Prevalence and Correlates of Family Violence Worldwide

This paper describes the International Dating Violence (IDV) Study, which is a multinational study of violence against dating partners by university students. It presents preliminary data on the prevalence of violence against dating partners, and on violence by parents against children, as reported by students in 31 universities in 16 countries, five in Asia and the Middle East, two in Australia-New Zealand, five in Europe, two in Latin America, two in North America). The past year rates for perpetration of violence against dating partner were: Physical assault -median = 29%, range 17% to 45%. Injury -median = 7%, range 2% to 20%. Sexual coercion -median = 24%, range 6% to 41%. Physically forced sex -median = 2%, range 0% to 12%. The paper also provides separate rates for males and females, and for severe levels of assault and injury. At the median university, 57% students were “hit a lot” by their parents when they were less than 12 years old was (range = 13% to 70%). During the teenage years, at the median university, 26% were hit a lot was (range = 12% to 56%). Correlation analysis found that the higher the percent of students at a university who were hit a lot by parents, the higher the percent who assaulted and injured a dating partner, and the higher the percent who coerced sex. The methodological results show that the measures of physical violence and sexual coercion have acceptable to high alpha coefficients of reliability, and that there is evidence indicating construct validity. * Paper presented at the International Seminar On Family Violence, University of Montreal, Montreal, Canada, 11-13 June 2003. For other publications on this or related issues, log into the Family Research Laboratory website www.unh.edu/frl. Financial support was provided by National Institute of Mental Health grant T32MH15161 and by the University of New Hampshire. I would like to express my deep appreciation to the members of the International Dating Violence Research Consortium listed in the appendix table. The study was only possible because of their contributions. They provided important advice and suggestions, translated and administered the questionnaire, and permitted use their data for this paper. Even with this data and assistance, the project would not have been possible without the dedicated and skilled contribution of the project staff at the University of New Hampshire. These include Doctors Emily M. Douglas, Patricia C Ellerson, and Ignacio Luis Ramirez (NIMH funded Post-Doctoral Research Fellows) and Rose A. Medeiros, and Sarah A. Savage (Graduate Research Assistants) \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 1 The rates of physical and sexual assault against dating partners by students reveal both important differences and similarities between universities. Perhaps the most important similarity is the high rate of assault by both male and female students in all the universities. Even the universities that had lower rates relative to other universities had a high rate of physical assault and sexual coercion of dating partners. These preliminary results indicate that the International Dating Violence Study has been successful in securing high rates of disclosure, and that these measures are reliable and valid. This paves the way for the main purpose of the study, which is to test theories to help explain the large differences between universities in the rate of violence against dating partners. * * * * * * * * * * This paper has three objectives. First, it describes a multi-nation investigation of violence in the dating relationships of university students called the International Dating Violence Study. Second, it presents preliminary results describing the prevalence of physical and sexual assault against dating partners by students at 31 universities in 16 countries (five in Asia and the Middle East, two in Australia-New Zealand, five in Europe, two in Latin America, two in North America). These are the first sites of a planned 30-nation study. Third, it presents preliminary methodological information on the reliability and validity for cross-national research of the scales used to measure violence against dating partners. VIOLENCE IN DATING RELATIONSHIPS It is now widely recognized that relationships between partners in marital, cohabiting, and dating relationships are often violent (Barnett, Miller-Perrin, and Perrin 1997; Gelles and Straus 1988). However, it is not widely recognized that dating couples are even more likely to be violent than married couples, despite the fact that the higher rate has been demonstrated by over 50 studies, starting in the 1980’s (Stets and Straus 1989; Sugarman and Hotaling 1989). For purposes of primary-prevention (Cowen 1978; O'Leary and Sweet Jemmott 1995), it is vital to increase understanding of the etiology of dating-couple violence because the behavior at that point in the life-cycle can establish patterns that persist over a lifetime (O'Leary, Barling, Arias, Rosenbaum, Malone, and Tyree 1989; O'Leary, Malone, and Tyree 1994). Numerous studies in the USA and Canada have found an extremely high prevalence of physical and sexual assault on dating partners by university students. For example, in Canada and the USA 20 to 40% of students report one or more assaults in the previous 12 months. This article provides preliminary data on the extent to which these high rates are found in other countries, and gender differences in dating partner violence in each of the 31 university samples. It also presents results on the reliability and validity of the measures of physical assault and sexual coercion. A long-term objective of the International Dating Violence Study is to test the applicability of theories purporting to explain the occurrence of partner violence in diverse societies, but work on that aspect of the study is only at the beginning stage. The International Dating Violence study is focused on the dating relationships of university students for a number of reasons. (1) Data on university students can be obtained in a uniform way by inexpensive questionnaires. This puts participation in the study within