 =/>50 1404 25.6 Missing data 6 0.1 5507 100 Total Married in Church 4184 76.0 Marital status Civilly Married 948 17.2 Living-In 375 6.8 5507 100 Total (n of valid cases) 3685 66.9 Place of Residence Rural Urban 1822 33.1 5507 100 Total

55

Table 6.1.2 shows the social and demographic characteristics of the respondents. The mean as well as the median age of the respondents when they first got married was 24 years. Majority (80%) has been married for more than 5 years with mean duration of marriage of 17 years. Almost half of the respondents had low educational attainment (have no schooling or have grade school education). The majority of the respondents were engaged in some kind of blue-collar job. Solely or jointly with their wives, they serve as the primary breadwinners in the household. Most of them had per capita monthly income of Php1, 000.00 or less. All but 43 (0.8%) of the respondents did not have their father, mother or both parents ever hitting or beating them. Table 6.1.2. Distribution of respondents according to socio-demographic profile Frequency Percent 25 years old or older 2400 43.6 Age At Marriage 24 years old or younger 3077 55.9 Missing data 30 0.6 5507 100 Total 5 years or less 4405 80.0 Duration of more than 5 years 1076 19.5 Marriage Missing data 26 0.5 5507 100 Total College 1101 20.0 Educational 34.0 High school/Vocational 1869 Attainment No education/Primary 2517 45.6 Missing data 20 0.4 5507 100 Total White Collar Worker 454 8.2 Occupation 74.8 Blue Collar Worker 4119 Unemployed 484 8.8 Missing data 484 8.2 5507 100 Total Husband/Both 5106 92.7 Primary WifeOthers 319 5.8 Breadwinner in the Missing data 82 1.5 Family 5507 100 Total > Php 1000 1670 30.3 Per Capita Php 1000 3504 63.6 Household Income Missing data 333 6.1 5507 100 Total 5453 99.0 History of parental Without history With history 43 0.8 abuse Missing data 19 0.2 5507 100 Total

56

Table 6.1.3 shows the social and demographic characteristics of the respondents female partners of the last year. Most of their female partners had low educational attainment (have no schooling or have grade school education), a third of them had some high school education while only a quarter had college education. Their female partners were mostly unemployed. Those who were working were working mainly as bluecollar workers. Table 6.1.3. Distribution of respondents according to socio-demographic profile of their spouses Frequency Percent College 1361 24.7 Educational High school/Vocational 1943 35.3 Attainment of No education/Primary 2177 39.4 Spouse/Partner 0.5 Missing data 26 5507 100 Total White Collar Worker 632 11.5 Occupation of Blue Collar Worker 1325 24.0 spouse/partner Unemployed 3543 64.3 0.1 Missing data 7 5507 100 Total

6.2.

PREVALENCE OF WIFE ABUSE

Overall, 468 or 8.4% of the study population admitted having ever beaten or hit his spouse or partner in the past year. On the other hand, 5100 or 91.6% of the respondents claimed they have not given in to wife beating in the past year. Table 6.2 shows the oneyear prevalence of wife beating among ever married males in selected regions in the Philippines.

Table 6.2. Prevalence of wife abuse among ever married males in selected regions Has ever Beaten/Hit Spouse/Partner in the Past Year Frequency Percent No 5042 91.6 Yes 465 8.4 5507 100 n of valid cases

57

6.2.1. Prevalence of Wife Beating According to Marital Factors

Based on review of related literature, there were three marital factors that could be associated with wife abuse, namely age at marriage, early stage of marriage and having multiple children. However, due to lack of adequate data on number of children, only age at marriage stage of marriage and duration of marriage were subjected to crude analysis. Table 6.2.1. shows that more than half (56.1%) of the respondents got married at 24 years of age or younger. The prevalence of wife beating according to age at marriage was slightly greater (9.8%) among those who got married at 24 years old or younger than among those (6.5%) who were married at 25 years old or older. Table 6.2.1. also shows that most of the respondents (80%) have been married for more than 5 years. The prevalence of wife beating according to duration of marriage was slightly greater (8.6%) among those who have been married longer (> 5 years) than among those (6.5%) who have been married for a shorter duration (</= 5 years).

Table 6.2.1. Prevalence of wife abuse among ever married males in selected regions according to marital factors Total b Wife Abuse Marital Factors
Absent Frequency Row % D Present Frequency Row % 158 302 460 85 377 462 6.5 Freq Column % 43.9 56.1 100 19.6 80.4 100

Age at Marriage >24 yrsa 24 yrs


Column Total

2242 2775 5017 991 4028 5019

93.5 90.2 100 92.1 91.4 100

9.8
100 7.9

2400 3077 5477 1076 4405 5481

Duration of Marriage 5 yrsa >5 yrs


Column Total
a b

8.6
100

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

6.2.2. Prevalence of Wife Beating According to Psychological-Social Factor

Table 6.2.2. shows that practically all (99.2%) of the respondents had no history of experiencing physical abuse from their parents. When there was history in the respondent of ever having been beaten/hit by his father, mother or both, the prevalence of wife

58

beating was thrice (25.6%) that of those who did not have such history (8.3%). Table 6.2.2. Prevalence of wife abuse among ever married males in selected regions according to psychological-social factor b Total Wife Abuse History of one or both parents ever (Absent) D (Present) Row Frequency Row% Freq Column hitting or beating the Frequency % % respondents
Without history With history Column Total
a a

5001 32 5033

91.7 74.4 100

452 11 463

8.3

25.6
100

5453 43 5496

99.2 0.8 100

Reference group b Discrepancy in total number of subjects equal to 5,507 is due to missing observations

6.2.3. Prevalence of Wife Beating According to Social Structural Factors

Table 6.2.3.1 shows that majority of the respondents would have their wives do the household tasks (78.4%). The prevalence of wife abuse was greater (9.6%) among men who, solely or jointly with his spouse, were primarily responsible for cooking meals, cleaning up after meals, going to market/buying the groceries, cleaning the house, washing or ironing the clothes. Similarly, Table 6.2.3.1 shows that practically all (90.4%) of the respondents allow their wives, or together with the wives, to handle the household budget. The prevalence of wife abuse was greater among men who handled the household budget (10.8%) than among men who would let their spouse, solely or jointly with them, handle the household budget (8.2%).

Table 6.2.3.1. Prevalence of wife abuse among ever married males in selected regions according to person primarily responsible for certain tasks b Total Wife Abuse Person Primarily Responsible (Absent) D (Present)
Frequency Row % 90.4 91.9 100 Frequency Row % Freq Col % 21.6 78.4 100

For Doing Household Tasks


Both/Predominantly husband Predominantly wife Column Total

955 3565 4520

103 314 417

9.6
8.1 100

1058 3879 4937

59

For Handling Household Budget


Predominantly wife/Both Predominantly husband Column Total
a b

4520 462 4982

91.8 89.2 100

403 57 460

8.2

10.8
100

4923 519 5442

90.4 9.6 100

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

Table 6.2.3.2 shows that the usual decision-maker on food matters were both spouses or predominantly the husband in more than half of subjects (51.3%). The prevalence of wife beating was greater among men who, solely or jointly with their spouse, would usually decide on what food to buy or prepare for family meals (9.3%) than among men who would let their spouse make such decisions (7.5%). Table 6.2.3.2 also shows that the usual decision maker on purchase or sale of assets and consumer goods were both spouses or predominantly the wife in most cases (66.4%). There was no difference in the prevalence of wife beating (8.4%) when the decision making on purchasing or selling of assets or consumer goods was made predominantly by the respondent or by the wife/jointly by the spouses. In addition, Table 6.2.3.2 shows that the usual decision maker on family planning matters were both spouses or predominantly the wife in majority of cases (79.7%). The prevalence of wife beating was higher (9.3%) among those respondents who would usually decide on the use of family planning methods and how many children to have than when both spouses or the wife predominantly would decide on this matter (8.3%).

Table 6.2.3.2. Prevalence of wife abuse among ever married males in selected regions according to usual decision-maker on certain matters Total b Wife Abuse Usual Decision Maker (Absent) D (Present)
Frequency Row % 92.5 90.7 100 Frequency Row % 7.5 Freq Col % 48.7 51.3 100

On Food Matters

Predominantly wife Both/Predominantly husband Column Total

2472 2557 5029

203 262 465

9.3
100

2675 2819 5494

On Purchase or sale of assets or consumer goods


Both/Predominantly wife Predominantly husband Column Total
a

3360 1701 5061

91.6 91.6 100

307 157 464

8.4 8.4 100

3667 1858 5503

66.4 33.6 100

60

On Family planning and number of children


Both/Predominantly wife Predominantly husband Column Total
a b

3933 992 4925

91.7 90.7 100

354 102 456

8.3

9.3
100

4287 1094 5381

79.7 20.3 100

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

Table 6.2.3.3 shows that most of the respondents do not object to their wives wanting to improve her education and to visit her friends without him (88.1%). Wife beating was almost twice as prevalent (13.3%) when the respondent would object if his spouse/partner wanted to improve her education and if she would visit her friends without him compared to those when the respondents had no objection to these matters (7.6%).

Table 6.2.3.3. Prevalence of wife abuse among ever married males in selected regions according to Attitude towards Better Education & Social Well Being of Spouse Total b Attitude towards Better Wife Abuse Education & Social Well (Absent) D (Present) Frequency Row Frequency Row Freq % Being of Spouse
Without objection With Objection Column Total
a a

3851 488 4339

% 92.4 86.7 100

319 75 394

% 7.6

13.3
100

4170 563 4733

88.1 11.9 100

Reference group b Discrepancy in total number of subjects equal to 5,507 is due to missing observations

Table 6.2.3.4 shows that less than a quarter of the men often discuss certain matters with their wives. Apparently, the more frequently men discussed certain matters with his spouse, the more prone he was to physical abuse. Thus, wife beating was more prevalent among those respondents who would often discuss with their spouse/partner the following matters: what to spend money for (9.4% vs 8.2%), political and social issues (9.4% vs 8.2%), and how many children to have and whether to use a family planning method (9.5% vs 8.1%). Table 6.2.3.4. Prevalence of wife abuse among ever married males in selected regions according to frequency of discussion between spouse Total b Wife Abuse Frequency of Discussion Between Spouse (Absent) D (Present)
Freq & Col %) Row % Freq & Col % Row % Freq Col %

61

On Expenses
Often Sometimes Column Total
a

826 4204 5030 826 4204 5030

90.6 91.8 100 90.6 91.8 100

86 379 465 86 379 465

9.4
8.2 100

912 4583 5495 912 4583 5495

16.6 83.4 100 16.6 83.4 100

On Political & Social Issues


Often Sometimes Column Total
a

9.4
8.2 100

On Family Planning & Number of Children


Often Sometimes Column Total
a b a

1069 3932 5001

90.5 91.9 100

113 350 463

9.5
8.1 100

1182 4282 5464

21.6 78.4 100

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

Table 6.2.3.5. shows that practically all (95.8%) of the respondents reach a compromise with their spouses during disagreements. However, wife beating was almost thrice as prevalent (20.1%) among couples whose norm for conflict resolution is in favor of the husband compared to couples who compromise (8.0%). Table 6.2.3.5. Prevalence of wife abuse among ever married males in selected regions according to Usual Manner of Resolving Disagreements Total b Usual Manner of Resolving Wife Abuse Disagreements (Absent) D (Present)
Frequency Couple compromise Respondent gets his way Column Total
a a

4244 162 4406

Row % 92.0 79.9 100

Frequency 368 41 409

Row % 8.0

Freq 4612 203 4815

20.1
100

Col % 95.8 4.2 100

Reference group b Discrepancy in total number of subjects equal to 5,507 is due to missing observations

6.2.4. Prevalence of Wife Beating According to Social Location Factors

Table 6.2.4.1 shows that most of the respondents had low educational attainment (45.9%), were engaged in blue collar work (81.4%), alone or jointly with their spouse served as the main provider (94.1%) and had a per capita household income of less than or equal to Php1,000.00 (17.7%). Wife beating was slightly more prevalent among married men (9.3%) who had low educational attainment (have no schooling or have grade school education), among those (8.7%) who were employed as a blue-collar worker, among those (9.0%) whose wives were the main breadwinner and among those (8.8%) who had per capita household income of equal to or less than Php1, 000/month.

