Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
2 Nature and Scope of Violence Against Women As discussed in Chapter 1, the 1996 NRC report Understanding Vio- lence Against Women outlines a series of methodological problems that have impeded the development of knowledge in this area. In the years since that report was produced, some of these problems have been addressed (e.g., the inclusion of questions about violence against women in surveys pertaining to other types of high-risk or violent behav- ior), yet others remain. This chapter focuses on the need for a coordinated and integrated research strategy that can build stronger scientific data- bases to enhance our understanding of violence against women. Much of what is known about the violent victimization of women has been derived from methodologically disparate survey data. Certainly, survey research has been instrumental in setting some parameters for the scope of two types of violence intimate-partner violence and sexual as- sault. Data on these problems indicate that the risk of such victimization varies substantially across racial and ethnic groups (Dugan and Apel, 2002; Rennison and Welchans, 2000; Tjaden and Thoennes, 1998~. Never- theless, survey research has been less successful in providing reliable esti- mates of the prevalence and incidence of intimate-partner violence and sexual assault, about the context of these violent events, about the devel- opmental patterns of such violence over time, and about the ways in which women's victimization experiences may be linked to women's offending behaviors (Campbell et al., 2002a; Richie, 1996~. To advance a more sys- tematic approach to the study of violence against women, this chapter considers in turn the strengths and weaknesses of the major national 35
36 RESEARCH ON VIOLENCE AGAINST WOMEN datasets that provide information on violence against women; other datasets that were not designed to examine violent victimization, but have included relevant questions pertaining to the violent victimization of women; information provided by current datasets about the extent and nature of violence against women; and methodological issues that have precluded linking extant datasets, improving measurement, and enhanc- . . ng ongoing survey researc I. MANOR DATASETS A wide array of datasets provides information on violence against adult and/or adolescent females, although this is not always the primary focus of the various data collection efforts. Many of these datasets include information from national samples of women or large groups of U.S. women that are representative of a particular population. Each has both strengths and limitations. The most serious limitations of existing datasets are as follows: (1) most were designed for primary purposes other than collecting information on violence against women; (2) most are not con- tinuous and so cannot show changes over time; and (3) although many collect data on similar issues (e.g., prevalence), definitions of violence against women, as well as data collection instruments, vary, making it difficult to compare results. These and other shortcomings, especially nonresponse and false-response errors, need to be addressed in future research. At a 1998 conference on research on violence against women, sponsored by the Centers for Disease Control and Prevention (CDC) and the National Institute of Justice (NIT), a matrix of datasets that include some information on violence against women was created with an eye to determining how to link information from disparate surveys, an issue dis- cussed further below. The committee has expanded this matrix somewhat and included it here as Table 2-1. In a paper commissioned for the work- shop, Campbell et al. (2002a) describe the following major datasets on which most researchers in the field currently rely: The National Crime Victimization Survey (NCVS), sponsored by the Bureau of Justice Statistics (BIS), is the second-largest ongoing gov- ernment-run U.S. survey (Bachman, 2000~. It is the most extensive victim- ization data source, documenting characteristics of victims and nonvictims aged 12 and older living within sampled housing units. In addition to detailed information on each household and the interviewed individual within that household, the survey documents respondents' re- cent experiences as crime victims, including details of each event and its consequences. From 1972 to the present, data collection, using a rotating panel design in selected housing units, has been conducted seven times
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 37 during each 3-year period. In 2000, the response rate was 93.0 percent of eligible households and 89.3 percent of eligible individuals (Rennison, 2002~. This survey is one of only a handful of continuous datasets that collects information on violence against women. Each year, BIS uses the data to publish reports on current crime distributions and to document patterns and consequences for several types of victims (Bureau of Justice Statistics, 2000~. NCVS data do permit an examination of whether and how violent victimization differs for women and men, but they have several limita- tions. While the redesign effort in 1992 led to improved estimates of do- mestic and sexual violence (see Bachman and Taylor, 1994, for a thorough description of the redesign effort and its improvements), analysis of long- term trends has been curtailed because the content of assault and sexual assault items was changed. In addition, no information is collected on victimization history and the communities and cultures within which the violence occurs, although NCVS data with census-tract codes attached are available to researchers with permission from the U.S. Census Bureau. Data pertaining to the immediate context of violent events (e.g., urban setting, home ownership, public housing) are also limited. The National Surveys of Family Violence (NSFV) encompassed two cross-sectional surveys conducted by the University of New Hampshire under the auspices of the Family Violence Research Program, sponsored by the National Institute of Mental Health (Gelles and Straus, 1988; Straus et al., 1988~: the first was conducted in 1975 and relied on in-person home interviews; the second was conducted in 1985 and relied on random-digi- tal-dial telephone interviews. Each was designed to provide a compre- hensive examination of violence in the family, including spouse abuse, child abuse (including physical punishment as a form of discipline), sib- ling-to-sibling abuse, and parent abuse (victimization of a parent or par- ents by the child/children of the family). The surveys were intended to provide data on the prevalence of these various types of family violence and to identify the risk (and protective) factors involved. Family violence was operationalized with the Conflict Tactics Scales (CTS) (Straus,1979~. These scales are based on self-reports of 18 behaviors (grouped into three subscales consisting of rational, verbal, and violent acts) that may have occurred within the context of a disagreement. Over the course of the two surveys, the CTS were modified to include an addi- tional series of questions regarding whether an act of violence produced an injury that required medical attention (Gelles, 1987~. These surveys represented the first nationwide examination of vio- iBJS uses the first interview for bounding purposes only; it is not included in these data.