the resources of investigators in many countries. (2) In many countries, heterosexual relationships in the form of “dating” are more likely to exist among university students than in other sectors of the population. (3) As indicated above, a large number of studies show that both physical assaults and sexual coercion occur very frequently among student dating couples. The rate of physical assault typically ranges from 20 to 40 percent (Sugarman and Hotaling 1989). (4) \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 2 University students are an appropriate population because in many countries they constitute a sizeable population. In the USA, for example, there are about 15 million currently enrolled. (5) Students are at a formative period in their lives, especially in relation to the development of appropriate patterns of behavior with a partner. The patterns manifested at this age are often enduring features of their relationship, (Murphy and O'Leary 1989; O'Leary et al. 1989; O'Leary, Malone, and Tyree 1994; Pan, Neidig, and O'Leary 1994). Definition Of Dating For purposes of this study dating was defined as a dyadic relationship involving meeting for social interaction and joint activities and an explicit or implicit intention to continue the relationship until one or the other party terminates or until some other more committed relationship is established such as cohabiting, engagement, or marriage. The social norms for dating, and actual dating behavior, differ according to many dimensions, including individual differences, racial/ethnic and socioeconomic group differences, historical era, and cultural context. Despite these differences, there are also some inherent structural similarities; for example, it is a dyadic relationship and the parties usually invest time and energy. Therefore social interact ional process typical of dyads are likely to apply regardless of whether the relationship was arranged by parents or friends, by newspaper or by internet, or by one party initiating the development of a relationship. Thus, the principles of theories such as exchange theory and conflict theory could apply to these types of relationships (whatever, they are named) in diverse national contexts. THE INTERNATIONAL DATING VIOLENCE STUDY The International Dating Violence study is being conducted by members of a research consortium located at universities in every major world region (see appendix). A detailed description of the study, including the questionnaire and all other key documents, is available on the website http://pubpages.unh.edu/~mas2. Some of the key characteristics of the study are: • It uses a well-validated instrument to measure partner violence – The revised Conflict Tactics Scales or CTS2 (Straus, Hamby, Boney-McCoy, and Sugarman 1996). The CTS has been successfully used in many countries with respondents of all socioeconomic levels, including low educated rural and urban populations and university students (Archer 1999; Straus 1990). • It examines an unusually wide range of etiological variables because it uses the Personal and Relationships Profile or (PRP) to measure 22 variables associated with partner violence (Straus, Hamby, Boney-McCoy, and Sugarman 1999; Straus and Mouradian 1999) • It provides the opportunity to examine the effect of socio-cultural differences on the etiology of dating-violence by examining the issue in different socio-cultural settings, including universities in all the major world regions. • Each of the participating researchers also investigates issues that are locally salient by adding questions to the standard questionnaire. For example, the questionnaire for Hong Kong has questions to measure the importance to the respondent of “saving face.” \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 3 • The procedures were field tested with more than a thousand students at three different universities in the USA, and as noted above, the CTS2 is part of the questionnaire has been used successfully in more than 31 countries. A Collaborative Project There is a core questionnaire that each member of the International Dating Violence Research Consortium translates and then back-translates to maintain “conceptual equivalence” (Straus 1969) across the sites. In addition, as mentioned above, the members add questions to measure variables that are uniquely important for their site or to measure constructs that are needed to test a theory of particular interest. These procedures allow the benefits of both standardized measures for all the sites, and also the benefits of culturally informed investigations of unique issues in each university. The research is being conducted under the terms of an explicit agreement (see the website cited above) that specifies the rights and obligations of the Principle Investigator and of each consortium member. This agreement protects the intellectual property rights of each member of the consortium, while also facilitating cross-national comparisons and publication of the results. Anticipated Results A large number of important issues are being investigated. The issues fall into two broad groupings: estimating the prevalence partner violence, and testing theories concerning the individual and societal characteristics associated with partner violence. Prevalence And Chronicity Of Four Aspects Of Dating Violence. The Revised Conflict Tactics Scales or CTS2 (Straus, Hamby, Boney-McCoy, and Sugarman 1996) provides prevalence rates and frequency of occurrence (chronicity) for the following aspects of partner violence: Physical assault Physical injury Psychological aggression Sexual Coercion Tests Of Theories. The major focus of the International Dating Violence study is to test theories that might explain the differences in dating violence between universities and differences within universities between individuals. The theories will be tested using the data on the characteristics of the couple in the demographic data section of the questionnaire and the measures of risk factors for partner violence in the Personal And Relationships Profile (Straus, Hamby, Boney-McCoy, and Sugarman 1999; Straus and Mouradian 1999) part of the questionnaire. To