Findings About Partner Violence from the Dunedin Multidisciplinary Health and Development Study

NT OF ME J T US U.S. Department of Justice R T A I P C E E D

O J C S Office of Justice Programs F A V M F O I N A C I J S R E BJ G O OJJ DP O F PR National Institute of Justice JUSTICE National Institute of Justice R e s e a r c h i n B r i e f Jeremy Travis, Director July 1999 Issues and Findings Findings About Partner Violence Discussed in this Brief: Findings about partner violence from the lon- From the Dunedin Multidisciplinary gitudinal Dunedin Multidisciplinary Health and Development Study of a Health and Development Study representative birth cohort of 1,037 New Zealand men and women by Terrie E. Moffitt and Avshalom Caspi born between April 1, 1972, and March 31, 1973. The 1992–96 National Crime Victimiza- Polynesian). The social class and Key issues: Dunedin researchers tion Survey (NCVS) indicates that in ethnicity of their families matched those decided to study partner violence 1996 victimization by intimates— of the South Island’s general population. as part of their longitudinal study spouses, ex-spouses, boyfriends, girl- because they suspected that some friends, and former boyfriends and The Dunedin cohort was reassessed at study members continued antiso- girlfriends—accounts for about 21 per- ages 5, 7, 9, 11, 13, 15, 18, and 21, with cial activities into adulthood in a cent of the violent crime experienced by 992 (97 percent) of the 1,020 living mem- form not well represented by women and about 2 percent of the vio- bers of the age 3 base sample (51 percent measures of delinquent behavior. male) participating in the last assessment Questions were included in the lence experienced by men. Rates of non- lethal violence are highest among women at age 21 in 1993–94. This history of phase 21 (age 21) interview to reassessment at regular intervals is espe- determine the extent of partner ages 16–24 and women in low-income 1 cially important for the research on part- violence among study members. households. These statistics highlight the importance of studying partner vio- ner violence. Since their childhood, study Key findings: Characteristics of lence among young adults, both married members have revealed problem behav- cohort members who were in- and unmarried, who represent all iors; their confidentiality has never been volved in partner violence include socioeconomic backgrounds. violated, and they have learned to expect the following: no intervention from the researchers (un-

● Although both partners in a re- The Dunedin Multidisciplinary Health less imminent serious danger was posed lationship may not recall the same and Development Study is a 21-year to themselves or others). As a result, by acts in precisely the same way, 70– investigation of a representative birth the time they were 21 years old, they 80 percent of one partner’s report cohort of infants born between April 1, were comfortable giving frank responses was in agreement with the other 1972, and March 31, 1973, in Dunedin, a to questions about partner violence. This partner’s report on whether physi- city of approximately 120,000 people on circumstance offers a special advantage cal violence took place and on the New Zealand’s South Island.2 Perinatal over studies involving self-reports from extent of the abuse. data for 1,139 births were obtained at de- perpetrators held in correctional institu-

● Risk factors in childhood and livery. When these children were traced tions or remanded for treatment. adolescence for male perpetrators at age 3, 91 percent, or 1,037, were as- This Research in Brief summarizes the included poverty and low academic sessed, forming the base sample for the achievement. Female perpetrators longitudinal study. The base sample was findings about partner violence from the showed risk factors of harsh family composed of 535 (52 percent) males and Dunedin study. Among the results are discipline and parental strife. Both 502 (48 percent) females. Fewer than that partner violence is strongly linked to male and female perpetrators also 7 percent of the study members identified cohabitation at a young age; a variety of had histories of aggressive behavior. themselves as nonwhite (Maori or mental illnesses; a background of family

continued… R e s e a r c h i n B r i e f

Issues and Findings adversity, dropping out of school, and intimates in the home. To test this possi- continued… juvenile aggression; conviction for bility, a decision was made to examine