38 RESEARCH ON VIOLENCE AGAINST WOMEN TABLE 2-1 Characteristics of Major Violence Against Women Datasets Has a Direct Question on Violence Against Women of Data Dataset Sample Collection Frequency Context of Survey Use of Health or Social Services (by victin Measuret Supplementary Homicide Homicide Reports incidents (SHR) reported by police departments National Crime Victimization Survey (NCVS) National sample of households National Incident- Criminal Based Reporting incidents System (NIBRS) reported by law enforcement agencies National Longitudinal Study of Adolescent Health ADD Health (NLSAH) National sample of adolescents, grades 7-12 Monitoring the National Future sample of 8th, 10th, and 12th graders Pregnancy Risk Assessment Monitoring System (PRAMS) National Youth Survey Continuous Criminal justice Yes Continuous Criminal justice Yes Continuous Criminal justice Yes Health No Continuous Health No Postpartum Continuous Health Yes women from 32 states; oversampling of various racial/ ethnic groups and women who delivered low- birthweight children National sample of youth Criminal Justice No No Health ar some soci No Some hea Some hea and social No No
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN rtasets 39 is a rect question Violence rainst amen Use of Health or Social Services (by victim) Measured Etiologya Comorbid Factorsb Can Be Prevalence Assessed Data Incidence Chronicity Data Data No Yes No Yes Yes No Health and Potentially Potentially Potentially Yes Yes some social No No No No Yes No Some health Potentially No No Yes No Some health Potentially Potentially Potentially Potentially Potentially and social No No No Yes No No Yes No Yes Yes Yes Yes Continued
40 TABLE 2-1 Continued RESEARCH ON VIOLENCE AGAINST WOMEN Has a Use of Direct Health or Question Social Frequency Context on violence Services of Data of Against (by victin Dataset Sample Collection Survey Women Measurec National Health Nationalsample Health No Health and Social Life of general Survey (NHSLS) population, aged 18-59 National College Nationalsample Health No No Health Risk of undergraduate Behavior Survey students (NCHRBS) National Physicians Continuous Health No Health Ambulatory Medical Care Survey (NAMCS) National Hospital National Continuous Health No Health Ambulatory sample of Medical Care patient record Survey forms (NHAMCS) National Hospital National Continuous Health No Health Discharge Survey sample of (NHDS) inpatient record forms for short-term hospital stays National Health Nationalsample Continuous Health No Health ar Interview Survey of households some soci (NHIS) National Survey Nationalsample Continuous Health No Health of Family Growth of general (NSFG) population, aged 15-44 National Vital National sample Continuous Health No No Statistics System of death (NVSS) certificates
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 41 is a Use of rect Health or question Social Comorbid Violence Services Factorsb ,ainst (by victim) Can Be Prevalence Incidence Chronicity amen Measured Etiologya Assessed Data Data Data Health Yes Yes Yes Yes No No No Health No Health No Health No No No Yes No Yes No No No Yes No No No Yes No No No Health and Potentially Yes Potentially Potentially Potentially some social Health No No No No No No No No No No No Continued
42 TABLE 2-1 Continued RESEARCH ON VIOLENCE AGAINST WOMEN Has a Use of Direct Health or Question Social Frequency Context on violence Services of Data of Against (by victin Dataset Sample Collection Survey Women Measurec National Morbidity Followback Survey (NMFS) National sample of people who have died in a given year, with oversampling of African-Americans Continuous Health No No National Nationalsample Continuous Health No Health Electronic Injury of emergency Surveillance room visits System (NEISS) involving injuries National Nationalsample Continuous Health No No Household of general Survey on Drug population, Abuse (NHSDA) aged 12 and up National Violence National sample Criminal justice Yes Health Against Women of general Survey (NVAWS) population New Hampshire Sample of Health Yes No Youth at Risk New Hampshire high school students National Surveys National Health Yes Social of Family survey of families Violence (NSFV) Youth Risk Nationalsample Continuous Health Yes Health ar Behavior of youth, social Surveillance grades 9-12 System (YRBSS) Behavioral Nationalsample Continuous Health Yes No Risk Factor of general Surveillance population System (BRFSS)C effects on violence against women. aRisk factors for intimate-partner violence. bother conditions that affect the magnitude of violence against women.
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 43 is a Use of rect Health or question Social Comorbid Violence Services Factorsb ,ainst (by victim) Can Be Prevalence Incidence Chronicity amen Measured Etiologya Assessed Data Data Data No Yes Health No Yes No No No No No No No No Potentially Yes No Yes Yes Health Yes No Yes Social Yes Health and Yes social No Yes Yes Yes Yes Yes Yes No No No Yes Yes No Yes No Yes Yes Yes Yes Potentially Potentially No CNational data are derived from aggregating state statistics. States differ in their inclusion of questions on intimate-partner violence and in the types of questions asked. NOTE: Where geocodes are available, linkages could be used to examine area spatial effects on violence against women.
