Introduction

A World Health Organization study in ten countries across five continents revealed the pervasiveness of sexual violence within intimate partner relationships is a serious problem (Garcia-Moreno, Jansen, Ellsberg, Heise, & Watts, 2006). Sexual violence by a male intimate partner tends to be more frequent (Hall, 1995; Russell, 1990) and more severe (Stermac, Del Bove, & Addison, 2001) compared to stranger perpetrated sexual violence. Despite evidence of its relatively high prevalence in numerous countries, its frequency and severity against female intimate partners is widely not believed to exist (Kirkwood & Cecil, 2001; Whatley, 2005) or still perceived as a lesser crime than that perpetrated by others (Monson, Byrd, & Langhinrichsen-Rohling, 1996).

Research indicates that sexual violence against female intimate partners often has severe and long-lasting consequences for women, such as injuries, fatigue, and vomiting (Adams, 1993; Bergen, 1996), pelvic pain, urinary tract infections, miscarriages, stillbirths, bladder infections, infertility, and the potential contraction of sexually transmitted diseases (Weiss et al., 2008), including HIV/AIDS (Silverman, Decker, Saggurti, Balaiah, & Raj, 2008). A study in India found that unplanned pregnancies were significantly more common among wives of abusive men, especially sexually abusive men who used force (Martin et al., 1999).

Short-term mental health effects of marital rape include anxiety, shock, intense fear, depression, suicidal ideation, sleep, and post-traumatic stress disorder (Campbell, Keegan, Cybulska, & Forster, 2007; Ellsberg, Jansen, Heise, Watts, & García-Moreno, 2008). Long-term effects often include eating and sleep disorders, depression, problems establishing trusting relationships, distorted body image, and increased negative feelings about self (Bergen, 1996; Ullman & Siegel, 1993).

The literature on factors associated with sexual violence by intimate male partners is often inconclusive. Further, some determinants remain unexamined in many developing countries, and given wide cultural variation in marital relationships and social beliefs, it may not be possible to extrapolate from much of the research in Western countries. Thus, for example, Russell (1990) found no age pattern among survivors of intimate partner sexual violence, while others found younger age increased the risk of such violence (Fawzi et al., 2005). Russell (1990) found that women in the upper middle class were slightly over-represented among marital rape survivors, while others (Fawzi et al., 2005; Santhya, Haberland, Ram, Sinha, & Mohanty, 2007) found the opposite.

In Egypt, low level of husband’s education was found to be associated with sexual abuse of the wife (Elnashar et al., 2007). In the U.S. and China, low level of education of the woman was linked to this violence (Coker, Derrick, Lumpkin, Aldrich, & Oldendick, 2000; Yimin et al., 2001). However, Gage and Hutchinson (2006) found that Haitian women with primary education were more likely to be sexually abused by intimate partners compared to women who had no education.

Some factors, such as spousal communication (Gordis, Margolin, & Vickerman, 2005) or power and control of the male partner (Gage & Hutchinson, 2006) have been identified as associated with sexual violence by an intimate partner were not put to test in other studies.

Diverse research results on factors associated with sexual violence in different settings is not surprising given the differences in the methods applied. Although it is important for conceptual clarity for program policies to identify common factors contributing to sexual intimate partner violence across countries, it is also important to recognize that individual context also contributes to violence, making it essential to investigate factors associated with violence in specific contexts.

The literature on Bangladesh mainly focuses on physically forced sex and report prevalence rates ranging between 18 and 26 % for lifetime and 11 and 15 % for current sexual violence (Bangladesh Demographic and Health Survey [BDHS], 2009; Hindin, Kishor, & Ansara, 2008). The only quantitative study on factors associated with sexual violence by husbands in Bangladesh reported that young, illiterate, and poor women were more likely to experience this violence while women participating in micro-credit programs or contributing to household income were less likely to be exposed to it (Hadi, 2000). This article explored the magnitude and nature of within marriage sexual violence against reproductive aged women (15–49) in rural and urban Bangladesh and the factors associated with physically forced sex (Fig. 1).