take a simple example, the younger the person, the greater the probability of crimes of all types, including assaulting a partner assaulting (Stets and Straus 1989; Suitor, Pillemer, and Straus 1990). Therefore, differences between universities in the average age of students could explain the difference between universities in assault. Another example is the hypothesis that sites with high rates of dominance by one partner tend to have high rates of physical assault on a dating partner. The hypothesis will be tested at both the individual couple level and at the societal level. For the individual couple level analysis, dominance will be measured by the Dominance scale of the Personal and Relationship Profile. For the societal level analysis, then mean Dominance scale score will be in the equation as a “level 2” variable using multi-level modeling (Bryk and Raudenbush 1992). The societal level analyses will also \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 4 use nation-to-nation differences on measures such as the percent that women’s education is of men’s education in each of the nations. Limitations All research has limitations, and the International Dating Violence study is no exception. There is space in this article to only describe one of the most serious of the limitations. It is that the study cannot make generalizations about nations, or even about university students in general in the nations where the data was gathered. This is because students are not necessarily representative of a nation, and because the student samples are not being chosen to be representative of all students. However, this is a theoretically driven study and it is appropriate to make generalizations based on conceptualizing each university as a social entity whose characteristics differ. For this paper, the characteristics of each of the universities are measured as the mean of students at each university (for example, their mean age), or as the percent of students with a certain characteristic such as the percent who were cohabiting with a partner. We also plan to use multilevel modeling (Bryk and Raudenbush 1992). For example, a preliminary analysis using the mean Dominance score of each university site as the independent variable found significant university-to-university differences in Dominance, and that these differences were significantly related to physically assaulting dating partners. The appropriate generalization from these results is not that students in country X are more prone to assaulting a dating partner, but that students at universities where dating relationships are characterized by a higher level of dominance by one partner tend to also have a high rate of assaults on dating partners. This is a generalization about a theory rather than a generalization about specific universities. It provides the type of data that, for example, Dutton argued is needed to provide empirical support for the patriarchical-dominance theory of partner violence, and which his review of the research found was virtually absent (Dutton 1994). METHODS FOR THIS PAPER The Data Before the data was gathered at each university, steps to ensure informed consent, privacy, and safety were reviewed and approved by a locally appropriate board or administrator. Informed consent included explaining the purpose of the study and the fact that the questionnaire contains questions on sensitive issues, including sexual relationships. The same information was printed on the cover page of the questionnaire. The students are asked to help by completing the questionnaire and at the same time they are told that answering the questionnaire is entirely voluntary, and that they can also choose to omit any question or questions they do not want to answer. The students were also told that the questionnaire is completely anonymous and that to preserve their privacy, they should not write their name anywhere on the questionnaire. Finally, the questionnaires were collected by placing a box for the questionnaires near the exit door. This preserves privacy because no one knows if the questionnaire dropped in the box was completed or blank. As students left, they were given a page listing of phone numbers of local resources to call on for help if they have experienced any of the problems covered in the questionnaire. Samples (Insert Table 1 about here) This article reports results for students at the 31 sites listed in Table 1. The number of cases at each site ranged from 132 to 741, with a mean of 279. The questionnaires were usually administered in classes taught by members of the consortium and in other classes for \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 5 which they could make arrangements. Thus, as indicated in the section on Limitations above, it is a convenience sample. The results describe what was found for the students in those classes in each country and cannot be taken as representative of students in general. Some of the characteristics of the students in each site are given in Table 1. Of the 8,666 students in Table 1, only those who had been in a dating relationship lasting a month or more were used for this paper. This varied from 100% to less than a third in Pune, India, where heterosexual dating is not part of the culture. In addition, as in other surveys, not every student answered every question. Indeed, to respect the privacy and the voluntary nature of participation, the instructions emphasized that respondents were free to omit any question they did not wish to answer. Aggregate Level Data This paper reports the results of macro-level analyses using as the cases for the analyses the 31 universities for which data was available at the time this paper was written. The macro-level variables for each site were created using the SPSS program AGGREGATE to create a data file of 31 cases, each university being one case. The variables in the macro-level file are the means for the students in each university such as the mean age of the students at each university, or the percent of students in each university who have a certain characteristic, such as the percent cohabiting or the percent who physically assaulted a dating