● The strongest risk factor for both other types of crime, especially violent partner violence among study members as male and female perpetrators and crime; drug abuse; long-term unem- they made the transition to young adult- victims was a record of physically ployment; and parenthood at a young hood, became involved in serious rela- aggressive delinquent offending age. tionships, and began to form new families. before age 15. More than half the Given the unique features of this longitu- males convicted of a violent crime Why study partner violence? dinal study, it became important to apply also physically abused their partners. the study’s resources to understanding the The Dunedin study began to examine origins of partner violence. ● About 27 percent of women and partner violence because partner violence 34 percent of men among the research is a natural extension of the The Dunedin cohort, it is important to Dunedin study members reported they had been physically abused by study’s earlier research on childhood be- note, is a birth cohort, not a commu- their partner. About 37 percent of havior problems and teen delinquency. nity sample. Followup was conducted women and 22 percent of men said Although official crime statistics from with all individuals in the cohort; for they had perpetrated the violence. police and courts suggest that offending example, those who had and had not declines rapidly during young adulthood, used battered women’s shelters and ● Domestic violence is most preva- Dunedin researchers suspected that some those who had and had not been con- lent among cohabitating couples. delinquents’ antisocial activities were victed of battery. ● Sixty-five percent of females who continuing, but in a form not easily de- When the study members turned 21 suffered serious physical abuse and tected by official crime statistics; for ex- 88 percent of male perpetrators ample, as abuse of family members or years old (phase 21 of the Dunedin had one or more mental disorders (as defined by the American Psychi- Other Related Publications Supported by the National atric Association in the Diagnostic and Statistical Manual of the Institute of Justice and the Bureau of Justice Statistics American Psychiatric Association [“DSM–III–R”]). Prevalence, Incidence, and Conse- Understanding Violence Against quences of Violence Against Women: Women, by Nancy A. Crowell and ● Women who had children by Findings From the National Violence Ann W. Burgess (eds.), National age 21 were twice as likely to be Against Women Survey (Research in Academy Press victims of domestic violence as Brief), by Patricia Tjaden and Nancy women who were not mothers. Thoennes, NCJ 172837 Violence in Families: Assessing Pre- Men who had fathered children by vention and Treatment Programs, by age 21 were more than three times Stalking in America: Findings Rosemary Chalk and Patricia King as likely to be perpetrators of abuse From the National Violence Against (eds.), National Academy Press as men who were not fathers. Women Survey (Research in Brief), by Patricia Tjaden and Nancy Violence by Intimates: Analysis of Target audience: Mental health Thoennes, NCJ 169592 Data on Crimes by Current or Former practitioners; emergency room and Spouses, Boyfriends, and Girlfriends general practice medical profes- Batterer Programs: What Criminal (Bureau of Justice Statistics Factbook), sionals; victim advocates; juvenile Justice Agencies Need to Know (Re- by Lawrence A. Greenfeld, Michael delinquency, substance abuse, and search in Action), by Kerry Murphy R. Rand, Diane Craven, Patsy A. violence specialists and researchers; Healey and Christine Smith, NCJ Klaus, Craig A. Perkins, Cheryl public health, juvenile justice, and 171683 Ringel, Greg Warchol, Cathy Maston, and James Alan Fox, NCJ 167237 criminal justice officials and practi- Legal Interventions in Family Vio- tioners; juvenile court administra- lence: Research Findings and Policy Publications that have NCJ numbers tors; judges; social scientists; Implications (Research Report), by can be obtained from the National researchers; and others concerned NIJ and the American Bar Associa- Criminal Justice Reference Service about violence prevention. tion, NCJ 171666 by calling 1–800–851–3420.

