|Home | About | Journals | Submit | Contact Us | Français|
This paper examines agreement on reports of male-to-female (MFPV) and female-to-male (FMPV) psychological, physical and sexual violence among White, Black and Hispanic couples in the U.S. Using a probability sample, separate face-to-face interviews were conducted in respondents’ homes with both members of 1,025 intact couples living in the 48 contiguous states. The overall survey response rate was 72%. Results indicate that agreement on each of the three types of violence is low, independent of perpetrator’s gender and ethnicity. Kappa coefficients of agreement ranged from .07 to .48. Higher agreement is obtained for psychological violence, followed by physical and sexual violence. Depending on the type of violence, between 30% (psychological aggression) and 90% (sexual coercion) of events would not have been identified if identification required agreement between partners. Logistic analysis shows that the severity of violence is the only variable that increases the likelihood of agreement across the three types of violence.
The objective of this study is to assess levels of agreement in reporting psychological, physical, and sexual intimate partner violence (IPV) among White, Black and Hispanic couples in the U.S. general population. IPV rates in the U.S. population remain high. This is especially so among some U.S. ethnic minority groups. The majority of national household surveys have demonstrated higher rates of male-to-female partner violence (MFPV) and female-to-male partner violence (FMPV) among Blacks and Hispanics compared to Whites (Caetano, Schafer, Clark, Cunradi, & Raspberry, 2000; Schafer, Caetano, & Clark, 1998; Straus & Smith, 1990). In the 1995–1996 National Violence Against Women Survey (Tjaden & Thoennes, 1998), lifetime rates of rape and physical assault were higher among Black, American Indian/Alaskan native and Mixed Race women than White women. Lifetime rates comparing Hispanic and non-Hispanic women showed a higher rate of rape among Hispanics (18.4% versus 14.6%) and a slightly lower rate of physical assault (51.8% versus 53.2%). This national survey research has relied on self-reports by the victim, the aggressor, or more rarely by both, to establish these prevalence rates. However, self-reports of deviant behavior, such as IPV, usually lead to underreporting, which happens because subjects forget that the violence occurred, misunderstand questions asked by interviewers or purposely conceal victimization or perpetration of violence (Caetano, Schafer, Field, & Nelson, 2002). The problem of underreporting can be partially addressed by interviewing both members of the couple (Szinovacz & Egley, 1995); in this case, the violence that is not reported by a member of the couple may be reported by the other. However, collecting data from both members of the couple inevitably raise issues about agreement on reports of IPV between the individuals being interviewed. Schafer et al. (1998) examined rates of physical MFPV and FMPV in a national sample of couples and found a significant level of disagreement between partners regarding the occurrence or absence of physical violence behaviors. Caetano et al. (2002) extended that analysis to examine the level of agreement among Black, White, and Hispanic couples. While levels of agreement varied by item among ethnic groups, the proportion of MFPV events reported by women only or men only (about one-third each) was similar for Blacks and Whites. The sex differential among Hispanics was greater, with men reporting a higher proportion of MFPV. For FMPV, women reported a higher proportion of events than men in all ethnic groups. Whether these patterns of reporting are consistent across different types of partner violence among White, Black and Hispanic couples in the general population has not been determined.
Methodological problems are present in assessing other types of IPV besides physical aggression (Alksnis, Desmarais, Senn, & Hunter, 2000; Coker et al., 2002). Self-report measures of sexual violence victimization, in particular, have been called into question (Fekete, 1994; Pearson, 1997), with critics postulating that self-report instruments overestimate the occurrence of such violence. Few studies have examined reports of sexual violence by both members of a couple, and those that have report divergent findings. For example, Meyer, Vivian, and O’Leary (1998) found that women in a clinical sample reported sexual IPV victimization more often than their male partners reported perpetration, particularly more severe sexual violence. Among the community control group in that study, however, male partners reported more sexual coercion than their female partners (but neither reported threatened/forced sex). On the other hand, in a community-based study by Marshall and Holtzworth-Munroe (2002), women were more likely than their male partners to report both sexual coercion and threatened/forced sex; although a small proportion of sexual IPV among couples was reported by the male partner only. Neither study, however, examined these differences by race/ethnicity.