Fig. 1
figure 1

Prevalence of sexual violence within marriage and its overlap with other forms of violence in Bangladesh, %

Full size image

Method

Participants

The data for this analysis come from the Bangladesh component of the WHO multi-country study on Women’s Health and Domestic Violence against Women. A total of 4,051 households (2,105 in the urban and 1,946 in the rural area) were included in the sample (see Garcia-Moreno et al., 2006; Naved, Azim, Bhuiya, & Persson, 2006). About 10 % of the households in both areas were either empty or destroyed. From the remaining, about 6 % in the urban area and about 1 % in the rural area refused to participate in the study. No eligible women (15–49 years) were found in 6 % of households in the city and in 8 % of households in the rural area. Among the eligible women, the response rate was 96 % for both urban and rural sites. The final sample consisted of 3,130 reproductive age women. From this sample, we selected the ever-married women for the current analysis. Thus, the sub-sample used in this analysis consisted of 2,702 ever-married women.

Table 1 shows the main socioeconomic characteristics of the study participants. The women were aged around 30 years, generally had low levels of education, and had low participation in income generating activities. The sample was predominantly Muslim, with a higher proportion of Hindu women in the rural area (16 % against 5 % in the urban area). Dowry was demanded in the majority of rural marriages (53 %), while it was demanded in 14 % of urban marriages. Score for conventional attitude towards gender role and controlling behavior of the husband were higher in the rural than urban area.

Table 1 Descriptive statistics of variables included in the multilevel analysis
Full size table

Procedure

A cross-sectional population-based survey was administered to Bangladeshi women aged 15–49 years between June 25 and November 25, 2001. The data were collected from one large urban city and one rural area. The urban and rural sites were located in the eastern part of the country. In the rural area, over 95 % are Muslims (International Center for Diarrheal Diseases Research, Bangladesh [ICDDR-B], 1998). The level of education is quite low with a high proportion of adults having no schooling (ICDDR-B, 1998). Most families are poor, though contrasts in socioeconomic status are much greater in the urban area. The urban area is a mega-city with an extremely high population density and in-migration rates.

A multi-stage sampling scheme was used with clusters being the primary sampling unit (i.e., mohollas or the smallest administrative units in the urban site and villages in the rural area). A multi-stage procedure was used to randomly select the clusters, first from within the selected urban and rural areas, and then from households within each cluster using the probability proportionate to size method. Forty-two clusters in the rural area and 39 clusters from the urban area were selected for the survey.

In the rural area, the household list was obtained from the ICDDR-B database, which is updated every month. In the urban area, the number of households in each cluster was obtained from Population Census, 1991, Community Series, published in April 1993 by the Bangladesh Bureau of Statistics. The number of households selected in each cluster was proportionate to size.

In the rural site, the households were randomly selected from the ICDDR-B database. In the urban site, on the other hand, every sixth household in a cluster starting from the northeast entry point was included in the sample. The first household in the cluster was selected randomly from the first six right-hand side households upon entry into the cluster from northeast. Households were defined as a unit consisting of people sharing the same stove.

The age and initials of all females in a selected household were recorded on a household form. From this list, women eligible for interviewing (i.e., aged 15–49 years) were identified. For the safety of the participant and the interviewer, only one woman in each household was interviewed. In situations, in which there was more than one eligible participant, one woman was randomly selected from the list for the interview. Cases in which the woman selected was not available, the interviewer made an appointment to return to conduct the interview. Two additional visits were made before the woman was considered unavailable.

Measures

A modified version of the Conflict Tactics Scales (CTS2) was used to measure sexual violence (Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Psychometric analysis of these behaviorally-specific questions showed good internal consistency among the items, indicating that the instrument provided a reliable and valid measure of sexual violence (Garcia-Moreno et al., 2006). Specific items regarding sexual violence perpetrated by the husband were as follows: (1) Physically forced you to have sexual intercourse when you did not want to? (2) Did you ever have sexual intercourse when you did not want because you were afraid of what he might do? and (3) Did he ever force you to do something sexual that you found degrading or humiliating? Frequency of each act of violence experienced was collected for the last 12 months and before the last 12 months. The response categories were: once, several times (2–4 times), and many times (>4 times). The survey also included as well a modified version of CTS2 for measuring physical and emotional violence (Straus et al., 1996).

All the women in the survey who ever had sex were asked to describe whether their first sexual experience (regardless of who the partner was) was wanted or unwanted but happened without being physically forced, or was physically forced. From response to this question, a dummy variable was constructed to reflect whether or not the first sexual experience was coerced or forced.