62

Table 6.2.4.1. Prevalence of wife abuse among ever married males in selected regions in terms of respondents characteristics according to social location factors Total b Wife Abuse (Absent)
Frequency Row % 93.3 91.9 90.7 100 92.8 91.3 93.3 100

D (Present)
Frequency Row % 6.7 8.1 9.3 100 7.2 8.7 6.7 100 Freq Col % 20.0 34.1 45.9 100 9.0 81.4 9.6 100

Educational Attainment
College High school/Vocational No education/Primary Column Total
a

1027 1717 2281 5025 421 3762 451 4634

74 152 236 462 33 357 33 423

1101 1869 2517 5487 454 4119 484 5057

Occupation
White Collar Worker Blue Collar Worker Unemployed Column Total

Primary Breadwinner in the Family


Husband/Both:Breadwinner Wife/Others: Breadwinner Column Total
a

4674 290 4964 1545 3195 4740

91.6 91.0 100 92.6 91.2 100

432 29 464 125 309 434

8.4 9.0 100 7.4 8.8 100

5106 319 5425 1670 3504 5174

94.1 5.9 100 32.3 67.7 100

Per Capita Income/month


> Php 1000* Php 1000 Column Total
a b

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

Table 6.2.4.2 shows that the spouses of the respondents mostly had low educational attainment and were unemployed. The prevalence of wife beating was slightly higher among those whose wives/partners had low educational attainment (have no schooling or have grade school education) (9.5%) and among those whose wives were employed as a blue-collar worker (8.9%) or were unemployed (8.3%). Table 6.2.4.2. Prevalence of wife abuse among ever married males in selected regions in terms of female partners characteristics according to social location factors Total b Characteristics of Female Wife Abuse Partners (Absent) D (Present)
Frequency Row % 94.3 90.8 90.6 100 Frequency Row % 5.7 9.2 9.4 100 Freq Col % 24.8 35.5 39.7 100

Educational Attainment
College High school/Vocational No education/Primary Column Total
a

1283 1763 1972 5018

78 180 205 466

1361 1943 2177 5481

63

Occupation
White collar worker Blue Collar Worker Unemployed Column Total
a b

582 1206 3247 5035

92.2 91.5 91.7 100

50 119 296 465

7.8 8.9 8.3 100

632 1325 3543 5500

11.5 24.1 64.5 100

Reference group Discrepancy in total number of subjects equal to 5,507 is due to missing observations

6.3.

FACTORS ASSOCIATED WITH WIFE ABUSE

Inferential analyses to assess the relationship between the independent variables and wife abuse, were done in three parts: crude analysis, stratified analysis and multiple logistic regression. 1999). All analyses were conducted using STATA version 6 (Startcorp,

6.3.1. Crude analysis

Of the eighteen variables tested, only thirteen variables had p-values less than or equal to 0.25 and were considered for inclusion in the initial multiple logistic regression model. Table 6.3.1. shows that for the martial factors, age at marriage had p-value< 0.25 while duration of marriage had a p-value > 0.25. For social psychological factors, results showed that the variable history of parental abuse had a p-value < 0.25.

Table 6.3.1. Crude odds ratios (cOR), 95% confidence intervals (95% CI) and pvalues of the association between wife abuse and marital & social psychological factors. OR p-value Factors (-) Wife (+) Wife Abuse (95% CI) Abuse % % 1.57 0.000 Age at Marriage 93.5 6.5 (1.28-1.92) > 24 yrs* 90.2 9.8 24 yrs Duration of Marriage 5 yrs* > 5 yrs History of one or both parents ever hitting or beating the respondents Without history* 92.1 91.4 7.9 8.6 1.09 (0.85-1.39) 3.82 (1.91-7.63) 91.7 8.3 0.486

.000

64

With history
* Reference group

74.4

25.6

For social structural factors, results of the crude analysis showed that the variables pertaining to who is primarily responsible for certain household tasks, handling of the household budget and decision making on food matters; to the attitude towards better education & social well-being of spouse; to the frequency of discussion on expenses, political-social issues and family planning matters; and to usual manner of resolving disagreements had p-values less than 0.25. The variables pertaining to who is the usual decision maker on purchase or sale of assets and consumer goods and on family planning & number of children, were not significant and hence, were not subjected to further analysis. Details are summarized in Table 6.3.2. Table 6.3.2. Crude odds ratios (cOR), 95% confidence intervals (95% CI) and pvalues of the association between wife abuse and social structural factors.
Social Structural Factors Primarily Responsible for Household Tasks Both/Predominantly husband* Predominantly wife Primarily Responsible for Handling the Household Budget Predominantly wife/Both* Predominantly husband Usual Decision Maker on Food Matters Predominantly wife * Both/Predominantly husband Usual Decision Maker on Purchase or sale of assets or consumer goods Both/Predominantly wife* Predominantly husband Usual Decision Maker on Family Planning & Number of Children Both/Predominantly wife* Predominantly husband Attitude towards Better Education & Social Well Being of Spouse Without objection* With Objection Frequency of Discussion on Expenses Often* Sometimes Frequency of Discussion on Political & Social Issues Often* Sometimes (-) Wife Abuse (%) (+) Wife Abuse (%) OR (95% CI) 0.83 (0.66-1.05) p-value 0.089

90.5 91.9

9.5 8.1 1.37 (1.02-1.83) 0.029

91.8 89.2

8.2 10.8 1.25 (1.03-1.52) 0.023

92.5 90.7

7.5 9.3 1.01 (0.82-1.24) 0.921

91.6 91.6

8.4 8.4 1.14 (0.90-1.45) 0.259

91.7 90.7

8.3 9.3 1.86 (1.41-2.45) 0.000

92.4 86.7

7.6 13.3 0.86 (0.67-1.11) 0.250

90.6 91.8

9.4 8.2 0.86 (0.67-1.11) 0.250

90.6 91.8

9.4 8.2

65

Frequency of Discussion on Family Planning & Number of Children Often* Sometimes Usual Manner of Conflict Resolution Couple compromise* Respondent gets his way
* Reference group

0.84 (0.67-1.05) 90.5 91.9 9.5 8.1 2.91 (2.00-4.22) 92.0 79.9 8.0 20.1

0.130

0.000

For social location factors, results of the crude analysis showed that the respondents highest educational attainment, the wifes highest educational attainment and the per capita household income had p-values less than 0.25. The variables respondents occupation, wifes occupation and occupational status discrepancy were not significant and hence, were not subjected summarized in Table 6.3.3. Table 6.3.3. Crude odds ratios (cOR), 95% confidence intervals (95% CI) and pvalues of the association between wife abuse and social location factors.
Social Location Factors Respondents Highest Educational Attainment No education/Primary High school/Vocational College* Respondents Occupation Unemployed Blue collar White collar* Wifes Highest Educational Attainment No education/Primary High school/Vocational College* Wifes Occupation Unemployed Blue collar White collar* Primary breadwinner in the family Wife: main breadwinner Husband/Both: breadwinner* Per capita monthly income Php1,000.00 > Php1,000.00*
* Reference group

to stratified analysis. Details are

(-) Wife Abuse (%)

(+) Wife Abuse (%) 9.3 8.1 6.7

OR (95% CI)

p-value

90.7 91.9 93.3 93.3 91.4 93.3

1.22 (0.91-1.64) 1.42 (1.08-1.88)

0.162 0.012

6.7 8.6 6.7

1.22 (0.84-1.76) 0.92 (0.56-1.52)

0.292 0.756

90.5 90.8 94.4 91.7 91.1 92.2 91.0 91.6 91.2 92.6

9.5 9.2 5.6 8.3 8.9 7.8 9.0 8.4

1.68 (1.27-2.22) 1.72 (1.32-2.26)

0.000 0.000

1.14 (0.81-1.62) 1.07 (0.78-1.46) 1.08 (0.72-1.59)

0.432 0.671 0.695

1.12 (0.97-1.51) 8.8 7.4

0.106

66

6.3.2

Multiple Logistic Regression Analysis 6.3.2.1. Association between wife abuse and marital factors

Multiple logistic regression analysis was performed to estimate the magnitude of association between marital factors and the outcome of wife abuse while controlling for confounding and accounting for effect measure modification (i.e., interaction). Below were the results of the modeling process.

Variable Selection

Of the two marital factors of interest, only age at marriage had p-value less than or equal to 0.25 on crude analysis and was included in the model as the exposure variable of interest (E).

Screening for effect measure modifiers using stratified analysis

In order to screen for effect measure modifiers, stratified analysis was performed as detailed in the Methodology section of the study. History of parental abuse did manifest interaction with age at marriage. Table 6.3.2.1.1. shows the stratified analysis for the interaction between history of parental abuse and exposure variable age at marriage. Among those with a history of being beaten/hit by one or both parents, the odds of those who married young (</=24 years old) to beat/hit their wives was 11.33 times those who married at 25 years old or older (95%CI=1.3 - 99.2). In contrast, among those without such history, the odds was only 1.5 times higher (95%CI= 1.3 - 1.9). The significant p-value of the test for homogeneity of the odds ratio that was less than 0.05 justified inclusion of age at marriage in Model 1 as potential effect measure modifier.

Table 6.3.2.1.1. Potential effect measure modifier for the association between age at marriage and wife abuse
Disease (%) Without wife With wife abuse abuse Stratum Specific Odds Ratio 95% CI pvalu e*

History of Parental Abuse

67

Without history of abuse Age at Marriage 25 & older Age at Marriage 24 & younger N of valid cases With history of abuse Age at Marriage 25 & older Age at Marriage 24 & younger N of valid cases * Test for homogeneity

1.5 2221(93.5%) 2755(90.4%) 4976 17 (94.45) 15 (60.0%) 32 156 (6.5%) 291 (9.6%) 447 11.3 1 (5.6%) 10 (40.0%) 11

1.3 1.9

0.04

1.3-99.2

Identification of effect measure modifiers by multiple logistic regression Identification of effect measure modifiers was based on the likelihood ratio test that compared a model that includes age at marriage (Model 1) with the model that adds the probable effect measure modifiers (determined from the stratified analysis) and the corresponding interaction terms one at a time (Model 2).

Model 1 which included age at marriage (E) and history of parental abuse (probable effect measure modifier) was compared with Model 2 which included age at marriage (E), history of parental abuse (probable effect measure modifier) and their corresponding interaction terms. Table 6.3.2.1.2. shows that the likelihood ratio test, with p-value of 0.03, identified history of parental abuse as an effect measure modifier of age at marriage and it was retained in the model.

Table 6.3.2.1.2. Log likelihood ratio test results for the assessment of statistical interaction between age at marriage and history of parental abuse
Variable Log Likelihood LRT p-value

Main effects model Main effect + Effect Measure modifier + Interaction term: (Age at marriage * History of abuse)a a retained in the model

-1558.75 -1556.33

0.03

Identification of Confounders

The full model included age at marriage (exposure variable), history of parental abuse (the identified EM) and their corresponding interaction term and the 11 probable confounders (determined from the crude analysis).

68

Table 6.3.2.1.3. Odds ratios and p-values of the variables in the full model on association between wife abuse and marital factors Variables in the full model Odds ratio p-value
Age at marriage (<24 yrs old) b (+) History of parental abuse Interaction (Age at marriage*History of parental abuse) Person primarily responsible for household tasks (wife)* Person primarily responsible for handling budget (husband)* Decision-maker on food (husband)* With objection to better education & social well-being* Frequency of discussion on expenses (sometimes)* Frequency of discussion on family planning (sometimes)* Conflict resolution (husband gets his way)* Respondents education (HS/vocational)* Respondents education (No education/Primary)* Wifes education (HS/vocational)* Wifes education (No education/Primary)* Per capita income (=/<Php1,000/mo)*
a: main effect b: effect measure modifier *: probable confounders
a

1.46 1.08 10.02 0.88 1.31 1.31 1.64 0.96 0.79 2.51 1.00 1.03 1.56 1.59 1.12

0.005 0.939 0.051 0.450 0.213 0.040 0.005 0.832 0.136 0.000 0.988 0.882 0.028 0.035 0.463

The confounding effect of the potential confounders was verified through the change in estimate criterion, arbitrarily set at 10%, comparing the adjusted and unadjusted odds ratios for the potential confounder. The process started with full model from which the potential confounder that had the highest p-value (respondents education) was deleted. Since the change-in-the estimate was < 10%, it implied that respondents education was not a confounder so it was not retained in the model. The process of comparing the ORs derived from the working model and the full model was repeated until potential confounders could not be deleted from the model since the change of estimate was 10%. These confounders identified by the change-in-the estimate criterion were attitude towards better education and social well being of spouse, per capita monthly income, and, person primarily responsible for household tasks. Specifically, income and attitude towards better education and social well being of spouse confound the effect of age at marriage among ever-married men whether there is a history of parental abuse or not. On the other hand, person primarily responsible for household tasks confound the effect of age at marriage only among ever-married men with history of parental abuse. Details are summarized in Table 6.3.2.1.4.

69

Table 6.3.2.1.4. Test for Confounding (Marital Factors) Factors Full Model (Adjusted) Model without attitude towards better education & social well being of spouse (Unadjusted) % change Model without income (Unadjusted) % change Model without person responsible for doing household tasks (Unadjusted) % change Age at Marriage No history of abuse With history of abuse 1.08 (0.13-8.76) 10.84 (4.06- 28.95) 0.93 (0.12-7.11) 9.12 (3.83-21.74)

-16.96% 0.95 (0.12-7.30) -15.18% 1.05 (0.14-8.12) -6.25%

-17.31% 8.43 (3.44-20.64) -23.57% 8.45 (3.45-20.67) -23.39%

Results show that without controlling for attitude towards better education and social well being of spouse, the association between age at marriage and wife abuse was underestimated by 17% in both strata of history of parental abuse.