Psychological or verbal abuse/aggression is also a critical component of IPV. Psychological IPV has been consistently associated with concurrent or subsequent physical and/or sexual IPV (Coker, Derrick, Lumpkin, Aldrich, & Oldendick, 2000; Marshall & Holtzworth-Munroe, 2002; Smith, Thornton, Devellis, Earp, & Coker, 2002), particularly among women (Coker et al., 2002). In an analysis of the National Violence Against Women Survey, Coker et al. (2002) found women less likely than men to report psychological IPV alone (12.1% and 17.3%, respectively). Studies to date, however, have not examined reporting patterns among couples or between ethnic groups with regard to psychological IPV.
Besides assessing levels of agreement in reporting psychological, physical, and sexual IPV among White, Black and Hispanic couples in the U.S. general population, this paper will also report on the association between couples’ sociodemographic characteristics, the severity of violence and agreement in reporting violence. Based upon initial findings from baseline data (Caetano et al., 2000), it is hypothesized that agreement among couples will be low for each form of IPV; and that level of agreement among couples for MFPV and FMPV will be comparable across ethnic groups.
Subjects in this study constitute a multistage random probability sample representative of married and cohabiting couples in the 48 contiguous United States. These couples were interviewed twice, five years apart, in 1995 and 2000. However, the analyses in this paper are cross-sectional, and are based on the second wave of interviews that took place in 2000. This is because 1995 data cover physical aggression only, while 2000 data cover physical, sexual, and psychological aggression.
At the time of the first interview, in 1995, all couples 18 years of age and older living in randomly selected households were eligible to participate. This process identified 1,925 couples, of which 1,635 couples completed the 1995 interview for a response rate of 85%. Included in the sample were oversamples of Black and Hispanic couples. In 2000, when the second interview took place, the 1,635 couples previously interviewed were contacted again to participate in the five-year follow-up. At follow up, both members of 15 couples were either dead or incapacitated, leaving 1,620 couples (1,635–15) to be re-interviewed. Interviews were successfully completed with 1,392 couples, or 72% of the 1,925 couples from the 1995 original eligible sample (or 85% of the couples actually interviewed in 1995). Among these couples 1,025 were intact and homogeneous regarding White, Black or Hispanic ethnicity (406 White, 232 Black and 387 Hispanic). The present analysis is limited to these couples. They are 71% of all the White, Black and Hispanic couples interviewed in 1995, during the baseline survey.
The Committee for the Protection of Human Subjects of the University of Texas Houston Health Science Center approved this study. All subjects interviewed signed a written informed consent before being interviewed. In both 1995 and 2000, face-to-face interviews were conducted in respondents’ homes with standardized questionnaires. Members of the couple were interviewed separately. This methodology has been identified as leading to more accurate identification of IPV than those that rely on one person’s report (Szinovacz & Egley, 1995).
This is described in Caetano, Ramisetty-Mikler, and McGrath (2003). Briefly, in regard to violence-related variables, women who reported being victimized by violence during their childhood were less likely to be among non-respondents compared to those who were not victimized.
The three forms of IPV including sexual coercion, physical assault and psychological aggression were assessed by scales adopted from The Revised Conflict Tactics Scales (CTS2) (Straus & Gelles, 1990b). Couples were asked about the occurrence of violent behaviors that they may have perpetrated against their partners, or that their partners may have perpetrated against them, during the past year. For all of these variables, violence was considered as present when at least one of the partners reported a violent incident.
Minor psychological aggression items included: insulted or swore at partner; shouted or yelled at partner; stomped out of room or house or yard during disagreement with partner; or did something to spite partner. Severe psychological aggression items included: called the partner fat or ugly; destroyed something belonging to the partner; accused the partner of being a lousy lover; or threatened to hit or throw something at the partner.
Minor physical assault items included: threw something at the partner that could hurt; twisted the partner’s arm or hair; pushed or shoved; grabbed; or slapped. Severe physical assault items included: used a knife or gun on the partner; punched or hit the partner with something that could hurt; choked; slammed the partner against a wall; beat up; burned or scalded on purpose; or kicked the partner.