Age of the woman was used as a categorical variable with age below 30 as the reference category. A dummy variable was used for reflecting whether or not she earned an income. Husband’s education was used as a categorical variable with education beyond 10th grade as the reference category.

An indicator of cluster-level patriarchal attitude of the woman toward gender roles was constructed aggregating women’s responses where they were asked whether they agree or disagree with the following five statements: (1) A good wife obeys her husband even if she disagrees; (2) It is important for a man to show his wife/partner who is the boss; (3) A woman should be able to choose her own friends even if her husband disapproves; (4) It is a wife’s obligation to have sex with her husband even if she is not in the mood; (5) If a man mistreats his wife, others outside of the family should intervene. The response categories were: agree, disagree, and don’t know. A response reflecting conventional attitude toward gender roles was coded as 1 and the reverse as 0 and the “don’t know” category was treated as missing. The questions were designed to tap two constructs that had emerged in the literature (Levinson, 1989) as predictive of communities with greater wife beating: (1) Settings where men were dominant in the family; and (2) Settings where there were norms of non-intervention by others in the family.

At the household level, variables that served as proxies for patriarchal attitude were: (1) controlling behavior of the husband; (2) whether the husband’s mother was physically abused by his father; and (3) whether dowry was demanded at marriage. Questions used for measuring controlling behavior were: (1) whether the husband commonly attempted to restrict the women’s contact with her native family; (2) whether the husband commonly attempted to restrict the women’s contact with friends; (3) whether he insisted on knowing where she was all the time; (4) whether he ignored her or treated her indifferently; (5) whether he expected her to seek permission for seeking health care for herself; (6) whether he constantly was suspicious that she was unfaithful; and (7) whether he got angry if she spoke with another man. The answers contained “yes” or “no” options. For scoring purposes, a positive answer was scored 1 and a negative one was scored as 0. Scores for the seven items were summed. If response of a woman to none of these questions was positive, the woman was considered as not experiencing controlling behavior. If response to one or more of these questions was positive, the woman was considered exposed to controlling behavior and the degree of control was assessed by the total score obtained from responses to these seven questions. A dummy variable was constructed for husband’s mother’s exposure to physical violence perpetrated by his father for use in the regression model.

The index representing the extent of communication between spouses was developed from “yes” or “no” questions presented to the woman about whether the couple discussed the daily experiences of the participant or her husband and whether they discussed worries of the participant or her husband. For each item, a “yes” answer was assigned a score of 1 and “no” with a 0. A summed score was created for communication, which ranged between 0 and 4, with 4 indicating the highest level of communication. Households were categorized into quartiles according to the size of annual household income.

The number and percentage of women who experienced sexual violence by the husband were calculated for lifetime and past 12 months, including frequency and overlaps with other forms of violence perpetrated by the husband.

Statistical Analysis

A multivariate logistic regression model was used for identifying factors associated with physically forced sex by husbands during the past 12 months. The independent variables were: age of the woman, whether or not she earned an income, an index of her attitude towards conventional gender roles, forced initiation of sex, husband’s education, number of acts of control the husband imposed on the woman, level of spousal communication, dowry demand at marriage, husband’s mother physically assaulted by his father, and annual household income. Data analysis was carried out using the software Stata.

The survey team received an intensive 16-day training. The training modules included gender issues, ways of handling sensitive issues and management of difficult situations arising out of discussion of these issues, confidentiality, and safety and protection of women participating in the study.

For the sake of confidentiality, the survey was introduced at the local and household level as a survey of women’s health and life experiences. The face-to-face interviews were conducted in private and in a non-judgmental manner. The study was approved by the Ethical Review Committee of the research institution in Dhaka, Bangladesh.

Results

Prevalence of within Marriage Sexual Violence against Women

Prevalence of sexual violence by husband is shown in Table 2. Overall, 37 % of urban and 50 % of rural ever-married reproductive aged women reported ever being sexual abused by the husband. Physically forced sex was most commonly reported (34 % in the urban area; 46 % in the rural area). Sex out of fear was reported by a substantial proportion of the women (21 % in the urban area; 32 % in the rural area). A low proportion of women (3–4 %) reported ever being forced to do something sexual that was degrading or humiliating. About 16 % of the urban and 18 % of the rural women reported sexual abuse by the husband during the preceding 12 months. An overwhelming majority (96 % in case of physically forced sex; 95–96 % in case of sex out of fear; and 89–90 % in case of unwanted sexual act) of the sexually abused women in both sites reported being abused more than once. There was no rural–urban difference or difference by reference period in the frequency of abuse.