Without controlling for income, the underestimation of the association between age at marriage and wife abuse was by 15% among those who, at the same time, did not have a history of parental abuse. The underestimation of the effect measure was even greater (by 24%) among those who, at the same time, had a history of parental abuse. From the findings of this study as well, failure to control for the effect of the factor person primarily responsible for household tasks leads to an underestimation of the association between age at marriage and wife abuse by 23% among those who had a history of parental abuse. Table 6.3.2.1.6 shows that there is absence of joint confounding, i.e., singly, or jointly, by the potential confounders that were deleted from the model. Thus, after controlling for the important extraneous variables, there was negative association between wife abuse and age at marriage among those with history of parental abuse (OR=10.62, 95% CI=4.15-27.17) but not among those without history of parental abuse (OR=1.07, 95% CI= 0.14-8.32).

70

Table 6.3.2.1.6. Comparison of the stratum specific odds ratio between the full model and the final model (marital factors)
Factors Full model (main effects, significant interaction terms & all 11 probable a confounders ) Final model (main effects model, significant interaction terms & 3 b confounders ) % change Age at Marriage No history of abuse With history of abuse 1.08 (0.13-8.76) 10.84 (4.06- 28.95) 1.07 (0.14-8.32) -2.68% 10.62 (4.15-27.17) -2.27%

a: wifes educational status, respondents educational status, per capita monthly income, and person primarily responsible for doing certain tasks and for handling the household budget, usual decision maker on food, attitude towards better education and social well being of spouse, frequency of discussion on expenses, on social-political issues, on family planning and conflict resolution b: per capita monthly income, person primarily responsible for doing certain tasks, and attitude towards better education and social well being of spouse

The final model is now composed of age at marriage as the main effect; history of parental abuse as the effect measure modifier, and; income, attitude towards better education & social well-being of spouse, and, person primarily responsible for household tasks as the confounders. values of the variables in the final model. Table 6.3.2.1.7 Coefficients and p-values of the variables in the final model (marital factors) Variables in the final model Coefficient p-value (B) Beta Constant -2.74 0.000 a Age at marriage (<24 yrs old) 0.40 0.001 (+) History of parental abuseb 0.07 0.937 Interaction (Age at marriage*History of parental abuse) 2.29 0.046 Person primarily responsible for household tasks (wife)c -0.16 0.253 c With objection to better education & social well-being 0.60 0.000 Per capita income (=/<Php1,000/mo)c 0.13 0.322
a: main effect b: effect measure modifier c: confounder

Table 6.3.2.1.7 shows the coefficients and p-

Thus, the final model is: Logit(P)=

0 + 1 E + 2 X + 3 E*X + 4X3+ 5X4 + 6X5

Where: 0 1 2 = intercept = coefficient for age at marriage (exposure variable, E) = coefficient for history of parental abuse (effect measure modifier, X)

71

3
4 5 6

= coefficient of interaction terms age at marriage (E) and history of parental abuse (X) = coefficient of person primarily responsible for doing household tasks = coefficient of attitude towards better education and social well-being of spouse = coefficient of per capita monthly income

Including the value of the coefficients, the final model is as follows:

Logit(P)= -2.74 + 0.40_age at marriage + 0.07_history of parental abuse + 2.29_age at marriage*history of parental abuse + -0.16_person doing household tasks + 0.60_attitude towards better education & social well-being of spouse + 0.13_per capita income

6.3.2.2. Association between wife abuse and social psychological factors

Multiple logistic regression analysis was performed to estimate the magnitude of association between history or parental abuse and the outcome of wife abuse while controlling for confounding and accounting for effect measure modification (i.e., interaction). Below were the results of the modeling process.

Screening for effect measure modifiers using stratified analysis

Stratified analysis was performed in order to screen for effect measure modifiers, as detailed in the Methodology section of the study. Because the literature that identified effect measure modifiers was so insufficient, variables that could logically modify the effect of the exposure variables from a biological perspective were screened for statistical interaction. For instance, some interaction terms explored were: history of parental abuse and age at marriage; history of parental abuse and respondents educational attainment; and, history of parental abuse and per capita household income. Only age at marriage manifested probable interaction with history of parental abuse. The rest of the variables tested did not have significant p-values on test for homogeneity. (see Table 6.3.2.2.1.)

72

Table 6.3.2.2.1. Screening for potential effect measure modifiers for the association between history of parental abuse and wife abuse
Disease (%) Without With wife wife abuse abuse Stratum Specific OR 0.8 2221 (93.4) 17 (94.4) 2755 (90.4) 15 (60.0) 156 (6.6) 1 (5.6) 6.3 291 (9.6) 10 (40.0) 0.166 1015 (93.2) 10 (100.0) 1706 (92.1) 10 (71.4) 2263 (90.8) 12 (63.2) 74 (6.8) 0 (0) 147 (7.9) 4 (28.6) 5.8 228 (9.2) 7 (36.8) 3.1 1530 992.6) 12 (80.0) 3173 (91.4) 17 (68.0) 122 (7.4) 3 (20.0) 4.9 299 (8.6) 8 (32.0) 2.1-11.7 0.9-11.2 0.551 2.2-14.9 Undefined 4.6 1.4-14.9 2.8 - 14.1 95% CI pvalue* 0.044 0.1 - 6.4

Age At Marriage
Married at 25 & older (-) History of parental abuse (+)History of parental abuse Married at 24 & younger (-) History of parental abuse (+)History of parental abuse

Respondents Education
College (reference group) (-) History of parental abuse (+)History of parental abuse High school/Vocational (-) History of parental abuse (+)History of parental abuse No education/Primary (-) History of parental abuse (+)History of parental abuse

Per Capita Income


>Php1,000/mo income (-) History of parental abuse (+)History of parental abuse Php1,000/mo income (-) History of parental abuse (+)History of parental abuse
* Test for homogeneity

Table 6.3.2.2.1 shows the stratified analysis for the exposure variable history of parental abuse. Among those who married at 24 years of age, the odds of beating/hitting their wives by men with a history of parental abuse was 6.30 times greater than those without a history of parental abuse (95% CI=2.81 - 14.16). The odds ratio was less than one for those who married at > 24 years of age (OR=0.85, 95% CI=0.11 - 6.42). The significant p-value of the test for homogeneity of the odds ratio that was less than 0.05 justified inclusion of age at marriage in Model 1 as potential effect measure modifier.

Identification of effect measure modifiers by multiple logistic regression Identification of effect measure modifiers was based on the likelihood ratio test that compared a model that includes history of parental abuse (E) with the model that adds age at marriage (probable effect measure modifier) and the corresponding interaction terms between these two variables (Model 2).

73

Table 6.3.2.2.2. shows that the likelihood ratio test, with p-value of 0.03, identified age at marriage as an effect measure modifier of history of parental abuse and it was retained in the model. No other interaction terms was explored and entered into the model for likelihood ratio testing since on screening by stratified analysis, the rest of the variables did not have significant p-values on test for homogeneity. Table 6.3.2.2.2. Log likelihood ratio test results for the assessment of statistical interaction between history of parental abuse and age at marriage
Variable Log Likelihood LRT p-value

Main effects model Main effect + Effect Measure Modifier + Interaction term: (History of abuse*Age at marriage)a
a

-1558.75 -1556.33

0.03

retained in the model

Identification of Confounders

The full model included history of parental abuse (exposure variable), age at marriage (the identified EM) and their corresponding interaction terms and the 11 probable confounders (determined from the crude analysis). Table 6.3.2.2.3. Odds ratios and p-values of the variables in the full model on association between wife abuse and social-psychological factor Variables in the full model Odds ratio p-value
(+) History of parental abuse b Age at marriage (<24 yrs old) Interaction (History of parental abuse*Age at marriage) Person primarily responsible for household tasks (wife)* Person primarily responsible for handling budget (husband)* Decision-maker on food (husband)* With objection to better education & social well-being* Frequency of discussion on expenses (sometimes)* Frequency of discussion on family planning (sometimes)* Conflict resolution (husband gets his way)* Respondents education (HS/vocational)* Respondents education (No education/Primary)* Wifes education (HS/vocational)* Wifes education (No education/Primary)* Per capita income (=/<Php1,000/mo)*
a: main effect b: effect measure modifier *: probable confounders
a

1.08 1.46 10.02 0.88 1.31 1.31 1.64 0.96 0.79 2.51 1.00 1.03 1.56 1.59 1.12

0.939 0.005 0.051 0.450 0.213 0.040 0.005 0.832 0.136 0.000 0.988 0.882 0.028 0.035 0.463

The

backward elimination process started with the full model from which the

potential confounder that had the highest p-value (respondents education) was deleted.

74

The change-in-the estimate criterion identified the variable person primarily responsible for doing household tasks as the only confounder for the association between history of parental abuse and wife abuse. Specifically, the variable person primarily responsible for doing household tasks confounds the effect of history of parental abuse only among ever-married men who married at 24 years of age or younger. Details are summarized in Table 6.3.2.2.4.

Table 6.3.2.2.4. Test for Confounding (Social Psychological Factors) Factors Full Model (Adjusted) Model without person responsible for doing household tasks (Unadjusted) % change History of Abuse >25 yrs old at Marriage 24 yrs old at Marriage 1.46 (1.12-1.90) 14.62 (1.47-145.24) 1.53 (1.25-1.87) 11.33 (1.29-99.23)

4.79%

-21.48%

Without controlling for this extraneous variable person primarily responsible for household tasks, the association between history of parental abuse and wife abuse among men who married young (24 years old) would have been underestimated by 21%. Thus, controlling for this variable, the association between history of parental abuse and wife abuse was strengthened. Table 6.3.2.2.5 shows that there is absence of joint confounding by the probable confounders that were excluded from the model. After control for all other variables in the model, the association between wife abuse and history of parental abuse was statistically significant. Among those who married young (24 years old), the odds of those with a history of being beaten/hit by one or both parents to beat/hit their wives was 13.14 times those who did not have a history of being beaten/hit by one or both parents (95% CI=1.48-116.75). Among those who married at 25 years of age or older, the odds of those with a history of being beaten/hit by one or both parents to beat/hit their wives was 1.47 times those who did not have a history of being beaten/hit by one or both parents (95% CI=1.19-1.82).

75

Table 6.3.2.2.5. Comparison of the stratum specific odds ratio between the full model and the final model (Social Psychological Factors)
Factors Full model (main effects, significant interaction terms & all 11 probable a confounders ) Final model (main effects model, significant interaction terms & 1 b confounder ) % change History of parental abuse >25 yrs old at marriage 24 yrs old at marriage 1.46 (1.12-1.90) 14.62 (1.47- 145.24) 1.47 (1.19-1.82) 13.14 (1.48-116.75)

0.68%

-8.11%

a: wifes educational status, respondents educational status, per capita monthly income, person primarily responsible for doing certain tasks and for handling the household budget, usual decision maker on food, attitude towards better education and social well being of spouse, frequency of discussion on expenses, on social-political issues, on family planning and conflict resolution b: person primarily responsible for doing certain tasks

The final model is now composed of history of parental abuse as the main effect; age at marriage as the effect measure modifier, and person primarily responsible for household tasks as the confounder. Table 6.3.2.2.6 shows the coefficients and p-values of the variables in the final model. Table 6.3.2.2.6. Coefficients and p-values of the variables in the final model (Social Psychological Factors) Variables in the final model Coefficients p-value (B) Beta Constant -2.48 0.000 (+) History of parental abusea -0.17 0.877 b 0.39 0.000 Age at marriage ( 24 yrs old) Interaction (History of parental abuse*Age at marriage) 2.19 0.050 Person primarily responsible for household tasks (wife)c -0.22 0.040
a: main effect b: effect measure modifier c: confounder

Thus, the final model is: Logit(P)= Where: 0 1 2 = intercept = coefficient for history of parental abuse (exposure variable, E) = coefficient for age at marriage (effect measure modifier, X) = coefficient of interaction terms history of parental abuse (E) and age at marriage (X) 4 = coefficient of person primarily responsible for doing household tasks

0 + 1 E + 2 X + 3 E*X + 4X3

76

Including the value of the coefficients, the final model is as follows:

Logit(P) = -2.48 + -0.17_history of parental abuse + 0.39_age at marriage + 2.19_history of parental abuse*age at marriage + -0.22_person doing household tasks

6.3.2.3. Association between wife abuse and social structural factors

Multiple logistic regression analysis was performed to estimate the magnitude of association between the different social structural factors and the outcome of wife abuse while controlling for confounding and accounting for effect measure modification (i.e. interaction). Below were the results of the modeling process.