Minor sexual coercion items included: made the partner have sex without a condom; insisted on sex when the partner did not want to but did not use physical force; or insisted that the partner have oral or anal sex but did not use physical force. Severe sexual coercion items included: used force like hitting, holding down, or using a weapon to make the partner have oral or anal sex; used force like hitting, holding down, or using a weapon to make the partner have sex; used threats to make the partner have oral or anal sex; and used threats to make the partner have sex.
Respondents who identified themselves as “black of Hispanic origin (Latino, Mexican, Central or South American, or any other Hispanic origin)” and “white of Hispanic origin (Latino, Mexican, Central or South American or any other Hispanic origin)” were classified as Hispanic. Respondents who selected “black, not of Hispanic origin” were classified as Black. Respondents who selected “white, not of Hispanic origin” were classified as White. The ethnicity variable represents couples ethnicity (homogenous) where both partners are from the same ethnic group.
The age of respondents was measured continuously in years.
Each respondent was asked about the highest grade or year in school (1 thru 12th grade or GED), technical or vocational educational, college, four-year degree or graduate or professional degree that he or she completed. Education was used as a continuous variable.
Respondents were given 12 income categories to identify into which their total pre-tax household income fell. On the basis of the responses, the original 12 income categories were collapsed into four categories: equal to or less than $20,000; $20,001 to $30,000; $30,001 to $40,000; and greater than $40,000. Income was treated as a categorical variable, with the less than $20,000 income category serving as the reference group.
Concordance rates were established using kappa coefficients. Chi-square analyses were conducted to compare differences in reporting by sex and across ethnic groups. Logistic regression was conducted to examine the relationship between severity of aggression and agreement while controlling for ethnicity, age, cohabitation, male and female education and couple income. However, the relatively low prevalence rates of sexual coercion precluded the examination of sociodemographic influences. Data were weighted to correct for probability of selection into the sample and non-response rates. In addition, a post-stratification weight was calculated to adjust the sample to known population distributions on certain demographic variables (ethnicity of the household informant, metropolitan status, and region of the country).
Rates of male-to-female and female-to-male psychological aggression are more common than physical assault and sexual coercion, independent of the perpetrator’s sex and ethnicity (Table 1). Between 50% and 60% of the sample, independent of ethnicity, report minor psychological aggression. Events of severe psychological aggression are less common, but still have a relatively high prevalence, with rates ranging from 14% to 30% across men and women.
Within each ethnic group, rates of male-to-female and female-to-male minor physical assault are relatively similar. These rates range between 7% and 17%, and are about 2 times higher among Blacks and Hispanics than among Whites. Rates of male-to-female and female-to-male severe physical assault are also relatively similar among Hispanic and Whites. Among Blacks, however, the rate of female-to-male severe physical assault is about 1.5 times higher than the rate of male-to-female physical assault.
Rates of male-to-female sexual coercion are higher than those of female-to-male, independent of ethnicity. These rates are also lower among Whites than among Blacks and Hispanics. Severe sexual coercion is a relatively rare event in this sample.
The interviewing process first selected at random an adult from all the eligible adults in the selected household. If this adult was married or lived with an intimate partner, then the partner was interviewed. It is possible then to provide violence rates based on the information provided by the first partner interviewed, and violence rates based on the interview with both partners. These should mirror rates obtained in surveys with a partner only versus surveys with couples. In general, rates for physical assault and sexual coercion are almost 2 times higher when both partners are interviewed. For instance, rates of physical assault based on reports with the first partner interviewed are: Blacks, 15.5%; Hispanics, 15.2%; Whites, 7.3%. Rates based on interviews with both partners are: Blacks, 26.9%; Hispanics, 26.1%; Whites, 11%.