44 RESEARCH ON VIOLENCE AGAINST WOMEN fence perpetrated by both male and female members of a couple. Numer- ous research publications have resulted from the data collected, some es- timating the prevalence of intimate-partner physical violence, and others examining associations between violence victimization and other topics (Gelles and Straus, 1988; Straus et al., 1988; Gelles, 1987; Straus, 1979~. The completion rate for the first survey was 65 percent of the entire sample; 84 percent of eligible respondents completed the second survey (Straus and Gelles, 1986~. Despite the prominent role of the NSFV in research on partner vio- lence, it has several well-known limitations. The CTS assess violence only in the context of a disagreement, and their earlier versions do not measure the severity of behaviors reported. Also, their original format did not as- sess a wide range of types of violence (including sexual violence). The new CTS2 includes emotionally abusive and sexually violent tactics and can be formatted so that the impact of a tactic is also measured. However, neither form of the CTS gathers information concerning whether the vio- lence used was defensive in nature a shortcoming of almost all current instruments employed in measuring intimate-partner violence. Because partner violence may best be conceptualized as a chronic condition that encompasses interrelated ongoing events, examining such violent behav- iors out of context may miss important dimensions of the overall situation (Smith et al., 1999~. The National Violence Against Women Survey (NVAWS) was de- signed by Tjaden and Thoennes (1999) to remedy a limitation of the NCVS: because the NCVS is a general crime survey aimed at generating annual estimates of many types of crime, sample size constraints limit its useful- ness as a source for better understanding historical and recent relational contexts likely to be associated with violence against women. Using a ran- dom-digit-dial household telephone survey, 8,000 women and 8,005 men aged 18 and older were sampled throughout the United States from No- vember 1995 to May 1996. The interviews were completed by 72 percent of the women and 69 percent of the men sampled (Tjaden and Thoennes, 2000~. The NVAWS employed a modified version of the CTS to collect data on physical assaults; data on sexual assaults and stalking, as well as injuries resulting from these victimization experiences, were also col- lected. This survey was unique among national surveys in its focus on respondents' lifetime histories of violence and its attention to gathering detailed information on perpetrators that could be linked across violent incidents. Despite these advantages, however, the NVAWS did not pro- vide estimates of violent victimization that are comparable to those ob- tained by either the NCVS or the NSFV; the referent populations, some of the screening questions, and the quantification of series victimizations all differ. The NCVS focuses on assaults on respondents aged 12 and older,
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 45 while the NVAWS focused on victimizations of respondents aged 18 and older. The NCVS also differs from the NVAWS in the approach used to count series victimizations. The National Youth Survey is a nationally representative longitudi- nal survey of 1,725 persons who were aged 11-17 in 1976 when the study began and are now aged 37~3. The study has collected information on these individuals over time to assess their changing attitudes, beliefs, and behaviors with regard to deviance, exposure to delinquent peers, self-re- ported depression, delinquency, drug and alcohol use, victimization, pregnancy, abortion, use of mental health and outpatient services, vio- lence by respondent and acquaintances, use of controlled drugs, and sexual activity. Data are available on the demographic and socioeconomic status of respondents, on parents and friends, and on neighborhood prob- lems. The sample is 53 percent male and 47 percent female. The ethnicity of participants is comparable to that of the general population of the United States. The completion rate of eligible youth sampled was 73 per- cent in the initial wave of the survey; the completion rate of original re- spondents was 78 percent for the ninth wave in 1992 (Menard, 2002~. This longitudinal survey reports on intimate-partner violence committed by both male and female respondents. Its estimates of lifetime prevalence are considerably higher than those derived from the surveys described above. OTHER DATASETS Supplementary Homicide Reports (SHR), part of the Federal Bureau of Investigation's Uniform Crime Reporting (UCR) program, is a major source of data on homicides (Bureau of Justice Statistics, 2002~. Supple- mental reports on homicide incidents have been voluntarily submitted monthly by local law enforcement agencies since 1976. These reports de- tail such information as age, race, and sex of victims and offenders, weapon use, circumstance of the crime, and the residential population and county and Metropolitan Statistical Area (MSA) codes of the report- ing agency. SHR is particularly useful for research on intimate-partner homicide because it also collects data on the victim-offender relationship, categorized as intimate (spouse, ax-spouse, boyfriend, or girlfriend), other family, other acquaintance, or stranger. However, some limitations have been noted. First, supplemental reports are voluntary; about 91 percent of homicides reported in the UCR are included in SHR (Bureau of Justice Statistics, 2002~. The actual incidence of homicides is underestimated com- pared with the National Vital Statistics System of the National Center for Health Statistics (Annest and Mercy, 1998~. Also, ethnicity is determined by the observations of the reporting officer. A number of other ongoing data collection efforts include questions