Variable Selection

Since this model had multiple exposure variables, only those with p-values less than or equal to 0.25 on crude analysis were included in the first main effects logistic model. Of the ten variables subjected to crude analysis, only eight were included in the main effects logistic model (Table 6.3.2.3.1.). Table 6.3.2.3.1. Odds ratios and p-values of the variables in the first main effects logistic model (Social Structural Factors) Variables in the first main effects model Odds ratio p-value
Person primarily responsible for household tasks (wife) Person primarily responsible for handling budget (husband) Decision-maker on food (husband) With objection to better education & social well-being Frequency of discussion on expenses (sometimes) Frequency of discussion on family planning (sometimes) Conflict resolution (husband gets his way)
0.97 1.47 1.31 1.64 0.98 0.83 2.51 0.842 0.062 0.030 0.003 0.915 0.207 0.000

The importance of exposure variables of interest was verified by backward elimination process utilizing Walds statistic. If the variable obtained a p-value greater than 0.05 in the logistic regression, it was subsequently deleted in the main effects model. Thus for social structural factors, four exposure variables of interest were retained in the main effects model, i.e., person primarily responsible for handling the household budget,

77

usual decision maker on food matters, attitude towards better education and social wellbeing of spouse, and, usual manner of conflict resolution (Table 6.3.2.3.2.) Table 6.3.2.3.2. Odds ratios and p-values of the variables in the final main effects logistic model (Social Structural Factors) Variables in the final main effects model Odds ratio p-value
Person primarily responsible for handling budget (husband) Decision-maker on food (husband) With objection to better education & social well-being Conflict resolution (husband gets his way)
1.51 1.35 1.71 2.67 0.020 0.009 0.000 0.000

Screening for effect measure modifiers using stratified analysis

Screening for effect measure modifiers for any of the four main effects variables using stratified analysis was done. Because the literature that identified effect measure modifiers was so scanty, variables that could logically modify the effect of the exposure variables from a biological perspective were screened for statistical interaction. As detailed in the Methodology section, for the identified main effect social structural variables, only interaction between person primarily responsible for handling the household budget and respondents educational attainment was explored. Table 6.3.2.3.2. shows that these interaction terms tested did not have significant p-value on test for homogeneity and hence was not considered a potential effect measure modifier of one of the four main effects variables. Table 6.3.2.3.3. Screening for potential effect measure modifiers for the association between a social structural factor and wife abuse.
Disease (%) Stratum Specific OR 2.57 910 (94.2) 101 (86.3) 1574 (91.8) 151 (91.5) 2047 (90.9) 209 (88.6) 56 (5.8) 16 (13.7) 1.04 138 (8.2) 14 (8.5) 2.28 206 (9.1) 27 (11.4) 0.84-1.96 0.59-1.85 95% CI pvalue* 0.068 1.42-4.66

Respondents education
College (reference group) Wife predominantly handles household budget Husband predominantly handles household budget High school/Vocational Wife predominantly handles household budget Husband predominantly handles household budget No education/Primary Wife predominantly handles household budget Husband predominantly handles household budget
* Test for homogeneit/y

78

Identification of effect measure modifiers by multiple logistic regression Since the stratified analysis did not point to any effect measure modifiers, likelihood ratio testing was not done anymore and no effect measure modifiers nor their corresponding interaction terms was entered into the multiple logistic regression model.

Identification of Confounders

The full model included person primarily responsible for handling the household budget, usual decision maker on food matters, attitude towards better education and social well-being of spouse and usual manner of conflict resolution (exposure variables), no effect measure modifiers and the 9 probable confounders (determined from the crude analysis). Table 6.3.2.3.4. Odds ratios and p-values of the variables in the full model on the association between wife abuse and social structural factors Variables in the full model Odds ratio p-value
Person primarily responsible for handling budget (husband) a Decision-maker on food (husband) a With objection to better education & social well-being a Conflict resolution (husband gets his way) Age at marriage ( 24 yrs old)* (+) History of parental abuse* Person primarily responsible for household tasks (wife)* Frequency of discussion on expenses (sometimes)* Frequency of discussion on family planning (sometimes)* Respondents education (HS/vocational)* Respondents education (No education/Primary)* Wifes education (HS/vocational)* Wifes education (No education/Primary)* Per capita income ( Php1,000/mo)*
a: main effect *: probable confounders
a

1.29 1.30 1.62 2.52 1.53 5.62 0.89 0.95 0.81 1.01 1.06 1.57 1.56 1.09

0.234 0.044 0.006 0.000 0.001 0.000 0.512 0.762 0.155 0.965 0.796 0.025 0.043 0.522

The confounding effect of the potential confounders was verified through the change in estimate criterion, arbitrarily set at 10%, comparing the adjusted and unadjusted odds ratios for the potential confounder. The process started with full model from which the potential confounder that had the highest p-value (respondents education) was deleted. Since the change-in-the estimate was < 10%, it implied that respondents education was not a confounder and it

79

was not retained in the model. The process of comparing the working model with the full model was repeated until potential confounders could not be deleted from the model since the change of estimate was 10%. This change-in-the estimate criterion identified age at marriage, history of abuse and person primarily responsible for doing household tasks as confounders for the association between the main social structural effects variables and wife abuse. Details are summarized in Table 6.3.2.3.5. Table 6.3.2.3.5. Test for confounding (social structural factors) Exposure Variables Probable confounders Person Decision Attitude towards Responsible Maker on better education for Handling Food & social well Budget Matters being of spouse Full Model (Adjusted) 1.29 1.302 1.62 Model without age at marriage (Unadjusted) % change Model without history of parental abuse (Unadjusted) % change Model without person responsible for doing household tasks (Unadjusted) % change 1.45 12.3% 1.44 1.304 0.2% 1.34 1.67 3.3% 1.58

Conflict Resolution

2.52 2.51 -0.7% 2.58

10.6% 1.46

2.5% 1.35

-2.2% 1.65

2.4% 2.80

12.4%

3.6%

2.3%

11.2%

In this study, the effect of handling of household budget on wife abuse was found to be confounded by age at marriage, history of parental abuse and attitude towards better education and social well being of spouse. Had these extraneous variables not been controlled in the analysis, the association between handling of household budget and wife abuse would have been overestimated by 10.6-12.4%. Specifically, the estimated odds ratio would be positively biased away from the null. The variable person primarily responsible for doing household tasks confounded the effect of conflict resolution on wife abuse. Had this extraneous variable not been controlled in the analysis, the association between usual manner of conflict resolution and wife abuse would have been overestimated by 11%. Specifically, the estimated odds ratio would be positively biased away from the null (ORc=2.80 vs. Ora=2.52).

80

Table 6.3.2.3.6 shows that there is absence of joint confounding by the probable confounders that were excluded from the model. Men who predominantly handle the household budget are 1.42 times more likely to beat/hit their wives than men who allow their wives to handle the budget, though the association is not significant (95% CI= 0.95-2.13). After control for all other variables in the model, the association between wife abuse and decision-maker on food matters, attitude towards better education and social well being of spouse and usual manner of conflict resolution were statistically significant. Men who, predominantly or jointly with their wives, decide on what food to buy or prepare for family meals, are 1.33 times more likely to beat/hit their wives than men who do not usually decide on food matters and allow their wives to do so (95% CI=1.041.69). Men who object to their spouse/partners improving their education or visiting their friends without him are 1.57 times more likely to beat/hit their wives than men who do not object to such matters (95% CI=1.12-2.19). Men who usually get their way when there are disagreements are 2.64 times more likely to beat/hit their wives than men who usually resolve such disagreements by couple compromise (95% CI=1.70-4.09). Table 6.3.2.3.6. Comparison of the odds ratio of the main social structural variables between the full model and the final model Factors Full model (main Final model % change effects & all 9 (main effects probable model & 3 confoundersa) confoundersb) Handling the household budget 1.29 (0.84-1.99) 1.42 (0.95-2.13) 9.2% (husband) Decision making on food matters 1.30 (1.01-1.68) 1.33 (1.04-1.69) 1.5% (both/husband) With objection to better education 1.62 (1.16-2.31) 1.57 (1.12-2.19) -1.3% & social well-being Conflict resolution (respondent gets 2.51 (1.59-3.95) 2.64 (1.70-4.09) 5.3% his way)
a: person primarily responsible for doing certain tasks, frequency of discussion on expenses, on socialpolitical issues, on family planning, age at marriage, history of parental abuse, wifes educational status, respondents educational status, and per capita monthly income, b: age at marriage, history of parental abuse, and person primarily responsible for doing certain tasks

The final model is thus composed of person primarily responsible for handling the household budget, usual decision maker on food matters, attitude towards better education and social well-being of spouse and usual manner of conflict resolution as the main effect variables; and, age at marriage, history of abuse and person primarily shows the

responsible for doing household tasks as confounders. Table 6.3.2.3.7. coefficients and p-values of the variables in the final model.

81

Table 6.3.2.3.7. Coefficients and p-values of the variables in the final model (social structural factors) Variables in the final model Coefficients p-value (B) Beta Constant -2.94 0.000 a Handling the household budget (husband) 0.35 0.090 Decision making on food matters (both/ husband)a 0.29 0.026 With objection to better education & social well-beinga 0.45 0.006 a Conflict resolution (respondent gets his way) 0.97 0.000 b 0.43 0.001 Age at marriage ( 24 yrs old) (+) History of parental abuseb 1.68 0.000 b Person primarily responsible for household tasks (wife) -0.07 0.726
a: main effect b: confounders

Thus, the final model is: Logit (P)=

0 + 1 X1 + 2 X2 + 3 X3 + 4X4+ 5X5 + 6X6 +7X7

Where: 0 1 2 3 4 5 6 7 Logit (P) = intercept = coefficient for person responsible for handling household budget = coefficient for decision maker on food matters = coefficient of attitude towards better education and social well being of spouse = coefficient of conflict resolution = coefficient of age at marriage = coefficient of history of parental abuse = coefficient of person primarily responsible for doing household tasks

= -2.94 + 0.35_person handling household budget + 0.29_ decision maker on food matters + 0.45_attitude towards better education and social well being of spouse + 0.97_conflict resolution + 0.43_age at marriage + 1.68_history of parental abuse + -0.07_person primarily responsible for doing household tasks

82

6.3.2.4. Association between wife abuse and social location factors

Multiple logistic regression analysis was performed to estimate the magnitude of association between the different social location factors and the outcome of wife abuse while controlling for confounding and accounting for effect measure modification (i.e. interaction). Below were the results of the modeling process.

Variable Selection

Six variables were subjected to crude analysis and of these, only three were included in the main effects logistic model (Table 6.3.2.4.1.). Table 6.3.2.4.1. Odds ratios and p-values of the variables in the first main effects logistic model (social location factors) Variables in the first main effects model Odds ratio p-value
Wifes education (HS/vocational) Wifes education (No education/Primary) Respondents education (HS/vocational) Respondents education (No education/Primary) Per capita income (=/<Php1,000/mo)
1.69 1.67 1.04 1.16 1.03 0.002 0.004 0.808 0.394 0.820

However, Table 6.4.2.4.2. shows that based on the result of Walds statistic, only one exposure variable of interest, wifes educational attainment, was retained in the main effects model. Table 6.4.2.4.2. Odds ratios, 95% confidence interval estimates, and p-values of the social location variables in the final main effects model Variable OR (95% CI) p-value Wifes education (HS/vocational) 1.68 (1.28 - 2.22) 0.000 Wifes education (No education/Primary) 1.73 (1.32 2.26) 0.000

Screening for effect measure modifiers using stratified analysis

Not one of the probable interaction terms tested had significant p-values on test for homogeneity. Table 6.4.2.4.3. shows that there was no potential effect measure modifier for the main effects variable of wifes educational attainment.

83

Table 6.4.2.4.3. Screening for potential effect measure modifiers for the association between the social location factor wifes education and wife abuse.
Disease (%) Stratum Specific OR 95% CI pvalue* 0.769 305 (93.0) 419 (89.1) 325 (89.8) 967 (94.7) 1336(91.3) 1615(90.6) 23 (7.0) 51 (10.9) 38 (10.2) 54 (5.3) 128 (8.7) 167 (9.4)

Frequency of discussion on family planning


Often College (reference group) High school/Vocational No education/Primary Sometimes College (reference group) High school/Vocational No education/Primary 1.61 1.55 0.94-2.79 0.88-2.76

1.72 1.85

1.22-2.42 1.33-2.58 0.804

Frequency of discussion on expenses


Often College (reference group) High school/Vocational No education/Primary Sometimes College (reference group) High school/Vocational No education/Primary
* Test for homogeneity

266 (93.0) 286 (88.8) 267 (89.9) 1012(94.6) 1475(91.1) 1700(90.7)

20 (7.0) 36 (11.2) 30 (10.1) 58 (5.4) 144 (8.9) 175 (9.3)

1.67 1.49

0.91-3.08 0.80-2.81

1.70 1.80

1.23-2.36 1.31-2.49

Identification of effect measure modifiers by multiple logistic regression As the stratified analysis did not reveal any probable effect measure modifiers, likelihood ratio testing was not done anymore and no effect measure modifiers nor their corresponding interaction terms was entered into the multiple logistic regression model. Identification of Confounders

The full model included wifes educational attainment (exposure variable) and the 12 probable confounders (determined from the crude analysis). Details are shown in table 6.4.2.4.3.