Rates of sexual coercion based on interviews with one partner only are: Blacks, 13.5%; Hispanics, 8.6%; Whites, 12.7%. Rates based on interviews with both partners are: Blacks, 24.4%; Hispanics, 15.2%; Whites, 12.7%. Rates of psychological aggression do not vary as much, maybe because they are higher than the rates of physical aggression and sexual coercion. Rates of psychological aggression based on interviews with one partner only are: Blacks, 74.1%; Hispanics, 60.7%; Whites, 68%. Rates based on interviews with both partners are: Blacks, 85.3%; Hispanics, 77.8%; Whites, 82.2%
Kappa coefficients of agreement are generally low independent of type of violence, perpetrator’s sex and ethnicity (Table 2). Coefficients for minor psychological aggression range from .38 to .48. Coefficients for severe psychological aggression range from .27 to .38. Coefficients of agreement for physical assault are slightly lower than those for psychological aggression, and especially low for female-to-male severe physical assault. The coefficients of agreement for sexual coercion are also low. However, many of the coefficients for severe sexual coercion could not be computed because there were too few events to analyze.
Among Blacks and Hispanics, the proportion of incidents of male-to-female psychological aggression identified by women is slightly higher than the proportion identified by men. Among Whites, the proportions are equal (Table 3). Identification of acts of physical assault varies by sex across ethnic groups. Women identify a lower proportion of male-to-female physical assault events than men do. In other words, women identify themselves as victims less often than men identify themselves as perpetrators. This difference is slightly more pronounced among Whites than among Blacks and Hispanics. Identification of male-to-female sexual coercion also varies by sex and by ethnic group. Identification is higher among White men than among Black and Hispanic men. Identification is similar for Black men and women, higher for Hispanic women than men, and lower for White women than men. So, ethnicity and sex seem to have a considerable impact in the identification of sexual coercion.
Percentage agreement on psychological aggression is relatively similar across ethnic groups, ranging from 62% to 68%. Percentage agreement for physical assault is lower than for psychological aggression, especially among Whites. However, differences across ethnic groups are not significant. Percentage agreement on sexual coercion ranges from 7% to 10%, and is the lowest among the three types of violence being considered.
Women self-identify themselves as perpetrators of psychological aggression more than men identify themselves as victims. Turning to female perpetrated physical assault, women identify a lower proportion of female perpetrated physical aggression than men do only among Whites. Among Blacks, women and men identify the same proportion of female perpetrated physical assault, and among Hispanics, women identify more female-to-male physical assault than men do.
For female perpetrated sexual coercion, identification is higher among Hispanic women than among White and Black women. Identification is also higher among White men, followed by Black and Hispanic men. Black and White men identify more events than women in the respective groups. Among Hispanics, the reverse is true. Hispanic women identify a higher proportion of events than Hispanic men do.
Men and women agree that events of psychological aggression happened in 60% to 71% of the time. Agreement between men and women in identifying physical assault is low and slightly lower for Whites and Hispanics than for Blacks. Agreement occurs in about one fifth of all events of physical aggression, independent of ethnicity and sex of the perpetrator. This suggests that about 80% of all events of physical aggression would be missed if agreement between partners is required to count such events as present. Agreement on the identification of sexual coercion is lower than agreement on identification of psychological aggression and physical assault. For female-to-male sexual aggression agreement is around 10%. Thus, the proportion of missed events of sexual coercion would be around 90% for female perpetrated sexual coercion if agreement between partners was required to consider the events as present.
Results from the logistic regression show that women’s education and severe psychological aggression by the men increase the likelihood of agreement on male perpetrated psychological aggression. Women’s education and severe psychological aggression by the woman increase the likelihood of agreement for female perpetrated psychological aggression (Table 4). Results for physical assault show that the only variable that increases couples’ likelihood of agreement is severe physical assault by the male. Finally, the severity of sexual coercion by the male increases the likelihood of agreement on male perpetrated sexual coercion. Unfortunately, because of the low levels of reported sexual coercion, this logistic regression does not control for potential confounders such as age, ethnicity, income and education, which are included in the other analyses.