partner. Data Quality Control The completed questionnaires were examined for questionable response patterns, such as reporting an injury but not reporting any assaults as having occurred; or cases with an implausible response such as attacking partner with a knife or gun 10 or more times in the past year. About 4% of the cases were identified as questionable and were removed from the sample. Measures Of Partner Violence Physical assault, injury, and sexual coercion were measured by the revised Conflict Tactics Scales or CTS2 (Straus, Hamby, Boney-McCoy, and Sugarman 1996). The CTS2 asks respondents how many times they committed any of the acts in the past year and how many times their partner had done so. This provides data on both perpetration and victimization. However, because of space limitations, the results in this paper are restricted to perpetration. The CTS2 was designed to provide scores for different levels of severity in the sense of risk of injury to the victim. The difference between the CTS2 score for “minor” and “severe” assaults is similar to the legal distinction between “simple” and “aggravated” assault. The differences between the CTS2 scores for minor sexual coercion and the scores for use of threats and use of force are similar to recent revisions of rape statutes in the USA which specify levels of sexual assault. When rates are based on both the minor and severe acts, it will be identified as the “overall rate.” Physical Assault. The specific items in the CTS2 used to measure assault are: Minor Assault Threw something at my partner that could hurt Twisted my partner's arm or hair Pushed or shoved my partner Grabbed my partner Slapped my partner \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 6 Severe Assault Used a knife or gun on my partner Punched or hit my partner with something that could hurt Choked my partner Slammed my partner against a wall Beat up my partner Burned or scalded my partner on purpose Kicked my partner Respondents were asked how many times they had committed each of the physical assault behavior items in the past year. The response categories are: Once in the past year, Twice in the past year, 3-5 times in the past year, 6-10 times in the past year, 11-20 times in the past year, More than 20 times in the past year, Not in the past year, but it did happen before, This has never happened. These response categories can be used to create five different types of physical assault scores . For this paper, they were used to estimate the percent of students who perpetrated one or more of the violent acts in the scale. Injury. The items in the Injury scale of the CTS2 are: Minor Injury Had a sprain, bruise, or small cut because of a fight with my partner Felt physical pain that still hurt the next day because of a fight with my partner Severe Injury Passed out from being hit on the head by my partner in a fight Went to a doctor because of a fight with my partner Needed to see a doctor because of a fight with my partner, but I didn't Had a broken bone from a fight with my partner Sexual Coercion. Sexual coercion is defined as behavior intended to compel the partner to engage in unwanted sexual activity. The sexual coercion scale covers a range of coercive acts, from verbal insistence to physical force. The specific items to measure sexual coercion are: Minor Insisted on sex when my partner did not want to (but did not use physical force) Made my partner have sex without a condom Insisted my partner have oral or anal sex (but did not use physical force) Threat Used threats to make my partner have sex Used threats to make my partner have oral or anal sex Force Used force (like hitting, holding down, or using a weapon) to make my partner have sex Used force (like hitting, holding down, or using a weapon) to make my partner have oral or anal sex \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 7 Cultural Norms And Beliefs The Personal and Relationship Profile (Straus, Hamby, Boney-McCoy, and Sugarman 1999; Straus and Mouradian 1999) part of the International Dating Violence questionnaire includes questions to measure cultural norms and beliefs that support violence, such as approval of slapping a husband or wife, the permissibility of a husband forcing sex when a wife does not want to have sex, a question on the belief that women who are raped “asked for it,” and questions about the necessity of parents using corporal punishment. The exact wording of these questions are given in places where the results for these questions are presented so that readers can have that information before them when they are thinking about the results. These questions asked the students about their individual attitudes and beliefs. However, the results to be presented are for geographic sites rather than for individual respondents. Consequently, the results can best be interpreted as indicating the strength of cultural norms and cultural beliefs about these issues at each site. The response categories for these questions were Strongly Disagree, Disagree, Agree, and Strongly Agree. This paper uses the percent of students at each university of who did not Strongly Disagree as the measure of cultural norms approving violence. Those students were coded as 1 to indicate that they accepted the attitude statement. All other responses were coded as zero. The reasoning is that students who did not mark the Strongly Disagree answer category, had a reason for not doing so, and the reason was probably that there was a situation where they did approve of slapping a partner, or did think that it as acceptable to force sex. Exploratory analysis found somewhat larger correlations for the items dichotomized in this way than for these items dichotomized in other ways, such as the percent who Strongly Agree = 1. Measures Of Respondent Characteristics Gender. Males were coded as 1 and females were coded 2. About two out of three students in the sample are female. The predominance of females occurred because most of the classes in which the questionnaire was administered were courses in psychology and sociology, which are courses in which women tend to predominate. However, there is also considerable variation