84

Table 6.4.2.4.3. Odds ratios and p-values of the variables in the full model on the association between wife abuse and social location factors
Variables in the full model a Wifes education (HS/vocational) a Wifes education (No education/Primary) Respondents education (HS/vocational)* Respondents education (No education/Primary)* Per capita income (=/<Php1,000/mo)* Age at marriage ( 24 yrs old)* (+) History of parental abuse* Person primarily responsible for household tasks (wife)* Person primarily responsible for handling budget (husband)* Decision-maker on food (husband)* With objection to better education & social well-being* Frequency of discussion on expenses (sometimes)* Frequency of discussion on family planning (sometimes)* Conflict resolution (husband gets his way)*
a: main effect *: probable confounders

Odds ratio 1.57 1.56 1.01 1.06 1.09 1.53 5.62 0.89 1.29 1.30 1.62 0.95 0.81 2.52

p-value 0.025 0.043 0.965 0.796 0.522 0.001 0.000 0.512 0.234 0.044 0.006 0.762 0.155 0.000

The process of identifying confounders started with full model from which the potential confounder that had the highest p-value (respondents education) was deleted. Among the twelve probable confounders, only age at marriage emerged as an actual confounder for the association between wifes educational attainment and wife abuse. Details are summarized in Table 6.4.2.4.4. Table 6.4.2.4.4. Test for confounding (social location factors) Factors Wifes Education Wifes Education HS/Vocationala No education/Primarya Full Model (Adjusted) 1.57 1.57 Model without age at marriage (Unadjusted) % change
Note: a - main effect variable

1.64 4.5%

1.74 10.8%

In this study, wifes educational attainment effect on wife abuse was found to be confounded by the variable age at marriage. Had this extraneous variable not been controlled in the analysis, the association between wifes educational attainment and wife abuse would have been overestimated by close to 11%. Specifically, the estimated odds ratio would be positively biased away from the null (ORc=1.74 vs. Ora=1.57). Table 6.4.2.4.5. shows that there is absence of joint confounding by the probable confounders that were excluded from the model. After control for all other variables in the

85

model, the association between wife abuse and wifes educational attainment was statistically significant. Men are 1.59 times more likely to beat/hit their wives when the wives had high school education only than when the wives had college education or higher. Men are 1.61 times more likely to beat/hit their wives when the wives had no education or had reached primary level of education only than when the wives had college education or higher. Table 6.4.2.4.5. Comparison of the odds ratio between the full model and the final model (social location factors) Factors Full model (main Final model % change effects & all 12 (main effects probable model & 1 b confoundersa) confounder ) Wifes education (HS/vocational) Wifes education (No education/ Primary ) 1.57 (1.06-2.33) 1.56 (1.01-2.39) 1.59 (1.20-2.10) 1.61 (1.22-2.11) 1.3% 3.8%

a: respondents educational status, per capita monthly income, age at marriage, history of parental abuse, person primarily responsible for doing certain tasks, handling the household budget, decisionmaker on food matters, attitude towards better education and social well-being of spouse, frequency of discussion on expenses, on social- political issues, on family planning and conflict resolution b: age at marriage

The final model is now composed of wifes educational attainment as the main effect variable; and, age at marriage as confounder. Table 6.4.2.4.6. shows the coefficients and p-values of the variables in the final model.

Table 6.4.2.4.6. Coefficients and p-values of the variables in the final model (social location factors) Variables in the final model Coefficients (B) p-value Beta Constant -2.99 0.000 a Wifes education (HS/vocational) 0.46 0.001 Wifes education (No education/Primary)a 0.47 0.001 Age at marriage( 24 years old)b 0.38 0.000
a: main effect b: confounder

Thus, the final model is: Logit (P)=

0 + 1 X1 + 2 X2 + 3 X3

Where: 0 = intercept

86

1 2 3

= coefficient for wifes education (high school/vocational) = coefficient for wifes education (no education/primary) = coefficient of age at marriage

Logit (P) = -2.99 + 0.46_ wifes education (high school/vocational) + 0.47_ wifes education (no education/primary) + 0.38_ age at marriage

87

7
DISCUSSION

The following sections discuss the analytic techniques, the results, and possible sources of bias in the estimation of the measures of association between the outcome, wife abuse, and the exposure variables namely: marital factors (age at marriage), socialpsychological factors (history of parental abuse), social structural factors (person primarily responsible for doing household chores, decision making on food matters, attitudes towards better education and social well-being of spouse, frequency of discussion on expenses, political & social issues, family planning and usual manner of resolving disagreements) and social location factors (wifes education, respondents education, per capita monthly income).

7.1 ANALYTIC TECHNIQUES

The possible inclusion of an irrelevant variable in the cluster analysis would increase the chance that outliers would have been created on these variables, which could have a substantive effect on the results. In this study therefore, efforts have been made not to include variables indiscriminately. Rather, careful choice of the variables were made, with the help of results of previous studies and experts opinion. The research objectives were used as the criteria for selection of which sets of items to subject to cluster analysis to create specific variables. In the inferential analysis, the p-values for statistical significance were set arbitrarily. For instance, the crude analysis was patterned after Hosmer and Lemeshows algorithm that uses a p-value of 0.25 in order to avoid excluding important exposure as well as extraneous variables. Thus, it was deemed necessary to subject variables that were weakly associated with the outcome such that their p-values on crude analysis were 0.25, to further analyses. Meanwhile, in the identification of probable effect measure modifiers, the p-value of the test for homogeneity of the odds ratio was set at less than 0.05 since sample size was relatively large. Assessment for effect measure modification (interaction) was performed prior to identification of confounders. According to Hosmer and Lemeshow, when a covariate is

88

an effect modifier, its status as a confounder is of secondary importance since the estimate of the effect of the risk factor depends on the specific value of the covariate. The choice of interaction terms to be tested should be based on a priori knowledge. However, because literature on this matter was scarce, even those without a priori knowledge but made sense from a biologic perspective were tested. Interaction terms were identified in two of the four models built through logistic regression. As cautioned by Hosmer and Lemeshow, it is important to keep in mind that the fundamental reason for developing a model is to provide as clear a description as is possible with the available data of the associations between outcome and covariates. If entering an additional term into the model improves our estimates of the relevant associations then that term should be put into the model regardless of its statistical significance. Significance means that interactions must contribute to the model, altering both point and interval estimates. These interactions however, do not necessarily imply biological interaction or interdependence of effect of two or more factors in the occurrence of the outcome. According to Rothman and Greenland (1998), statistical interaction is a phenomenon whose presence or absence, as well as magnitude, is entirely determined by the scale chosen for measuring additivity of effects. The relative risk for interaction or ICR is zero when there is no biological interaction. In the interaction between age at marriage and history of parental abuse for instance: ICR = RR11 - RR10 - RR01 + 1 = 0, where: RR11 = R11/R00 = risk among ever married males who were married at 24 years of age and have a history of parental abuse relative to the risk among those who were married at > 24 years of age and did not have a history of parental abuse. RR10 = R10/R00 = risk among ever married males who were married at 24 years of age and did not have a history of parental abuse relative to the risk among those who were married at > 24 years of age and did not have a history of parental abuse. RR01 = R01/R00 = risk among ever married males who were married at > 24 years of age and have a history of parental abuse relative to the risk among

89

those who were married at > 24 years of age and did not have a history of parental abuse. The measure of effect for this study is the prevalence odds ratio, since this investigation utilized a cross-sectional design. The risk period, identified as the year before the interview, is restricted, and the use of the prevalence odds ratio is desired (Kleinbaum et al, 1982). The prevalence odds ratio may be used to estimate the risk ratio if the following conditions are met: 1) subjects are randomly selected from a stable source population under steady state condition, 2) exposure does not affect the mean duration of the disease, 3) disease does not affect exposure, and 4) disease is rare. Since the prevalence of wife abuse among ever-married males in selected regions in the Philippines was 8.4%, wife abuse cannot be considered as a rare outcome. Hence, it is not possible to assess whether the observed statistical interaction between age at marriage and history of parental abuse can be considered a biological interaction. If we assume this to be so, then it means that the effect of age at marriage on wife abuse varies with history of parental abuse and vice-versa.

7.2 VALIDITY ISSUES

Results may be compromised due to impediments in the internal validity known as biases. These biases distort the estimate of the effect of the parameter due to the selection of subjects into the study population or into the analysis (selection bias); due to the manner in which the information was obtained (information bias); or due to the effects of uncontrolled extraneous variables that are themselves risk factors of the outcome, are associated with the exposure but are not intermediate variables (confounding bias).

7.2.1

Selection bias

Selection bias is the distortion in the estimate of the effect that result from flawed procedures used to select subjects and from factors that influence study participation. The common element of such biases is that the relation between exposure and disease is different for those who participate and those who should be theoretically eligible for study, including those who did not participate. The result is that associations observed in the study represent a mix of forces determining participation, as well as forces determining

90

outcome. Since wife abuse was just one of the many variables included in

the

Reproductive Health Survey 2000 from which the data used in this study was taken, it is difficult to imagine how the probability of selection of respondents could have been differentially influenced by the occurrence of domestic violence among the exposed and nonexposed.

7.2.2

Information bias

The distortion in the estimate of effect which can occur from errors in obtaining the needed information, is referred to as information bias. It can occur whenever there are errors in the measurement of subjects, but the consequences of the errors are different, depending on whether the distribution of errors for one variable (e.g., exposure or disease) depends on the actual value of other variables. For discrete variables, measurement error is usually called classification error. Classification error that depends on the values of other variables is referred to as differential misclassification while classification error that does not depend on the values of other variables is referred to as nondifferential misclassification. Nondifferential exposure or disease misclassification occurs when the proportion of subjects misclassified on exposure does not depend on disease status or when the proportion of subjects misclassified on disease does not depend on exposure. The bias caused by differential misclassification can either exaggerate or underestimate an effect. On the other hand, any bias introduced by non-differential misclassification is predictable in direction, namely, towards the null value, if the following conditions were met: first, if the misclassified variable has two categories (dichotomous); second, if the misclassification of the variable is independent of the misclassification of the other variables; and lastly, if there were no other sources of bias, a condition which would be difficult to satisfy. Hence, nondifferential misclassification can sometimes produce bias away from the null, as when exposure is polytomous (i.e., has more than two categories) or when the classification errors are dependent. Misclassification of the outcome could have biased the results of this study. The classification of a respondent being an abusive husband/partner was based solely on mens reports of violence, which may underestimate the true extent of this abusive behavior. As the question specifically asked the men if they had ever beaten or hit

91

(pinagbuhatan ng kamay which is literally translated as lifted ones hands against, which in the Filipino culture is slapping) their spouse or partner, there is the possibility that other forms of physical violence (e.g. hair pulling, shoving, use of any object that can harm or kill) could have been under-reported. The WorldSAFE Study on domestic violence among urban Filipino women reported that the leading types of physical maltreatment were slapping, kicking and beating repeatedly (Ramiro et al, 1998). Hopefully, the field instruction to interviewers to probe for all types of physical maltreatment has minimized such under-reporting. Also, since the classification of a respondent being an abusive husband/partner was based solely on mens reports of violence, it is quite possible that they may not be willing to admit the fact that they beat their wives. This self-presentation bias may again underestimate the true extent of this abusive behavior. However, the Reproductive Health Survey 2000 from which the data in this study was taken, reported that the ever-married females who were living in the same regions as the ever married male respondents, had a lifetime prevalence of physical violence from any member of the household of 8.68%. Of these married females who have ever been hit or beaten, in close to 90% of them, their husbands are the perpetrators of violence. Though this study dealt with one-year prevalence estimates while that of ever-married females dealt with lifetime prevalence estimates, it would appear that the prevalence estimates based on mens reports of physical violence against their female partners did not differ markedly from the womens reports of their experiences. In the event of misclassification of the outcome, it would most probably be nondifferential, i. e, it is independent of the exposure variables (history of abuse, person primarily responsible for handling the household budget, usual decision maker on food matters, attitude to better education and social well-being of spouse, usual manner of conflict resolution and wifes highest educational attainment). Then, as in most situations, such nondifferential misclassification of a binary outcome would most probably produce bias towards the null, provided that the misclassification is independent of other errors. In this study, the social structural variables primarily responsible for handling the household budget, usual decision-maker on food matters, attitude to better education and social well being of spouse, usual manner of conflict resolution could probably be subject to temporal ambiguity. Since this is a cross-sectional study in which exposure and outcome status were determined at a specific point in time, there is no assurance that the

92

exposure variables actually preceded the outcome. However, since the risk period for the outcome was identified as just the year before the interview, then it is very probable that there has been no change in these social structural exposure variables as this period of time was quite short.

7.2.3. Confounding bias

Confounding is the bias in the crude estimate that can result when the exposuredisease relationship under study is mixed up with the effects of extraneous variables. To be a confounder, the extraneous variable must have three characteristics, namely (1) it must be a risk factor for the disease; (2) it must be associated with the exposure under study in the source population; and (3) it must not be affected by the exposure or the disease. Confounding (regardless of the number of variables eligible for control) requires consideration both of causal relationships believed to be operating in the target population and databased associations. The theoretical eligibility requirement for any extraneous factor is that, on the basis of information external to the study, this factor be determined to be a non-intervening risk factor for the disease. The databased criterion for confounding involving a single risk factor is that the estimated crude measure of the effect differ from an adjusted estimate of the effect. In this study, in identifying extraneous factors eligible for control, effort was made to distinguish intervening variables from risk factors. The change-in-the estimate criterion, arbitrarily set by many at 10%, was used to identify confounders. When several risk factors have been identified and measured for possible control, two fundamental principles about control for confounding need to be considered: (1)Failure to consider the combined effects of all risk factors (joint confounding) measured in the study when assessing confounding may lead to the retention or even the introduction of confounding (Fisher and Patil, 1974); and (2) Not all the variables in a given list of risk factors may need to be controlled. Thus, with regard to bias removal, it is sufficient to use any subset of risk factors on the list that eliminates such bias, including the one consisting of all the risk factors. In this study, assessment for joint confounding by the variables that were deleted from the model was made by comparing the reduced model (without the probable confounder) with that of the full model (with all the risk factors). There was less than 10% change in the measure of association (OR) implying that there was no joint confounding.