46 RESEARCH ON VIOLENCE AGAINST WOMEN on violence against women. Examples are the Behavioral Risk Factor Sur- veillance System, National Longitudinal Study of Adolescent Health, Pregnancy Risk Assessment Monitoring System, Youth Risk Behavior Sur- veillance System, National Health Interview Survey, and National House- hold Survey on Drug Abuse (for a more complete list, see Table 2-1~. How- ever, these surveys vary in the extent to which they systematically include measures of violence against women, gather data on the relationship be- tween the victim and perpetrator, and address the situational context in which the violence occurs. The definitions and measures of violence against women used in these surveys also differ. Furthermore, these gen- eral population surveys rarely oversample racial minorities, and the data collection methods used (telephone and mail surveys) frequently preclude obtaining information on institutionalized and homeless women's experi- ences of violent victimization. Other one-time surveys have been con- ducted to measure various aspects of violence against women; these sur- veys are not described here (see, e.g., Fisher et al., 2000~. There is also a handful of datasets from other countries that collect information on vio- lence against women and could be expected to shed light on the problem in the United States. The following are examples of such datasets. The Canadian Violence Against Women Survey examined the safety of women inside and outside the home. This survey, a one-time-only study based on a sample of households and conducted in 1993, served as a model for the NVAWS described above. Data elements collected in the Canadian survey included "perceptions of fear, sexual harassment, sexual violence, physical violence and threats by strangers, dates/boyfriends, other known men, husbands and common-law partners" (Statistics Canada, 1993~. The target population for the survey was all Canadian women 18 years of age and over. Institutional populations and residents of the Yukon and Northwest Territories were excluded. The overall re- sponse rate was 54 percent. Canada also collects data annually on police- reported family violence. The Dunedin Multidisciplinary Health and Development Study is a longitudinal investigation of a representative birth cohort of infants born between April 1, 1972, and March 31, 1973, in Dunedin, on New Zealand's South Island. The base sample was composed of 535 males and 502 fe- males. The Dunedin birth cohort has been reassessed at 2-year intervals from ages 5 through 15 and at 3-year intervals between ages 15 and 21, with 97 percent of the living members of the age 3 base sample participat- ing at age 21. The Dunedin data provide a unique opportunity for assess- ment of the developmental origins of partner violence that may have im- plications for understanding of the problem in the United States and elsewhere.
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN EXTENT AND NATURE OF VIOLENCE AGAINST WOMEN 47 Prevalence of Violence Against Women Estimates of the prevalence of intimate-partner violence range widely, depending on the type of survey, the content of survey questions, and the sampling frame used (see Table 2-2~. In nationally representative surveys in which women self-report physical abuse victimization by an intimate partner, past-year prevalence ranges from 0.43 percent of respondents to the NCVS reporting aggravated or simple assault to 11-12 percent of re- spondents to the NSFV reporting nonfatal physical abuse. In the five sur- veys that use the same measure the CTS2 self-reports of intimate-part- ner physical abuse against women place past-year prevalence at 1.1-13.6 percent. In the four surveys that use the same measure plus comparable samples, the range is 1.1-9.1 percent for victimization in the past year. Existing longitudinal studies place the past-year prevalence of inti- mate-partner physical abuse against women in the 20-50 percent range. Among all the surveys/studies of the prevalence of intimate-partner physical abuse victimization shown in Table 2-2, only the NCVS and the NVAWS report women's victimization as being higher than that of men; the others indicate that men report higher rates of severe and overall vic- timization. For the crime of rape, the NVAWS reports a prevalence 1.5 times higher than that reported in the NCVS. Only one of these surveys, the NCVS, is repeated at regular intervals, so that comparability among the various surveys already compromised by differences in age range, sampling frame and sample sizes, length of recall period, and the manner in which questions are asked and violence is defined is further dimin- ished by differences in the reporting period. These kinds of conflicting prevalence rates make it impossible to develop credible estimates of trends over time or to determine the effectiveness of interventions. Understanding Risk Factors While existing datasets represent important strides in providing basic information about the overall scale and severity of violence against women, scholars are only beginning to learn about the risk factors for this particular type of violence. There is a long tradition of such research on other types of violence, such as violent delinquency. Risk factors for vio- lence against women can reside in the individual victim, certain groups of women, characteristics of offenders, or social and spatial circumstances surrounding a violent event. However, as with risk factors identified in other criminology research, they cannot be used to predict a specific event
48 RESEARCH ON VIOLENCE AGAINST WOMEN TABLE 2-2 Prevalence of Nonfatal Intimate-Partner Violence (IPV) Physical Abuse (Past Year) % Report Survey Year Sample Measurement Female National Prevalence Surveys National Surveys 1975 Households containing Conflict Tactics 12.1% ova of Family Violence cohabiting (married or Scale 3.8% sevt nonmarried) couple n = 2,143 1985 Households containing Conflict Tactics 11.3% ova cohabiting (married or Scale, modified 3.0% sevt nonmarried) couple n = 3,520 National Alcohol 1992 Married and cohabiting Conflict Tactics 9.1% over and Family Violence persons aged 18+ Scale, modified 1.9% sevt Survey n = 1,970 NationalAlcohol 1995 Married and cohabiting couples Conflict Tactics 5.21%a to Survey n = 1,599 couples Scale, modified National Violence 1995-1996 Persons aged 18+ Conflict Tactics 1.3% Against Women n = 8,000 women, 8,000 men Scale, modified Survey Married and cohabiting Conflict Tactics 1.1% persons aged 18+ Scale, modified n = 5,982 men, 5,655 women NationalCrime 2001 Persons aged 12+ 0.43% Victimization n = 79,950 Survey Longitudinal Studies Dunedin 1993-1994 Study participants aged 21 Physical abuse 40.9% Multidisciplinary who were in romantic scale (CTS plus Health and relationships and their 4 additional items) Development Study partners n = 360 couples Study participants aged 21 Conflict Tactics 38.8% who were married or cohabiting Scale n = 250 Study participants age 21 who Conflict Tactics 27.1% ova were married, cohabiting, Scale or dating n = 861 NationalYouth 1992 Study participants who were Conflict Tactics 20.2% ant Survey married or cohabiting Scale n= 1,340 aBoth partners reported the act occurred. bOnly respondent reported the act occurred.