Minor psychological aggression is the most common form of partner abuse. While continued minor and severe psychological aggression can be damaging to individuals and relationships, the perpetration of this type of violence is less stigmatizing because the damaging effects are not observed immediately, which may lead couples to engage in it more readily. Regarding ethnicity, Blacks have higher rates than Whites and Hispanics for almost all forms of violence under analysis, with exception of minor psychological aggression. Previous analyses of this data set have produced similar results for rates of physical aggression (Caetano et al., 2002; Schafer et al., 1998). Results from the 1975 National Family Violence Survey indicated that Blacks had annual rates of severe husband-to-wife violence twice that of other ethnic minorities, and 400% greater than the rate for Whites (Straus, Gelles, & Steinmetz, 1980). Other analyses of this same survey reported an annual rate of 11% for severe husband-to-wife violence among Blacks and 3% among Whites (Cazenave & Straus, 1990). Previous longitudinal analyses indicated that Black couples were 2.5 times more likely than Whites to report new incidents of intimate partner violence and 2 times more likely than Whites to report recurrence of intimate partner violence at five year follow up (Caetano, Field, Ramisetty-Mikler, & McGrath, 2005).
Rates of male-to-female psychological aggression and physical assault approximate female-to-male rates of these two types of violence. This similarity of rates of intimate partner violence in general population samples has been reported before (Schafer et al., 1998; Straus & Gelles, 1990b). It has also been the subject of much discussion in the literature (Archer, 2000; Straus & Gelles, 1990a). Critics argue that in spite of the similarity in rates, male perpetrated violence (e.g., physical aggression) is bound to be more damaging than female perpetrated violence, and authors are in general agreement that the health consequences of violence are greater for women than for men (Stets & Straus, 1990).
Rates of male perpetrated minor sexual coercion are higher than those for female’s about 4% points to 6 % points. This sex difference should not come as a surprise as well, given the general expectation that men will initiate most consensual or non-consensual sexual acts. Perhaps somewhat surprising is that the sex difference between rates is not larger, or in other words, that female perpetrated minor sexual coercion is as high as reported in the sample, especially among Blacks (13%). Certainly, the definition of sexual coercion used in the analyses influences the magnitude of the rates. Severe sexual coercion is rare, and most couples report minor sexual coercion only. This is defined by items such as, “insisted on sex when the partner did not want to but did not use physical force” or “insisted that the partner have oral or anal sex but did not use physical force”. The interpretation of these items is subjective, and may have influenced the reporting of these events by women.
Rates of physical assault and sexual coercion based on interviews with both partners are about two times higher than the rates based on interviews with one partner only. This is not true for psychological aggression, most probably because these rates are considerably higher, which leaves less room for variation. Nevertheless, this variation in rates confirms that survey research on intimate partner violence that relies on self-reported information about the violence needs to seriously consider interviewing both partners to elicit as much information about violence as possible. The difference in rates is also a good indicator of potential disagreement between partners on the occurrence of violence. That is, if the source of difference in rates does not result from willful withholding of information about violence by one partner, then the difference can only be due to memory lapse or disagreement about the nature of act. This then leads to the low levels of agreement discussed below.
Agreement between partners on reports of perpetration and victimization by psychological aggression, physical assault and sexual coercion is low, independent of the sex of the perpetrator and the ethnicity of the couple. Previous reports in the literature provide similar findings (e.g., Caetano et al., 2002; Schafer et al., 1998). Survey research methodologists and cognitve psychologists have made important contributions to understanding the factors that influence responses to survey questions, and that have the potential to affect agreement in partners’ responses about violence. Sudman and Bradburn (1982) identify the following four factors as leading to response error, or as discussed here, disagreement: memory lapses, motivation to tell the truth, communication (understanding the question), and knowledge of the answer. Tourangeau (1984) identifies four stages in answering questions: comprehension, retrieval of information, judgment of information relevance, matching response to answer. Each of these stages can contribute to disagreement between two partners if they go through a stage differently from one another. If partners have different motivation to tell the truth or if they access the sensitivity and desirability of the response in different ways, this may lead to willful underreporting of violence by that partner. If partners have different levels of recall of past events, or recall events differently, this too may lead to genuine disagreement about the nature of the event, which may be seen as violence by only one partner. The frequency and severity of the violence being recalled may also influence agreement. More frequent violence may be easier to recall because of its recurrence in the lives of the people affected. More severe violence may also be easier to recall because it leads to more severe physical, psychological or social consequences (involvement of family, friends, police). Comprehension of question content may lead to disagreement because partners may misunderstand item content, and provide either false positive or false negative reports of violence. Many of the items in the CTS2 can be affected by subjective interpretation, as for instance, the items on sexual coercion mentioned above.