93

There were other risk factors identified in other researches which was included in the conceptual framework but of which no data was available for this study. As such, we were not able to control for these potential confounders. If the confounding is strong and the exposure-disease relation is weak or zero, misclassification of the potential confounder can lead to extremely misleading results. For example, a strong causal relation between age at marriage and wife abuse coupled with a strong association between age at marriage and wifes educational attainment makes age at marriage a strong confounder of any possible relation between wifes educational attainment and wife abuse. Since the control of confounding by age at marriage depends on accurate information and since some misclassification of the relevant information is inevitable no matter how age at marriage is measured, some residual confounding is inevitable (Morrison, 1982). Also, the problem of residual confounding could have been aggravated when the variable age at marriage was dichotomized (24 years old versus >25 years old) since the lack of detailed specification of age at marriage prohibits adequate control of confounding.

7.3.

CORRELATES OF WIFE ABUSE

The prevalence odds ratio was computed from the coefficients in the final logistic regression model to determine the correlation between wife abuse and the various independent variables. These study variables included social-psychological factors (history of parental abuse),social structural factors (person primarily responsible for doing household chores, decision making on food matters, attitudes towards better education and social well-being of spouse, frequency of discussion on expenses, political & social issues, family planning and usual manner of resolving disagreements) and social location factors (wifes education, respondents education, per capita monthly income). The identified correlates of wife abuse were age at marriage, history of parental abuse, person primarily responsible for handling the household budget, usual decision maker on food matters, attitude to better education and social well-being of spouse, usual manner of conflict resolution and wifes highest educational attainment.

7.3.1. Association Between Wife Beating and Marital Factors

94

7.3.1.1. Association Between Wife Beating and Age at Marriage

Several authors and their colleagues, e.g., Fagan & Browne (1994), Martin et al (1999), Howell & Pugliesi (1988), MacEwen & Barling (1988), and Pan, Neidig, & OLeary (1994), found that age at marriage is negatively associated with partner aggression. That is, being younger increases the risk for engaging in partner aggression. Kantor and Jasinski (1994) stated that youth is probably one of the strongest risk factor for aggression. The relationships between domestic violence and age when entering marriage is proposed to be mediated by its effect as a stressor that potentially influences to varying degrees patterns of managing disputes within the family. A consequence of early marriage on the part of the husband is the inability to finish his studies, which, in turn would negatively affect his capability to get a stable, and well paying job. What could happen is that such men would have their families living in poverty. They may suffer from stress because of their inability to meet their needs with the resources available to them. This stress may lead to frustration and possibly aggression. Likewise, very young and inexperienced men who enter into a marriage may not have the necessary knowledge and sense of responsibility to plan their family, consequently, they usually have numerous children. Having multiple children would impose increasing burdens on income, hereby inducing stress. In this study, the association between age at marriage and wife abuse was not significant among those without a history of parental abuse (OR=1.09, CI= 0.14-8.44). It was very strong (OR=10.78; 95% CI: 4.21-27.59) among those with the history of parental abuse. Thus, the presence or absence of history of parental abuse was identified as an effect measure modifier of such an association. As discussed in the Methodology Section, statistical interaction does no necessarily imply biological interaction. However, if we can assume that the observed statistical interaction implies that there is synergism between age at marriage and history of parental abuse, then this could probably be mediated through mens feeling of powerlessness because of low self-esteem. When a man would get married at a young age, at which time he may not have finished his schooling yet, he would be unable to get a stable and well-paying job. Unemployment and poverty might reduce the ability of men

95

to attain certain ideals of successful manhood, particularly those based on ideas of men as providers for the family, leading to low self-esteem. This may give rise to frustration and may subsequently enhance the probability not just of wife abuse but child abuse as well. On the other hand, a history of violence in the family of origin may lead a man to get married young in order to get away from his dysfunctional family-of-origin. And so the vicious cycle goes. Selection bias was unlikely to have occurred in this study because the respondents were cross-sectionally selected at random. Thus, the probability of selecting those who married young and those who married at 25 years of age or older are equal among those married men who gave in to wife abuse and among those men who do not abuse their wives. Misclassification of the category of the respondents according to age at marriage was also unlikely. However, as previously discussed, misclassification of outcome status may be possible, resulting in a negative bias (bias towards the null). In this study, income, person primarily responsible for household tasks and attitude towards better education and social well being of spouse confounded the effect of age at marriage on wife abuse. Had these identified confounders not been controlled, the association between age at marriage and wife abuse would have been underestimated. Since this study has shown the very strong association between age at marriage and wife abuse particularly among those with history of parental abuse, programs for prevention of violence against women need to focus on counselling adolescents as to the drawback of getting married at a young age. This can be incorporated in the school curriculum. Efforts to reach out to out-of-school youths during information-educational campaigns should be made, with the help of government and non-government organizations.

7.3.1.2. Marital Factors Not Associated with Wife Beating

A review of the literature showed conflicting results on the association of duration of marriage with wife abuse. One study reported positive association between being in a marriage of a longer duration (> 5 years) and abuse (Martin, Tsui, Maitra, & Marinshaw,

96

1999) while another investigation noted higher rates of spousal violence in the early marriage period, i.e., one and a half (1 1/2) years after marriage (OLeary et al, 1989). In this study, duration of marriage was not found to be independently associated with domestic violence. Our assumption is that, due to publication bias, previous investigations that likewise found no association were not published.

7.3.2. Association Between Wife Beating and Social Psychological Factors

The association between wife abuse and history of parental abuse was significant after controlling for all other variables in the model. Age at marriage was an identified effect measure modifier of such an association. Among those who married at 24 years of age or less, the odds of wife beating/hitting for men with a history of having been physically abused by one or both parents was 13.08 times (95% CI=1.47-116.14) the odds of those without history of parental abuse. The odds ratio of 1.49 among those who were 25 years old or older when they were married was much lower but it was nevertheless significant (95% CI=1.20-1.84). These relatively older men may have had a longer time learning how to cope with having been abused by their parents and growing in emotional maturity and thus may be more in control of possible violent tendencies. An important theory of domestic violence causation relates to the intergenerational cycling of violence. Our finding that being beaten in childhood was important for mens physical aggression against his partner later on, are in line with those reported elsewhere (Barnett & Fagan, 1993; Caesar, 1988; Dutton, Starzomski, & Ryan, 1996; Malone, Tyree, & OLeary, 1989; MacEwen & Barling, 1988; OLeary & Curley, 1986; Telch & Lindquist, 1984). In these previous studies, the Pearson correlation coefficient ( r ) ranged from 0.14 to 0.31, indicating a relatively weak to slightly strong positive correlation between child physical abuse and current partner aggression. In some studies overall abuse by parents was assessed, whereas in others abuse by mother and father were assessed separately. Overall, physical abuse by either parents or both seems to be about equally predictive of later partner physical aggression. Dutton et al. (1996) and Barnett and Fagan (1993) found that experience of child physical abuse as well as verbal and child psychological abuse predict later partner physical aggression. This is important because it is often assumed that child physical abuse has a more lasting, profound impact than other types of abuse. Caesar (1988)

97

found that the parents of physically aggressive men were not only more likely to be more abusive, but that they also were significantly less likely to use nonviolent punishments such as restriction or grounding than the parents of non-aggressive men (r = -0.44). Instead they favored physical punishments such as hitting with a belt or switch (r = 0.27). The result of this study lends empirical evidence for the social learning of violence within families. A basic premise underlying this association between wife abuse and history of parental abuse is that the family is the training ground for violence and even the use of ordinary physical punishment such as spanking is associated with wife beating. Some men from violent homes learn in childhood to model the parental behaviors observed which place them at subsequent risk for beating their wives. These may relate to viewing marital violence as normative with tolerance of types of behaviors which may ultimately be associated with domestic violence. The association between physical punishment in childhood and adult domestic violence suggests that beating teaches children the normality of using violence in punishment and conflict situations. Observation of violence in childhood may also impact on a mans self-esteem, which may in turn increase marital stress and approval of marital violence. Selection bias was unlikely to have occurred in this study. The probability of selecting those who had history of parental abuse and those who did not are equal among those married men who gave in to wife abuse and among those men who do not abuse their wives. The degree of bias resulting from misclassification of the category of the respondents according to history of parental abuse is likely to be low. However, as previously discussed, misclassification of outcome status may be possible, resulting in a negative bias towards the null, i.e., the association would have been stronger if not for such misclassification bias. In this study, the crude effect of history of parental abuse on wife abuse was found to be confounded by the variable person primarily responsible for household tasks. Without controlling for this extraneous variable, the association between history of parental abuse and wife abuse among men who married young (</= 24 years old) was underestimated by 21%. Thus, controlling for the variable person primarily responsible for household tasks, the association between history of parental abuse and wife abuse

98

was strengthened. Since it has been identified that history of parental abuse is a risk factor to domestic violence, there is a need to strengthen family ties and values so as to reduce normative use of violence in conflict situations between couples as well as between parents and children on the excuse of disciplining them or correcting misbehaviour. Anti-Domestic Violence Bills could be proposed to policy-makers and the lawmaking body of the country to provide legal protection to prospective and current victims, not just of wife abuse but child abuse as well. Strategies for prevention of violence against women need also to address prevention and treatment of child abuse as it has been shown to be a risk factor. The first priority is to identify and protect child and adult victims of family abuse. Battered children in particular should have easy access to medical and legal assistance, as well as being given shelter when needed. Time, patience and persistence are needed to allow victims to disclose their secrets and to cope with the trauma and psychological consequences. Personnel trained in family violence issues must educate, train and consult with health and mental health service providers, as well as school authorities, to ensure appropriate detection, referral and treatment. All health and mental health service providers, along with school authorities, must educate themselves about domestic violence, its nature and effects, and the complex societal factors that prevent or facilitate disclosure, help seeking, and change in family systems. (Giles-Sims, 1998). Children who had been harshly punished or abused in other ways and children who had exposure to marital violence in the family of origin need to be subjected to counselling and even child psychiatric treatment for debriefing from their traumatic experience.

7.3.3. Association Between Wife Beating and Social Structural Factors

The identified social structural correlates of wife abuse were person primarily responsible for handling the household budget, usual decision maker on food matters, attitude to better education and social well-being of spouse and usual manner of conflict resolution.

7.3.3.1 Association Between Wife Beating and Person primarily

99

responsible for handling the household budget

Another factor that has been implicated in intimate partner violence is in connection with household income distribution. Intimate partner violence is higher among women who reported that their husband keeps some or all of his earnings (20%), than when the husband turns over all his earnings (10%) (Hindin, 2002). This study explored if the corollary hypothesis, i.e., that when the husband does not turn over all of his income and instead handles the household budget he is more prone to intimate partner violence, was supported by the data. While the results were not statistically significant (OR=1.42, 95% CI=0.95-2.13), there is an indication that husbands who control the household budget could be twice more likely to physically violate their spouse. In a society where husbands turning over all their income is socially expected so that women can manage the budgets, our analyses show that deviations from this expectation was a predictor of higher rates of wife abuse, resembling the findings in the study of Hindin et al (2002). The result of this study lends additional support to the evidence from previous researches which suggested that wife beating is more common in households where power is concentrated in the hands of the husband or male partner (Coleman & Straus, 1990; Levinson, 1989; Straus et al, 1980; Yllo & Straus, 1990). A basic premise underlying this association between wife abuse and person primarily responsible for handling the household budget is that higher levels of dominance and control over the partner to gain the upper hand in the relationship serves as the background for violence. Selection bias was unlikely to have occurred in this study. The probability of selecting those households in which the respondent or his wife have the independent variables of interest and those households in which the respondent or his wife do not have the independent variables of interest are equal among those married men who gave in to wife abuse and among those men who did not abuse their wives. The degree of bias resulting from misclassification of the category of the respondents according to exposure variables such as person primarily responsible for handling the household budget, is likely to be low. However, as previously discussed, misclassification of outcome status may be possible, resulting in a negative bias (bias

100

towards the null). Therefore, the association would even be stronger. In this study, the crude effect of handling of household budget on wife abuse was found to be confounded by age at marriage, history of parental abuse and attitude towards better education and social well being of spouse. Had these extraneous variables not been controlled in the analysis, the association between handling of household budget and wife abuse would have been overestimated by 10.5 - 12%. Specifically, the estimated odds ratio would be positively biased away from the null.