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN V) 49 % Reporting IPV Physical Abuse Victimization ment Female Male Total Tactics 12.1% overall, 11.6% overall, 16.0% overall 3.8% severe 4.6% severe 6.1% severe Tactics 11.3% overall, 12.1% overall, 15.8% overall edified 3.0% severe 4.4% severe 5.8% severe Tactics 9.1% overall, 9.5% overall, edified 1.9% severe 4.5% severe Tactics 5.21%a to 13.61%b 6.22%a to 18.21%b 7.84%a to 21.48%b edified Tactics 1.3% 0.9% edified Tactics 1.1% 0.6% edified 0.43% 0.08% 0.26% abuse 40.9% 47.4% ES plus anal items, Tactics 38.8% 55.8°X Tactics 27.1% overall, 12.7% severe 34.1% overall, 21.2% severe Tactics 20.2% any, 5.7% severe 27.9% any, 13.8% severe
50 RESEARCH ON VIOLENCE AGAINST WOMEN or to predict that a particular person will become a victim or an offender. They can often, however, indicate propensities for and patterns of risk by gender, race and ethnicity, and circumstances. Gender In general, overall victimization rates of women are low (compared with those of men or juveniles, for example) an important reason for the almost total absence of research on this problem historically. However, the NCVS indicates that victimization by intimates accounts for 20 per- cent of violence experienced by women and 3 percent of that experienced by men (Rennison, 2003~. Moreover, because women are most often vic- timized in "safe spaces" where no one witnesses the crime, they are more vulnerable to repeat attacks and are more likely to be severely injured or killed by intimate partners than by others. It is for these reasons that most research on the victimization of women has focused on intimate-partner violence (Bureau of Justice Statistics, 2000; Tjaden and Thoennes, 2000; National Research Council, 1996~. Interestingly, in nationally representative longitudinal studies in the United States and the above-described 21-year birth cohort study in New Zealand, women have reported higher levels of perpetration of intimate- partner violence than men, and men higher levels of victimization than women (Moffitt and Caspi, 1999~. Some scholars have attributed this dis- crepancy to methodological problems in the CTS, which were used in these studies. Moreover, when serious violence (i.e., resulting in severe injury or death) is the focus, women do not report such higher levels of perpetration (Kruttschnitt, 2002~. Moffitt and Caspi (1999) also note that male perpetrators are much more deviant (e.g., more likely to use illegal drugs or be chronically unemployed) than their female counterparts. Fi- nally, it is intriguing to note that rates of violent victimization have been declining overall. However, violence against women for all crime types has been declining at lower rates than that against men (Bureau of Justice Statistics, 2000~. Race and Ethnicity In a paper prepared for the workshop, Dugan and Apel (2002) use data from the NCVS over an 8-year period (January 1992 to June 2000) to model risk factors for all cases of nonlethal violent victimization of women. They note that until recently, researchers limited investigations of violence to African-American and white women, lumping groups such as Asian/Pacific Islanders and Native Americans into a generic "other"
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 51 category or omitting them entirely (Dugan et al., 2000; Bureau of Justice Statistics, 2000; Greenfield et al., 1998~.2 The usefulness of disaggregating female victims by race is under- scored by the analysis of Dugan and Apel, who found that Native Ameri- can women have considerably higher rates of victimization than other groups of women, they appear most likely to be victimized by someone they know, and their assailant is often using drugs or alcohol. Asian women have the highest proportion of incidents in public places and are more likely to be victimized by sober strangers or multiple offenders. White, African-American, and Hispanic teenage girls all have higher odds of victimization than young adult women, and those aged 60 or older display a significantly low risk. Having some college, but not 4 years, is positively related to violence for African-American and Hispanic women. Residential stability as measured by number of months living at the same location appears to lessen the risk for white and African-American women only. Circumstances Dugan and Apel (2002) found that the strongest risk factor for violent victimization of women is living in a household with one adult and chil- dren. This risk is greatest for Asian/Pacific Islander women. Controlling for other indicators of poverty, living in public housing is also a risk, es- pecially for African-American and Hispanic women. Dugan and Apel (2002) conclude that living in the city, having more or younger children, or having low income appears to raise the risk of violence for all but His- panic and Asian women. Moffitt and Caspi (1999) found that risk factors for female perpetrators of partner violence include disturbed family rela- tionships, especially weak attachments, harsh discipline, and conflict be- tween parents. Although most research in this field reflects the belief that female vic- timization may be driven by some factors that differ from those affecting rates of male victimization, existing longitudinal studies point to risk fac- tors that are similar to those for other kinds of criminal offending and victimization (Moffitt and Caspi, 1999; Straus and Gelles, 1992; Elliott et 2An important exception is the survey by Tjaden and Thoennes (2000), which describes differences in lifetime prevalence of violent victimization for Asian/Pacific Islander, Native American (including Alaskan Native), and mixed-race women (see also Rennison, 2001~. Such analyses are, however, typically bivariate, leaving unanswered the question of which risk factors are of greater concern for any one group over the others.