The low level of agreement between partners has serious consequences for studies of intimate partner violence that rely on self-report. As the results in Table 3 indicate, if agreement between partners is required for an event to be considered as present, the percentage of incidents of partner violence that can be missed is considerable. This percentage ranges from about a third of events missed for psychological aggression to about 90% of all events of sexual coercion. This is why Szinovacz and Egley (1995) emphasize the importance of interviewing both partners in studies of intimate partner violence. However, because few studies have interviewed both partners, this issue of agreement and its impact on prevalence rates is considerably minimized.
Data in Table 3 are also interesting because they do not clearly indicate that men, as the conventional wisdom suggests, underreport perpetration of violence. For instance, the results show that the proportion of events of male-to-female psychological aggression reported by women, the victims, is only slightly higher than the proportion reported by men, the perpetrators. Also, against conventional wisdom, more men self-report perpetration of physical assault than their partners report victimization. Although, as stated above, this could result from women’s underreporting of these events for fear of reprisal. This can also be taken as an indication that men are more aggressive than women in intimate relationships (see O’Leary (2000) for a discussion, also Heyman and Schlee (1997) and Stets & Straus (1990)). But it also indicates that men do not always hide about their roles as perpetrators. In the context of survey research with general population samples, this lack of underreporting by men may occur because most of these men are not reporting very severe acts of violence, which are shameful and could also result in an arrest.
The results also indicate that sex and ethnicity interact with type of violence to modify self-report. For instance, more White men self-report perpetration of sexual aggression than White women report victimization. When compared to reports of perpetration by women, Hispanic men seem to underreport victimization by physical assault, perhaps due to shame. Thus, a simple interpretation such as that the men perpetrate the violence but do not report it does not fit the pattern of results. There are important and complex interactions of sex roles, ethnicity and the type of violence that must be taken into account to understand the patterns of reaction to violence and the reporting of violence observed in the sample. Further, the type of sample (clinical versus community) and the severity of violence also seem to interact with sex to affect level of reporting of violence. Thus, in Heyman and Schlee’s (1997) comparison of reporting of violence by men and women interviewed in clinical and community samples, the women reported more severe violence than the men, particularly in the clinical sample (9% versus 58%). Differences in the community samples were not as marked for both any type of violence and severe violence.
Women’s educational level and the severity of violence are positively related to level of agreement. However, women’s education only affects agreement on psychological aggression, which makes this association difficult to interpret. It is possible the more educated women would be more perceptive and/or less tolerant of psychological aggression, which would lead them to report this type of violence more often. However, this should not necessarily lead to a higher level of agreement in reporting with their partners, especially because the results do not show any effect of men’s education on agreement. Regarding the association between severity of violence and agreement, severe violence is probably easier to remember and less likely to be misinterpreted by one of the partners involved in the violence, which should raise agreement levels.
This study has several strengths. The sample under analyses is representative of couples in the continental U.S. Self-reports of IPV were collected from both partners, thus enhancing the probability of identification of spousal violence (Stets & Straus, 1990; Szinovacz & Egley, 1995). Bilingual interviews were conducted, and oversampling of Blacks and Hispanics allowed for ethnic group-specific analyses. However, the study also has limitations. The analyses do not consider the frequency of violence, and do not differentiate couples with reciprocal violence from those where the violence was perpetrated either by males or females alone. Finally, analysis of IPV based on a sample of intact couples may underestimate this violence in the general population because couples who separated due to intimate partner violence are not included in the analysis. Limitations of the CTS, the scale used to assess IPV, have also been discussed in the literature. These include lack of assessment of the context and of the consequences of violent acts committed by husbands and wives (Straus & Gelles, 1990a, 1990b).
Work on this paper was supported by a grant (R37-AA10908) from the National Institute on Alcohol Abuse and Alcoholism to the University of Texas School of Public Health.