Table 2 Prevalence and frequency of sexual violence against wives in Bangladesh, %
Full size table

About 75 % of the sexually abused women experienced other forms of violence by husbands (83 % in urban; 69 % in rural). About 12 % of the sexually abused urban women experienced both sexual and physical violence; 20 % experienced sexual and emotional violence; and 51 % experienced all the three forms of violence. Overlap between sexual and physical violence (22 %) was much greater for rural women compared to urban women. Sexual and emotional violence was experienced by 9 % of rural women and all the forms of violence by 38 % of them. In the rural area, a greater proportion of women reported sexual violence (only 16 %) compared to the women in the urban area (6 %).

Factors Associated with Physically Forced Sex Perpetrated by Husbands

Results of logistic regression analysis (Table 3) showed that, in both the urban and rural areas, the common factors positively associated with physically forced sex by husband were: history of physical abuse of husband’s mother by his father; controlling behavior of the husband; and forced first sex. Thus, history of physical abuse of husband’s mother by his father increased the likelihood of sexual violence 2.13 times in the urban area and 1.58 times in the rural area. With increase in number of different acts of controlling behavior imposed on the woman by her husband the likelihood of sexual violence increased 1.61 times in the urban area and 1.20 times in the rural area. If a woman had forced first sex, her risk of experiencing forced sex by the husband in the past 12 months was 5.08 times higher in the urban and 2.25 times higher in the rural area compared to the women who did not experience forced initiation of sex. Interestingly, urban women belonging to second and third income quartiles were more likely to be exposed to physically forced sex (2.50 and 1.92, respectively). In the rural area, women aged 25–29 were more likely to experience this violence compared to women aged 15–19. Dowry demand at marriage increased the likelihood of rural women’s exposure to this violence by 1.91 times. Husbands with secondary education were less likely to engage in physically forced sex compared to men having no education in the urban area.

Table 3 Determinants of self-reported physically forced sexual violence in past 12 months against wives in Bangladesh
Full size table

Discussion

This study presents the most in-depth epidemiological survey of violence against women in marriage in Bangladesh. The rates of physically forced sex reported in this study (37 % in the urban and 50 % in the rural site) were much higher (three times in the urban area and two times in the rural area) compared to those reported by Bangladesh Demographic and Health Survey 2007 (BDHS, 2009). These differences are more likely to be attributable to methodological differences between the two studies. The current study was entirely devoted to violence against women while the BDHS addressed it as one of the multiple issues it deals with. Elaborate staff training, careful sequencing of violence related questions for avoiding participant fatigue, use of behaviorally explicit multiple questions on sexual violence, fieldwork approaches in creating a supportive environment for disclosure of violence, and making psychological support available for field workers in their very difficult role in the current study might have contributed to greater reporting. As shown in the literature, some of these elements do yield higher reporting (Ellsberg, Heise, Pena, Agurto, & Winkvist, 2001; Fisher, 2009).

Sex out of fear and forced involvement in degrading sexual acts were explored in this study for the first time in Bangladesh. The prevalence of sex out of fear was quite high (21 and 32 %) while degrading sexual acts were reported only by 3–4 % of the women. Prevalence of sex out of fear found in this study was comparable to the rate found in Thailand, while it was much lower than the reported rate in Ethiopia and much higher than in Namibia or Brazil (García-Moreno, Jansen, Ellsberg, Heise, & Watts, 2005). Low reporting of forced engagement in degrading sexual act was common in other countries as well.

Similar to countries such as Peru, Namibia, Samoa, and Tanzania, Bangladesh reported physically forced sex to be the most commonly reported item of sexual violence. This pattern differed in the case of Ethiopia and Thailand, where sex out of fear was most frequently reported. It is not possible to discern from this study whether these differences in the pattern reflect differences in actual behavior or any reporting bias.