52 RESEARCH ON VIOLENCE AGAINST WOMEN al., 1986~. Thus, an important research question is which risk factors spe- cific to partner violence or other violence against women do not apply also to criminal offending or victimization in general. For example, most studies of homicide against women have been descriptive in nature and have focused on the murder of women by intimate male partners. More information about the characteristics of the killing of women in other cir- cumstances would be helpful in understanding the lethal victimization of women in the context of homicide studies in general. Health Consequences of Violence It is widely recognized that violence against women, including inti- mate-partner violence, sexual assault, and rape, is associated with nega- tive physical and mental health outcomes. Many studies have found a correlation between violence against women and emotional and physical health problems that go beyond the immediate effects of the abuse. Rela- tionships have been found between previous physical abuse of women and stress-related physical health problems (Campbell et al., 2002b; Sutherland et al., 1998; Koss and Heslet, 1992), gynecological problems (Campbell et al., 2002b; Coker et al., 2000; Letourneau et al., 1999; Golding, 1996), and neurological injuries (Campbell et al., 2002b; Coker et al., 2000; Diaz-Olavarrieta et al., l999~. A recent meta-analysis revealed that women who experience domestic violence have elevated rates of insomnia, de- pression, post-traumatic stress disorder, panic disorder, and substance abuse symptoms that can persist for years after the abuse ends (Golding, 1999~. One concern noted in Understanding Violence Against Women is that part of what is known about the health of abused women is provided by studies using samples drawn from women seeking medical care or from health plan populations. Such samples may not be representative of all victims: there may be differences in injury types, and uninsured women may not seek care. Hathaway et al. (2000) found that there was no differ- ence between abused and nonabused women in rates of routine health care, although abused women were less likely to have health insurance. Lemon et al. (2002) report no differences in checkups and clinical breast examinations, but note that abused women are more likely to undergo Pap smear screening. However, findings of recent studies using popula- tion-based samples are largely consistent with findings of studies using clinic or health plan populations in showing that women who have been abused are more likely to report physical or emotional disabilities, smok- ing, unwanted pregnancy (Hathaway et al., 2000), high-risk alcohol use (Lemon et al., 2002), gynecologic problems (Plichta and Abraham, 1996), and mental health problems (Hathaway et al., 2000; Danielson et al., 1998~.
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 53 In the Dunedin sample, 65 percent of women who had experienced severe abuse met criteria for one or more disorders listed in the Diagnos- tic and Statistical Manual of the American Psychiatric Association (DSM III-R). Abused women in this cohort were three times more likely than nonabused women to suffer a mental illness. Their disorders included depression, drug dependence, antisocial personality disorder, and schizo- phrenia (Moffitt and Caspi, 1999~. An important concern remains with regard to establishing causality. Most studies focus on correlates of violence, failing to establish the tem- poral sequence of events and leaving the pathways between abuse and health outcomes unspecified. Many victims of intimate-partner violence have reported problems, such as unemployment, lack of transportation, substandard housing, and financial difficulties, that may predispose women to poor health outcomes (Browne et al., 1999; Eby, 1996; Sullivan et al., 1992~. One population-based study, by Sutherland et al. (2001), ad- dressed this problem by investigating whether intimate-partner violence has a significant effect on women's health beyond that which can be ex- plained by poverty. Both income and physical abuse contributed to women's rates of physical health symptoms, and abuse contributed to the variance in physical health beyond that predicted by income level alone. Additional research is needed to further explain the direct and indirect causes of health problems experienced by victims of abuse. Risk of Injury and Death Dugan and Apel (2002) measured the likelihood of injury in all types of violent crimes against women. They found that crimes involving physi- cal contact with known persons and the presence of a weapon were pre- dictors of severe injury. According to the NVAWS, 36 percent of women who had been raped since age 18 and 42 percent of women who had been physically assaulted since age 18 reported that they had been injured dur- ing their most recent victimization. However, most of the injuries were relatively minor (scratches, bruises, welts), while more serious injuries (broken bones, dislocated joints, concussions, lacerations, bullet wounds) were sustained by relatively few of the victims (Tjaden and Thoennes, 2000~. In a study by the RAND Corporation, intimate-partner violence was found to be one of the most common causes of injury in women (Rand, 1997~. When survivors of serious injury are compared with those who were killed, findings suggest that women who are harmed by their hus- bands, as opposed to a live-in boyfriend or acquaintance, are overrepre- sented as victims of homicide (Dugan and Apel,2002~. The Dunedin study found that men who severely injured their partners demonstrated extreme
54 RESEARCH ON VIOLENCE AGAINST WOMEN levels of deviance as characterized by polydrug use, antisocial personal- ity disorder, chronic unemployment, and violent acts against persons out- side the family (Moffitt and Caspi, 1999~. In preliminary findings from her study of women killed by their intimate partners, Campbell (2002) found the most important pre-incident demographic risk factor in predicting le- thality in abusive relationships to be perpetrator unemployment. Dawson (2002) describes a recent framework that has been applied to understanding declines in intimate-partner homicide (see also Dugan et al., 1999, 2000~. The "exposure reduction" framework highlights key social changes that may have contributed to the decline in intimate-part- ner homicide in recent decades. These include changes in the nature of intimate relationships fewer and/or delayed marriages and more di- vorce; improved socioeconomic status of women, including increasing gender equality; and the increased availability of domestic violence re- sources, including legal and social services (e.g., domestic violence courts, shelters). Considering that intimate-partner homicide is often preceded by a history of intimate-partner violence, the exposure reduction frame- work holds that the impact