among universities. For example, in the Portuguese sample only 40% are female. Socioeconomic Status (SES). A scale to measure of socioeconomic status was computed using the number of years of education completed by the student’s father and mother and family income. Each of these three variables were transformed to Z scores and summed. The sum was transformed to a Z score. The advantage of this approach to measuring SES is that the score has the same interpretation at all sites regardless of variations in the educational system or the wealth or poverty of the nation. Specifically, the scores indicate the number of standard deviations the student’s family of origin was above or below the mean of the families of all students at their university. Social Desirability Scale. Research that uses self-reported data needs to take into account the tendency of some respondents to minimize socially undesirable behavior. This study used the Social Desirability scale of the Personal And Relationships Profile (Straus, Hamby, Boney-McCoy, and Sugarman 1999; Straus and Mouradian 1999). This is a 13-item scale adapted from Reynolds short form of the Marlowe-Crowne social desirability scale (Reynolds 1982). The scale measures the degree to which a respondent tends to avoid disclosing undesirable behavior. The items in the scale consist of behaviors that are undesirable but true of almost everyone such as “I have never deliberately said something that hurt someone’s feelings.” Consequently, the more of these almost universal items a respond denies, the less likely the respondent is to disclose more seriously undesirable information such \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 8 as assaulting a partner and other forms of crime. The theoretical range of the social desirability scale is from 13 to 52. For this sample, the scores ranged from 18 to 52, with a mean of 34.2 and a SD of 4.8. Because of the importance of confounding with social desirability, all subsequent analyses control for score on the social desirability scale. Age. The students ranged in age from 18 to 40. The mean was 21.9, but there were two universities where the students were significantly younger (Juarez, Mexico and New Hampshire, USA). It is well established that the younger a couple, the more likely there is to be violence in the relationship (Stets and Straus 1989). Because the sites varied significantly in age, this variable was controlled in the analysis of university-to-university differences in violence against a partner. Relationship Length. The students had been in the relationship they described for times ranging from one month more than two years. The median number of months was 13 and mean almost 14. Because the nature of a relationship can change over time, it is important to control for this variable when comparing the universities. Data Analysis Because the purposes of this paper are to introduce readers to the International Dating Study and describe prevalence of assault and sexual coercion of dating partners, the main mode of data analysis was to compute the rates at each of the 31 universities and present them in tables listing the 31 sites in rank order. The rank order tables permit easy identification of the sites that are low, middle, and high in physical assault and sexual coercion of dating partners. Because gender is such an important aspect of violence between partners, all tables provide separate rates for male and female perpetrators. The psychometric adequacy of the rates of assault and sexual coercion was investigated by computing the Alpha coefficients of internal consistency reliability of the physical assault and sexual coercion measures for each site. Correlation analyses, controlling for social desirability response bias, was used to provide preliminary evidence on construct validity. Because the N for the correlation analyses is low (31 sites), the .10 level of significance was used to evaluate statistical significance. RELIABILITY AND CONFOUNDING WITH SOCIAL DESIRABILITY Reliability Although the Conflict Tactics Scales has been used in hundreds of studies, most have been studies in North America, and the studies outside of North America have seldom reported reliability coefficients. Table 2 provides the coefficients for each of the 31 universities in 16 countries. (Insert Tables 2 and 3 about here) \\mstraus\C\My Documents\A2\ID10-19\ID16 ID overview Montreal\ID16A c.doc,10-Jun-03. Page 9 The Physical Assault scale was the most consistently reliable of the three measures of perpetration of violence against a dating partner. At all but two of the 31 sites, the coefficients were above .70. For the Injury scale, coefficients below .70 were found at five sites. The Sexual Coercion scale was the least consistently reliable. There were 12 sites with coefficients below .70, of which seven were below .60. At one of the sites, all three of the coefficients were below .70 and at two others, two of the coefficients wee below .70. Confounding With Social Desirability Response Bias A scale that purports to measure socially undesirable behavior such as perpetration of physical assaults or injury might have a high level of internal consistency reliability because some respondents consistently avoid disclosing that type of behavior. Consequently, it is important to determine if the differences between sites are an artifact of confounding with siteto-site differences in willingness to disclose socially undesirable behavior and beliefs. Table 4 shows that scores on a social desirability response bias scale which measures the tendency of a respondent to avoid disclosing undesirable behavior did not differ importantly from university to university or between men and women. As expected, the higher the score on the social desirability scale, the lower the rate of physical assault and sexual coercion, indicating that the social desirability scale is operating as intended. However, these correlations were not high enough to be an important threat to validity. In addition, the correlation of the social desirability scale with the measures of physical and sexual violence did not differ much between men and women or between universities.