Sexual violence was found to be higher in the rural area, which is consistent with findings from the nationally representative BDHS (2009) and with other countries (Garcia-Moreno et al., 2006). Similar to other countries (García-Moreno et al., 2005), sexual violence by the husband in Bangladesh was a pattern of recurrent behavior rather than a one-time event and it tended to overlap with other forms of violence (Garcia-Moreno et al., 2006).

Overlap among different forms of violence found in this study was expected as violence is used as a tool for controlling a woman (Dobash & Dobash, 1998; Dutton & Golant, 1995; Lloyd, 1999; Smith, 1990) and this tool is used in both different spheres and situations for controlling various aspects of the life of a woman. Sexual violence is linked to other forms of violence by numerous threads. A study by Parish, Wang, Laumann, Pan, and Luo (2004) clearly demonstrated that severe hitting was a significant risk factor for unwanted sex. Overlap between sexual and physical violence in Bangladesh (between 12 and 22 %) is of concern as this particular combination has been identified to be associated with HIV infection in neighboring India (Silverman et al., 2008).

In regards to predictors of sexual violence, this study clearly demonstrated the positive association between number of acts of control of a wife and sexual violence. In the given sociocultural context of high control over women’s sexuality and sanctions against sex out of wedlock, it can be assumed that use of force by a partner during first sexual encounter in most cases occurred within marriage. As violence is a pattern of behavior, continuation of sexual violence in marriage seems logical. Koenig et al. (2004) showed that first coerced sex was related to a number of negative reproductive health outcomes, including unintended pregnancy. This study expanded this finding showing that coerced or forced first sex was a predictor of sexual violence against women in marriage.

The finding that history of physical abuse of husband’s mother by his father increased the likelihood of sexual violence in both the sites lends support to the literature on intergenerational transmission of violence theory (Ellsberg, Peña, Herrera, Liljestrand, & Winkvist, 2000; Garcia-Moreno et al., 2006; Koenig, Stephenson, Ahmed, Jejeebhoy, & Campbell, 2006) that violence is, in fact, a learned behavior.

Increased likelihood of exposure to this violence by rural women aged 20–24 compared to women aged 15–19 was counterintuitive as younger age is widely known to be associated with violence against women (García-Moreno et al., 2005). Further research is needed to explore factors underlying this. The finding that urban wives from certain higher income groups had increased odds of being raped challenges the widespread notion that any form of violence against women is clustered in economically disadvantaged groups.

Some possible limitations of this study need to be mentioned. Being a cross-sectional survey, causal inference cannot be drawn from these study findings with regard to determinants of sexual violence. Also, strictly speaking, the findings cannot be generalized beyond the study areas; however, there is no reason to believe that these study areas are particularly unique or have different patterns from what might be expected in other areas of Bangladesh. As a study based on self-report, recall might be an issue.

Feminists suggest that patriarchal attitudes explain wife abuse (Dobash & Dobash, 1992; Haj-Yahia, 2005; Santana, Raj, Decker, La Marche, & Silverman, 2006). As demonstrated by Naved and Persson (2010), dowry demanded at marriage is indicative of patriarchal attitude of the groom’s household and is highly predictive of physical wife abuse in Bangladesh. In line with this were findings that dowry demand at marriage in rural Bangladesh was associated with sexual violence by the husband. Conventional attitude of the women in studied clusters towards gender role, however, was not associated with physically forced sex against wives.

This study highlights the importance of adopting a multidimensional approach in measuring sexual violence for better capturing the extent of sexual violence. It also indicates the need for programs addressing this violence having huge public health implications. Finkelhor and Yllo (1985) convincingly argued that the marriage license was historically perceived as a “license to rape.” There is no law in Bangladesh criminalizing sexual violence within marriage. Long-standing traditions that have failed to recognize wife sexual abuse as a problem reflects cultural beliefs about masculinity, femininity, and sexuality that have interfered with the acknowledgment of and societal response to wife sexual abuse. Thus, legislation against sexual wife abuse should be passed in Bangladesh with proper provision for enabling implementation of the law. The impetus for controlling women originates from particular notions of masculinity. Importance of husband’s characteristics as predictors of sexual violence indicates the need for targeting and working with men for preventing this violence. This study has identified that control of women by husbands and children witnessing parental violence as a child perpetuate sexual abuse and they should be primary foci of prevention programs.