of social changes that help abused women exit violent relationships or prevent women from entering such relation- ships may also reduce the rate of intimate lethal victimization. Further research is needed on the potential of this framework for increasing un- derstanding of murders and assaults of women. If we are to be able to prevent such crimes, moreover, longitudinal research in the United States is needed to determine which risk factors (if any), for which groups of women, are truly unique to lethal events or outcomes involving severe injury. The committee recommends that work be initiated to examine the feasibility and cost-effectiveness of successfully conducting such longi- tudinal studies. METHODOLOGICAL ISSUES Improving Measurement As noted above, many federally and some privately sponsored data collections include information on violence against women. As noted ear- lier, in 1998, NIT and CDC sponsored a conference on improving research on violence against women. The matrix produced at that conference (the basis for Table 2-1) categorizes the existing datasets relevant to this re- search according to the following characteristics: · Whether the survey was a one-time only study or continuous. Con- tinuous was defined as conducted at regular intervals (e.g., every 6
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN 55 months in the case of the NCVS, annually in the case of the Youth Risk Behavior Survey) and expected to be conducted (repeated) in the future. · Whether the survey was precise (generally defined as designed to minimize standard measurement errors). · Whether a supplement or follow-back could be undertaken to bet- ter estimate violence against women. · Whether the survey was a health or criminal justice survey. · Whether there was a social service utilization measure for violence against women. · Whether risk factors for violence against women could be esti- mated from the survey. Despite the variations in the quality of the various surveys listed in Table 2-1, creating linkages among existing survey data would provide important additional resources for scholars interested in the contexts and outcomes of violent victimization. Generally, such linkages will not be possible across individuals in the different datasets but may be possible across common geographic areas, for example, states or cities. In addi- tion, linkages between national-level surveys can be developed that relate differing characteristics of events to one another. A good example is the research on the outcomes of violent victimization based on the merging of data from the SHR with NCVS data (see Felson and Messner, 1996; Kleck and McElrath, 1991~. If the NCVS could be linked to data from the Na- tional Health Interview Survey, information could be obtained on health- related outcomes of violent injuries. Research is needed on the feasibility of linking different datasets and on how to validate survey data with data on clinical, legal, and social outcomes. An important aspect of this linking process will be developing a framework for standard definitions. To advance understanding of violence against women, the constructs researchers use must be valid and reliable across different social settings, samples, and measurement conditions. Currently, the behaviors used to measure or operationalize "violence" or abuse are wide-ranging, and this seriously compromises our understanding of the prevalence and distribu- tion of violence against women. For example, even where the same mea- sures and comparable samples are used, prevalence estimates differ by a factor of 2 (see Moffitt and Caspi, 1999~. Part of the problem is a lack of information on the amount of harm or the nature of an injury resulting from a violent act. As Johnson (1995) demonstrates, research that uses the CTS and does not tap the consequences of various violent tactics comes to a very different conclusion about the prevalence of violent victimization among men and women than would be derived from agency or official data. Although much of Johnson's argument revolves around sampling issues rather than measurement issues, there is no question that differ-
56 RESEARCH ON VIOLENCE AGAINST WOMEN ences in what is counted as a violent act (e.g., pushed or shoved versus broken bones) affect how many women and men are classified as victims of intimate-partner violence. These discrepancies in findings raise a cen- tral question: What elements should be included in the definition of vio- lence, or what kinds of behaviors should be considered violent? Cook (2000) documents at least 29 different measurement instruments used in research on violence against women. These measures vary not only in the types, levels, and degrees of coercion they measure, but also in the sever- ity of the acts they include. Rigorous inquiry into violence against women is precluded when scholars fail to distinguish among what constitutes an act of violence, abuse, or battering. The NSFV suggests that abuse has a normative criterion but uses the terms "abuse," "assault," and "violence" relatively interchangeably, dif- ferentiating only between what are termed "minor" and "severe" violent acts based on the potential risk of injury (Straus and Gelles, 1992:75-85~. These terms fail to distinguish among physical violence, physical aggres- sion, and psychological abuse. This lack of conceptual and operational clarity is particularly problematic when attempts are made to compare survey findings with data on clinical populations, among whom abuse may be determined by specific medical criteria. It is also problematic con- sidering that what constitutes being victimized and what constitutes of- fending may be culturally determined. If we want to be able to determine whether critical aspects of abusive and violent behaviors against women (e.g., their prevalence, incidence, and distribution) differ from those of other kinds of violent behavior, we need to employ consistent definitions and measures. This distinction between nominal and operational definitions applies not just to questions of how violence or abuse is measured, but also to questions of how different research settings introduce measurement prob- lems. Violence may be operationalized differently in clinical, legal, and research settings or, as occurred in the redesign of the NCVS in 1992, even within the same setting over different periods of time. Causes, correlates, and epidemiological and survey estimates of vio- lence may all be sensitive to the conceptual and operational clarity of definitions. Far more than half of violent crimes against women re- main unredressed, in large part because they are unknown to criminal justice authorities. Scholars have developed some important methods for assessing this so-called "dark figure of crime," but a large proportion of violent crimes perpetrated between intimates and family members are still unreported.