7.3.3.2 Association Between Wife Beating and Decision making on food matters

In this study, the odds of wife abuse was found to be 1.32 (95% CI: 1.03-1.69) times greater among those households in which both spouses or the husband primarily were the usual decision maker on what food to buy or prepare for family meals than among those households in which the wife primarily was the usual decision maker on such matters. There were not much previous studies specifically dealing on this matter. We note a pattern of male dominance, by himself or jointly with his wife, in decisions regarding what food to buy or prepare for family meals. A possible explanation is that most husbands are quite particular and even demanding as to what food is served to them. The result of this study lends empirical evidence for cultural norms supporting unequal power structures playing an important role in spousal violence. A basic premise underlying this association between wife abuse and the husband, by himself or jointly with his wife, being the usual decision maker on food matters is that when power was concentrated in the hands of the husband or male partner, common in that household. Selection bias was unlikely to have occurred in this study. The probability of selecting those households in which the respondent or his wife have the independent variables of interest and those households in which the respondent or his wife do not have the independent variables of interest are equal among those married men who gave in to wife abuse and among those men who did not abuse their wives. wife beating was more

101

The degree of bias resulting from misclassification of the category of the respondents according to exposure variable usual decision-maker on food matters is likely to be low. However, non-differential misclassification of the exposure variable attitude to better education and social well being of spouse and of the outcome status may be possible, resulting in a negative bias (bias towards the null). Increasing the awareness of men and women regarding their duties and responsibilities as well as their rights in a marital relationship may be of help in fostering equality between partners as opposed to male dominance and traditional gender roles which have been shown to promote spousal violence.

7.3.3.3 Association Between Wife Beating and Attitude towards better education and social well being of spouse

In this study, the odds of wife abuse was found to be 1.60 times (95% CI: 1.142.23) greater among those households in which the respondent would object to his partners wish to improve her education and to visit her friends alone than among those households in which the respondent would not have any objections to such. An important risk factor for domestic violence relates to social isolation of the woman preceding the onset of abuse (Neilsen, Russell & Ellington, 1992). It is also a strategy of possessive and controlling male partners to increase their power and dominance and it reduces womens ability to leave abusive relationships. In countries where violence is not regarded as a private matter, domestic violence is less common (Counts et al 1992). Our finding that the odds of wife abuse was greater among those households in which the respondent would object to his partners wish to improve her education lends empirical evidence for the premise that social isolation of the woman is indeed a risk factor for domestic violence. A basic supposition underlying this association between wife abuse and husbands objection to better education for his spouse is that the husband would not want his wife to obtain further education, given that this would provide her with economic independence and with greater social empowerment, i.e. social networks, self-confidence, or an ability to utilize sources of information and resources available in society (Jewkes et al, 1998).

102

Similarly, our finding that the odds of wife abuse was greater among those households in which the respondent would object to his spouse visiting her friends without him lends empirical evidence for the premise that social isolation of the woman is truly a risk factor for domestic violence. The basis for this association between wife abuse and husbands objection to visiting her friends without him is that the husband would not want his wife to have friends, given that this would provide her with persons to confide in about her relationship problems with her abusive husband (Jewkes et al, 1998) and from whom she could be given help to leave her abusive husband.

Selection bias was unlikely to have occurred in this study. The probability of selecting those households in which the respondent or his wife have the independent variables of interest and those households in which the respondent or his wife do not have the independent variables of interest are equal among those married men who gave in to wife abuse and among those men who did not abuse their wives. Misclassification of the category of the respondents according to attitude towards better education and social well being of spouse may be likely due to self-presentation bias. However, it would occur nondifferentially and therefore would result in a bias towards the null. Misclassification of outcome status may also be possible, resulting likewise in a negative bias (bias towards the null). Again, increasing the awareness of men and women regarding their duties and responsibilities as well as their rights in a marital relationship, including the right to develop their potentials through higher education and greater social dealings, may be of help in fostering equality between partners as opposed to male dominance and traditional gender roles which have been shown to promote spousal violence.

7.3.3.4 Association Between Wife Beating and Conflict resolution

In this study, after control for all other variables in the model, the odds of wife abuse was found to be 2.60 times (95% CI: 1.68-4.03) greater among those households in which the respondents get his way when there are disagreements than among those households in which the spouses would reach couple compromise. A common dynamic of conflicted intimate relationships is an inability of the couple

103

to communicate or negotiate in rational, non-judgmental ways. Male physical violence can serve to dominate, intimidate, control and silence the partner to gain the upper hand in a relationship. Physical violence may also be a strategy of first resort or last resort among men lacking verbal communication and problem solving skills. Our finding that when the husband usually get his way during disagreements he is more likely to beat/hit his wife, are in line with those reported elsewhere (Barling & Rosenbaum, 1986; Hotaling & Sugarman, 1990; Morrison et al, 1987; OLeary & Curley, 1986; Pan et al, 1994 and Rosenbaum & OLeary, 1981). In these previous studies, the Pearson correlation coefficient (r) ranged from 0.32 to 0.71, indicating a strong positive correlation between partner physical aggression and marital discord or conflict. These findings are not surprising, given that most partner aggression occurs in the context of a conflict or argument. Selection bias was unlikely to have occurred in this study. The probability of selecting those households in which the respondent or his wife have the independent variables of interest and those households in which the respondent or his wife do not have the independent variables of interest are equal among those married men who gave in to wife abuse and among those men who did not abuse their wives. The degree of bias resulting from misclassification of the category of the respondents according to exposure variables such as usual manner of conflict resolution is likely to be low. However, non-differential misclassification of the exposure variable attitude to better education and social well being of spouse and of the outcome status may be possible, resulting in a negative bias (bias towards the null). In this study, there were no identified effect measure modifiers. However, person primarily responsible for doing household tasks confounded the crude effect of this variable conflict resolution on wife abuse. Had this extraneous variable not been controlled in the analysis, the association between usual manner of conflict resolution and wife abuse would have been overestimated by 11%. Specifically, the estimated odds ratio would be positively biased away from the null (ORc=2.80 vs. Ora=2.52). Couples can be taught ways to enhance their communication skills so that they can communicate or negotiate in rational, non-judgmental and non-violent ways, allowing

104

them to reach a compromise when they have disagreements, which ordinarily can occur in any marriage.

7.3.3.5 Social Structural Factors Not Associated with Wife Beating

Literature dealing with the association between wife abuse and social structural factors was so scanty. One study (Hindin et al, 2002) explored decision-making within couples as an indicator of power relations between men and women. They proposed that deviations from the normative marital dynamic might result in more reports of intimate partner violence. In this study, given this subset of respondents, decision making, as to purchase and sale of assets or consumer goods, and as to family planning matters, was found not to be independently associated with domestic violence. Person primarily responsible for doing household tasks was also found not to be associated with wife abuse. Frequency of discussion, on expenses, on social-political issues, and on family planning matters, as indicator of general levels of communication, was also not independently associated with family violence. 7.2.4 Association Between Wife Beating and Social Location Factors

7.2.4.1. Association Between Wife Beating and Wifes Educational Attainment

In this study, after control for all other variables in the model, the odds of wife abuse was found to be greater in men whose wives had low educational attainment (OR=1.59, 95%CI:1.21-2.11 for those with wives who reached high school; OR=1.61, 95%CI:1.23-2.13 for those with wives who had no education or for those who had primary education), than in men whose wives had college education. The result of this study lends empirical evidence to the idea that violence towards women is about power and control. A basic premise underlying this association between wife abuse and wifes highest level of educational attainment is that education provides the woman with economic independence. Other authors have found that women who had higher education (McCall & Shields 1986, Straus et al 1980) or were economically independent were less likely to be abused (Rao 1997, Schuler, Hashemi, Riley & Akhter, 1996) (although Ellsberg et al(1999) did not find it). It seems likely that the mechanism of protection is not just through economic independence, as many other women with less

105

education were economically independent, but also a greater social empowerment (i.e. social networks, self-confidence, or an ability to utilize sources of information and resources available in society). One indicator of this is the lower personal subservience scores of more educated women. Ethnographic research suggests that women who are active in community fora may be relatively protected (Counts et al, 1992; Levinson 1989; Schuler et al 1997). In this study, the crude effect of wifes educational attainment on wife abuse was found to be confounded by the variable age at marriage. Had this extraneous variable not been controlled in the analysis, the association between wifes educational attainment and wife abuse would have been overestimated by close to 11%. Specifically, the estimated odds ratio would be positively biased away from the null (ORc=1.74 vs. Ora=1.57). Selection bias was unlikely to have occurred in this study. The probability of selecting those households in which the respondents wife have the low educational attainment and those households in which the respondents wife have high educational attainment are equal among those married men who gave in to wife abuse and among those men who did not abuse their wives. The degree of bias resulting from misclassification of the category of the respondents wife according to educational attainment is likely to be low. However, misclassification of the outcome status may be possible, leading to an underestimation of the measure of association.

Since possessing a higher education has been demonstrated to have a protective effect against domestic violence, programme planning aimed at decreasing the prevalence of wife abuse in the country, may be focused on a broad intersectoral agenda of providing women in general with access to college education.

7.2.4.2. Social Location Factors Not Associated with Wife Beating

Research in many different settings has found domestic violence to be associated with poverty (Martin et al 1999, Ellsberg et al 1999, Hoffman et al 1994, Allen & Straus 1980), male unemployment (Stark, Flitcraft, Zuckerman, Grey, Robinson & Frazier, 1981) and status differences between partners (Yllo & Bogard 1988). In this study a range of

106

social location factors were examined including respondents employment, wifes employment, per capita household monthly income and primary breadwinner in the family. None of these were found to be independently associated with domestic violence. A possible explanation is that the ever-married male respondents who formed part of the Reproductive Health Survey 2000 were quite homogeneous as regards their employment, their wifes employment, income and primary breadwinner status. Hence, the lack of sufficient variability between the exposed and unexposed groups could account for the finding of no significant association between these social location factors and wife abuse. 7.5. Summary of Findings

While this study was cross-sectional and so on its own cannot demonstrate causation, it adds to the growing body of evidence on intimate partner or domestic violence. In the data discussed in this study and the broader body of literature to which this is added, some important factors associated with wife abuse were made manifest. In summary, marital factor (age at marriage), social-psychological factor (history of parental abuse), some social structural variables (usual decision maker on food matters, attitude to better education and social well-being of spouse and usual manner of conflict resolution) and a social location factor (wifes educational attainment) contribute to increased risk for partner physical aggression. There was also interaction between age at marriage and history of parental abuse. Our findings underscore the multifactorial etiology of domestic violence. Future researches should consider additional factors, such as number of children and intraindividual factors (substance abuse/use, abusers personality and biological/neurophysiological disorders). They should also include an adequate period of follow-up and an assessment of the past experiences of both the women and the men involved. The findings presented on the different risk factors for wife abuse points to the need for varying interventions, not just medical, but legal (policies and laws) and sociocultural as well.

107

8
CONCLUSIONS AND RECOMMENDATIONS

8.1 CONCLUSIONS
The following conclusions can be drawn from the results of this study after identifying effect measure modifiers and controlling for confounders: 8.1.1. Men who married at a young age, 24 years old or younger, were more likely to beat their wife than men who married at an older age, 25 years old or older if they themselves have experienced parental abuse (OR=10.78, 95%CI: 4.21-27.59). 8.1.2. Men who had a history of violence in the family of origin, either from his father, his mother or both, were more likely to beat/hit their wife than men who had no such history. Among those who married at 24 years of age or less, the odds of wife beating/hitting for men with a history of having been physically abused by one or both parents was 13.08 times (95% CI=1.47-116.14) the odds of those without history of parental abuse. The odds ratio of 1.49 among those who were 25 years old or older when they were married was much lower but it was nevertheless significant (95% CI=1.20-1.84). 8.1.3. Although it was not statistically significant, there was positive association between wife abuse and person primarily responsible for handling the household budget (OR=1.47, 95% CI=0.95-2.13). 8.1.4. Men who are involved in decision-making, by themselves or jointly with their wives, on what food to buy or prepare for family meals are 1.32 times (95% CI: 1.031.69) more likely to beat/hit their wives than men who do not usually decide on food matters and allow their wives to do so. 8.1.5. Men who object to their spouse/partners improving their education or visiting their friends without him are 1.60 times (95% CI: 1.14-2.23) more likely to beat/hit their wives than men who do not object to such matters.

108

8.1.6. Men who usually get their way when there are disagreements are 2.60 times (95% CI: 1.68-4.03) more likely to beat/hit their wives than men who usually resolve such disagreements by couple compromise. 8.1.7. Men are more likely to beat/hit their wives when the wives had low educational attainment than when the wives had college education or higher. The odds of wife abuse was 1.59 times (95%CI:1.21-2.11) in men whose wives had reached high school while it was 1.61 times (95%CI:1.23-2.13) in men whose wives had no education or who had primary education only. 8.1.8. There were no significant associations between wife abuse and the marital factor duration of marriage. 8.1.9. There were no significant associations between wife abuse and the social structural variables person primarily responsible for doing household tasks, frequency of discussion on what to spend money for, on political-social issues and on family planning matters as well as the variables usual decision-maker on purchase or sale of assets and consumer goods and on family planning matters 8.1.10. Wife abuse was not significantly correlated with social location variables respondents educational attainment and per capita monthly income as well as with the occupation of either husband or wife and the person who is the primary breadwinner in the family.

8.2

RECOMMENDATIONS

8.2.1. Personnel trained in family violence issues must educate couples about domestic violence, its nature and effects, and the complex societal factors that prevent or facilitate disclosure, help seeking, and change in family systems. One possible venue by which practically all couples could be reached by this is during seminars given to them prior to their marriage (Pre-Canaan). 8.2.1.1. Men and women must educate themselves regarding their duties and responsibilities as well as their rights in a marital relationship, to maintain equality between partners as opposed to male dominance and traditional gender roles.

109

8.2.1.2.

Likewise, spouses must acquire communication skills, as spouse and parent, so that they can communicate or negotiate, with their spouse and children, in rational, non-judgmental and non-violent ways.

8.2.1.3.

Since it has been identified that history of parental abuse is a risk factor to domestic violence, couples must work at strengthening family ties and values, so as to reduce normative use of violence in conflict situations between couples, as well as between parents and children, on the excuse of disciplining them or correcting misbehaviour.

8.2.2. Programs should strive at providing women in general with access to higher education as this has been demonstrated to have a protective effect against domestic violence. Battered women in particular should have easy access to medical and legal assistance, as well as being given shelter when needed. Anti-Domestic Violence Bills could be proposed to policy-makers and the law-making body of the country to provide legal protection to prospective and current victims, not just of wife abuse but child abuse as well. 8.2.3. Strategies for prevention of violence against women should endeavour to

prevent and treat child abuse as it has been shown to be a risk factor, especially among those who marry at a young age. School authorities, perhaps through the guidance and counselling office, must be able identify victims of child abuse to ensure appropriate referral and treatment. Battered children in particular should have easy access to medical and legal assistance, as well as being given shelter when needed. Children who had been harshly punished or abused in other ways and children who had exposure to marital violence in the family of origin need to be subjected to counselling and even child psychiatric treatment for debriefing from their traumatic experience. 8.3.4. Programs for prevention of violence against women should focus on

counselling adolescents as to the drawback of getting married at a young age, particularly those with history of parental abuse given that they have been identified as a high-risk group in this study. This can be incorporated in the school curriculum. Efforts to reach out to out-of-school youths during information, education, and communication campaigns should be made, through the leaders in the community and with the help of government and non-government organizations.

110

9
BIBLIOGRAPHY 1. Banatin CA. (1998) Domestic violence. Proceedings from the 37th Annual Convention of the Philippine Academy of Family Physicians, 55-58. 2. Kitchener-Waterloo Multicultural Center (1998). Understanding violence against women. Retrieved November 20, 1998 from the World Wide Web: http://www.kwmc.on.ca/women.html 3. Pineda-Mercado S. (1997). Violence against women and children. The Filipino Family Physician, 35 (4): 155-159. 4. University of the Philippines-Philippine General Hospital. (1997). Womens Desk Training Workshop on Violence Against Women Part 1. 5. Bascos-Deveza T. (2001). Violence against women and children: methods of measurement using police statistics and a national survey. Proceedings from the 8th National Convention on Statistics, 15-17 to 15-28. 6. Alpert EJ, Cohen S, Sege RD. (1997) Family violence: an overview. Academic Medicine, 72 (1) S3-S6. 7. Womens Health. (2002). Assessing risks of intimate partner violence in primary care setting. Retrieved January 9, 2002 from the World Wide Web: http://www.medscape/womens health.html 8. Population Reports. (1999). Ending violence against women. Volume XXVII, Number 4. 9. Straus MA , Gelles RJ. (Eds.) (1990). Physical violence in American families: Risk factors and adaptations to violence in 8,145 families. New Brunswick, NJ: Transaction. 10. Edleson JL. (1995). Do batterers program work? In International study group on the future of intervention with battered women and their families, (pp. 1-10), Haifa, Israel. 11. Stark E, Flitcraft A. (1988). Violence Against Intimates: an epidemiological review. In V.B. Van Hasselt, R. L. Morrison, A.S. Bellack, & M. Hersen (Eds.), Handbook of family violence (pp. 293-317). New York. Plenum. 12. Miller TR, Coehn MA, & Wiersema B. (1996). Victim costs and consequences: a new look (Research Report No. NCJ 155282)). Washington, DC: National Institute of Justice. 13. Meyer H. (1992). The billion-dollar epidemic. American Medical News, 35 (1),7. 14. Ontario Womens Directorate, Violence Against Women, Services Elgin County (1992). Fact Sheets: Wife assault: the impact on children. Retrieved November 20, 1998 from the World Wide Web: http://www.vawsec.on.ca/wifechil.html; Fact Sheets:

111

Power and control equality wheel. Retrieved November 20, 1998 from the World Wide Web: http://www.vawsec.on.ca/whltxt.html. 15. Gelles RJ, Straus. (1978). Violence in the American family. Journal of Social Issues, 35 (2). 15-39. 16. Hamberger LK, Hastings J. (1986). Personality correlates of men who abuse their partners: A cross-validation study. Journal of Family Violence, 1(4), 323-341. 17. Roy, M. (Ed.). (1977). Battered women: A psychosociological study of domestic violence. New York: Van Nostrand Reinhold. 18. National Commission on the Role of Filipino Women. (2001). Info resource- womens Resource Directory Assessing risks of intimate partner violence in primary care setting. Retrieved March 21, 2002 from the World Wide Web: http://www.medscape/womens health.html 19. United Nations Foundation (UN) Issues Women and Population. (2002).Women and Population: The Role of the UN. Retrieved March 21, 2002 from the World Wide Web: http://www.unfoundation.org/programs/women/role.html 20. Holzworth-Munroe A, Stuart GL. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116 (3), 476-497. 21. Dutton DG. (1994). The origin and structure of the abusive personality. Journal of Personality Disorders, 8(3), 181-191. 22. Kaufman Kantor G, Jasinski J, Aldarondo E. (1994). Sociocultural status and incidence of marital violence in Hispanic families. Violence and Victims, 9(3), 207-222. 23. Straus MA. (1973). A general systems theory approach to a theory of violence between family members . Social Science Information, 12, 105. 24. Smith C. (1990). Status discrepancies and husband-to-wife violence. Paper presented at the Meeting of Eastern Sociological Society. 25. Kalmuss D. (1984). The intergenerational transmission of marital aggression. Journal of Marriage and the Family. 48, 113-120. 26. OLeary KD. (1988). Physical aggression between spouses: A social learning theory perspective. In V.B. Van Hasselt, R.L. Morrison, A.S. Bellack, & M. Hersen (Eds.) Handbook of family violence. (pp. 31-56). New York: Plenum. 27. Straus MA, Gelles RJ, Steinmetz S. (1980). Behind closed doors: Violence in the American family.Garden City, NJ: Anchor. 28. Kaufman Kantor G, Jasinski JL. (1998). Dynamics and Risk Factors for Partner Violence. In J.L. Jasinski & L.M. Williams (Eds.) Partner violence. A comprehensive review of 20 years of research. Thousand Oaks, CA: Sage. 29. Hotaling GT, Sugarman DB. (1986). An analysis of the risk markers in husband to wife violence: The current state of knowledge. Violence and Victims., 1(2), 101-124.

112

30. Bowker JH. (1983). Beating wife beating. Lexington, MA: Lexington. 31. Fagan JA, Stewart DK, Hansen KV. (1983). Violent men or violent husbands? Background factors and situational correlates. In D. Finkelhor, R.J. Gelles, G.T. Hotaling, M.A. Straus (Eds.), The dark side of families (pp. 49-68). Beverly Hills, CA: Sage. 32. Walker LE. (1984). The battered woman syndrome. New York: Springer. 33. Stith SM, Farley SC. (1993). A predictive model of male spousal violence. Journal of Family Violence, 8 (2), 183-201. 34. Alexander PC, Moore S, Alexander ER III. (1991). What is transmitted in the intergenerational transmission of violence? Journal of Marriage and the Family., 53, 657-668. 35. Langhinrichsen-Rohling J, Neidig P, Thorn G.(1995). Violent marriages: gender differences in levels of current violence and past abuse. Journal of Family Violence. 10 (2), 159-176. 36. Simons RL, Johnson C, Beaman J, Conger RD. (1993).Explaining womens double jeopardy: Factors that mediate the association between harsh treatment as a child and violence by a husband. Journal of Marriage and the Family, 55, 713-723. 37. Shupe A, Stacey WA, Hazlewood LR, (1987). Violent men, violent couples: The dynamics of domestic violence. Lexington, MA: Lexington. 38. Heyman RE, OLeary KD, Jouriles EN. (1995). Alcohol and aggressive personality styles: Potentiators of serious physical aggression against wives? Journal of Family Psychology, 9 (1), 44-57. 39. Dutton DG, Starzomski AJ. (1993). Borderline personality in perpetrators of psychological and physical abuse. Violence and Victims, 8 (4), 327-337. 40. Kitchener Waterloo Multicultural Centre (1998). Profile of a batterer. Retrieved November 20, 1998 from the World Wide Web: http://www.kwmc.on.ca/batterer.html 41. Fagan J. (1990). Intoxication and aggression. In M. Tonry & J.Q. Wilson (Eds.). Drugs and crime (pp. 241-320). Chicago: University of Chicago Press. Hudson WW, McIntosh SR. (1981). The assessment of spousal abuse: Two quantifiable dimensions. Journal of Marriage and the Family, 43, 873-885. 42. Aldarondo E, Sugarman DB.(1996). Risk marker analysis of the cessation and persistence of wife assault. Journal of Consulting and clinical Psychology, 64 (5), 1010-1019. 43. Kaufman Kantor G, Straus MA. (1987). The drunken bum theory of wife beating. Asocial problems, 34 (3), 213-230. 44. Jewkes R, Levin J, Penn-Kekana L. Risk factors for domestic violence: findings from a South African cross-sectional study. Soc Sci Med 2002; 55 (9) :1603-17.

113

45. Kyriacou DN, Anglin D, Taliaferro E, Stone S, Tubb T, Linden JA, Muelleman R, Barton E, Kraus JF. Risk Factors for Injury to Women from Domestic Violence. New England Journal of Medicine. 1999; 25 (341): 1892-1898. 46. Pihl RO, Smith M, Farrell B. (1983). Alcohol and aggression in men: A comparison of brewed and distilled beverages. Journal of Studies on Alcohol, 278-282. 47. Yllo K, Straus MA, (1990). Patriarchy and violence against wives: The impact of structural and normative factors. In M.A. Straus & R.J. Gelles (Eds.) Physical violence in American families: Risk factors and adaptations to violence in 8, 145 families (pp. 383-399). New Brunswick, NJ: Transaction. 48. Hornung CA, McCullough BC, Sugimoto T. (1981). Status relationships in marriage: Risk factors in spousal abuse. Journal of Marriage and the Family, 43, 675-692. 49. Kaufman Kantor G, Asdigian NL, (1996). When women are under the influence: Does drinking or drug use by women provoke beatings by men? In M. Galanter (Ed.), Recent developments in alcoholism. New York: Plenum. 50. Ramiro LS, Amarillo LE, Madrid BJ. Domestic violence in urban Filipino families (A WorldSAFE Multicenter Study on Family Violence). 1998. (unpublished). 51. Kaufman Kantor G. (1990). Ethnicity, alcohol and family violence: A structural and cultural interpretation. Paper presented at the Forty-Second Annual Meeting of the american soociety of Criminology, Baltimore, MD. 52. Martin SL, Tsui AO, Maitra K, Marinshaw R. Domestic Violence in Northern India. American Journal of Epidemiology. 1999, 150 (4): 417-426. 53. Rollins BC, Oheneba-Sakyi Y (1990). Physical violence in Utah households. Journal of Family Violence, 5 (4), 301-309. 54. McLaughlin IG, Leonard KE, Senchak M.(1992). Prevalence and distribution of premarital aggression among couples applying for a marriage license. Journal of Family Violence, 7 (4), 309-319. 55. Conger RD, Elder GH, Lorenz KJ, Conger RL, Simons RL, Whitbeck LB, Huck S & Melby JN. (1990). Linking economic hardship to marital quality and instability. Journal of Marriage and the Family, 52(3), 243-656. 56. Sugarman DB, Aldarondo E, Boney-McCoy S. (1996). Risk marker analysis of husband-to-wife violence: A continuum of aggression. Journal of Applied Social Psychology, 26 (4), 313-337. 57. Widom CS. (1993). Child abuse and alcohol use and abuse. In S.E. Martin (Ed.) Alcohol and interpersonal violence: Fostering multidisciplinary perspectives (NIH Research Monograph No. 24, pp. 291-314) Rockville, MD: U.S. Department of Health and Human Services. 57.Whitbeck LB, Hoyt DR, Simons RL, Conger RD, Elder GH. (1992). Intergenerational continuity of parental rejection and depressed affect. Journal of Personality and Social

114

Psychology, 63 (6), 1036-1045. 58. Bushman BJ, Cooper HM. (1990). Effects of alcohol on human aggression: an integrative research review. Psychological Bulletin, 107 (3), 341-354. 59. Downs WR, Miller BA, Panek DD. (1993). Differential patterns of partner-to-woman violence: A comparison of samples of community, alcohol abusing, and battered women. Journal of Family Violence, 8 (2), 113-135. 60. Fagan J, Browne A. (1994). Violence between spouses and intimates: Physical aggression between women and men in intimate relationships. In A.J. Reiss & J.A. Roth (Eds.), Understanding and preventing violence (Vol 3, pp. 115-292). Washington, DC: National Research council, National Academy of Sciences. 61. Khalidi A. Domestic violence among some Palestinian refugee communities in Lebanon: An exploratory study and ideas for further action. A report submitted to Najdeh Association, 2000

115

Você também pode gostar