NATURE AND SCOPE OF VIOLENCE AGAINST WOMEN Improving Research 57 The committee recommends more research to examine the situational contexts and dynamic interactions that lead to violence against women. Research on violent events complements studies of the individual pro- pensities of victims or offenders, and focuses instead on the occurrence of violence, identifying the specific conditions that channel individual moti- vation and predispositions into violent actions. This approach addresses the social or psychological pathways that bring individuals to specific vio- lent events and the transactions or decisions that comprise the onset, course, conclusion, and aftermath of the event. Recent studies on interpersonal violence among strangers illustrate the confluence of motivation, perceptions of risk and opportunity, and the social control attributes of the setting that shapes the decision to par- ticipate in a violent event, as well as its outcome (Wilkinson and Pagan, 2001~. Other research shows that violence against women serves specific functions for assailants, and that those functions may covary with the type of assault. Tedeschi and Felson (1994) hypothesize that all violence is re- lated to one or more of the following three goals: compliance, identity, and justice. To understand the catalyst for a violent event among inti- mates, researchers must examine the social construction and discourse on male-female relationships, perceived imbalances in power, control dy- namics, identity threats, relationship problems, and communication pat- terns . It is important to recognize that "when violence occurs it is not an isolated event in peoples' lives, but is embedded firmly in the process of interpersonal communication which people use to regulate their lives" . Research on the "sparks," motivations, interaction patterns, and decision making associated with violent events can identify leverage points for reducing the threat of violence or averting it entirely. Some violent events against female victims are stranger assaults, and understanding the situational and structural contexts of those events pro- vides another window on social factors that elevate risks beyond those attributable to individual offenders or victims. While women's victimiza- tion results from various forms of violence, there are commonalities across those events that can be examined to provide a differentiated understand- ing of the unique and shared risks involved. The following are some ex- amples of research on situational contexts and violent interactions: · Research on the processes of victim selection Stranger assaults may appear at first glance to be random occurrences, but there are processes of victim selection that can be studied to identify attributes of individuals, settings, and social interactions that may motivate victim selection.
58 RESEARCH ON VIOLENCE AGAINST WOMEN · Research on location selection The selection of locations for as- saults against women may reflect a rational decision-making process that can be modified to reduce risk and prevent physical or sexual assault. · Research on victim-offenderinteraction patterns Studies of the patterns of social interaction in stranger violence have generated robust theories of violent interaction that can be extended to the unique circum- stances of gender-related assaults. Studies of these interaction patterns in domestic violence show how personality factors interact with situational contexts to launch interaction dynamics that end in assaults by male inti- mate partners against women (see Jacobson and Gottman, 1998; Wilkinson and Hamerschlag, 2002; Wilkinson and Pagan, 2001~. Replications and extensions of this research should encompass a more diverse set of rela- tionships and different types of assault. These studies should examine and decompose the stages of violent events from arousal to aggression- to identify behavioral scripts or cognitive frames that are amenable to in- tervention or prevention efforts. Several datasets currently available can be examined to begin the pro- cess of theory construction. Research designs using survey methods, event history research with samples of individuals, and laboratory experiments can begin to generate the empirical data that will produce a more refined and productive knowledge base from which prevention efforts can be launched. CONCLUSION Although progress has been made in the effort to measure and under- stand the nature of violence against women, a more coordinated research strategy would help remedy the measurement problems that remain. The committee recommends that an effort be made to investigate how to link different datasets and how to link information from these datasets with findings from clinical research to provide more information on the risks of, responses to, and consequences of violence against women and the impact of interventions. Such an effort should include the formulation of a framework for developing standard definitions to overcome the lack of conceptual and operational clarity, as well as other problems involved in measuring violence against women, especially differences in sample se- lection among studies. In addition, more attention should be devoted to developing event-based measures of violence against women.