2 R e s e a r c h i n B r i e f study), questions about partner vio- How trustworthy are the with. Of these, 360 (76 percent) brought lence were embedded in a 50-minute data? Do partners’ reports their partners along to participate in standardized interview about intimate about abuse in their phase 21. Study members and their part- relationships. Information was gath- relationship agree?5 ners were interviewed separately (simul- ered on both positive and negative taneously) with identical questions by conflict-negotiation behaviors occur- The scientific study of partner abuse is different interviewers who did not know ring during the past 12 months. In- controversial in part because there are the responses provided by the other cluded in the interview were the items concerns about the accuracy of data. member of the couple—and their confi- of the Conflict Tactics Scales (CTS)3 Abuse data are usually collected by ask- dentiality was guaranteed. Couples did and other items from published do- ing respondents to “self-report” their not know before they arrived that they mestic violence interviews. Although experiences. The majority of studies usu- would be asked about partner abuse, the CTS is controversial,4 it has been ally interview only one member of each eliminating any opportunity to coordi- used in numerous clinical studies as couple. Can these self-reports of abuse be nate their responses prior to the inter- well as in U.S. national surveys on the trusted? To answer this question, the ex- view. Before interviewers turned to the prevalence of domestic violence. The tent to which partners’ responses were in topic of partner abuse, each participant CTS was included because it allowed agreement was analyzed. was given the opportunity to decline comparison of the Dunedin findings Prior to phase 21, 474 study members discussion of that topic, but none of the with research on partner violence in indicated they were involved with a part- participants refused. The full set of the United States and elsewhere. ner they had been dating for at least 6 questions measured both physical and months, were married to, or were living psychological abuse (see exhibit 1).

Exhibit 1. Item content of physical and psychological partner abuse scales

Physical Abuse Scale Psychological Abuse Scale Participants were asked if they had in the past year: Participants were asked if they had in the past year: Physically twisted your partner’s arma Damaged a household item or some part of the home out of angera Pushed, grabbed, or shoved your partnera, b Deliberately disposed of or hidden an important item of your partner’sa Slapped your partner a, b Become very upset if dinner/housework/home repair work was not donea Physically forced your partner to have sexa Purposely damaged or destroyed your partner’s clothes/car/othera Shaken your partnera Insulted or shamed your partner in front of othersa Thrown or tried to throw your partnera Locked your partner out of the housea Thrown an object at your partnera, b Told your partner that he/she could not work or studya Choked or strangled your partnera, b Tried to stop your partner from seeing/talking to family or friendsa Kicked, bit, or hit your partner with a fista, b Restricted your partner’s use of the car or telephonea Hit or tried to hit your partner with somethinga, b Made threats to leavea Beaten up your partnera, b Tried to turn family, friends, or children against your partnera Threatened your partner with a knife or guna, b Ordered your partner arounda Used a knife or gun on your partnera, b Frightened your partnera Treated your partner like he/she was stupida Given in to your partner but planned revengea Ridiculed your partnera Threatened to hit or throw something at your partner in angerb Told your partner he/she was ugly or unattractivea Become abusive after using drugs or alcohola Thrown, smashed, hit, or kicked something during a disagreement b aFrom Margolin’s “Domestic Conflict Scale” or “Conflict Inventory” (Margolin, G., B. Burman, R.S. John, and M. O’Brien, The Domestic Conflict Instrument, Los Angeles: University of Southern California, 1990). bFrom Straus’s “Conflict Tactics Scales” (Straus, M.A., “Measuring Intrafamily Conflict and Violence: The Conflict Tactics Scales,” Journal of Marriage and the Family 41(1979): 75–88).

3 R e s e a r c h i n B r i e f

Perpetrators’ reports of their own abuse representing the facts in an attempt to scale in the past year. Exhibit 2 shows behaviors were compared with their mislead the interviewer. The statistical the prevalence rates of physical part- partners’ reports of victimization to de- correlations indicate that about 70–80 ner violence by and against men and termine if couple members concurred percent of one partner’s report agreed women. Data from the Dunedin study about the perpetrator’s behaviors. with the other partner’s report. Contrary were compared with data from respon- Couples’ responses to the interview to expectations, agreement between dents under age 25 in the 1985 Na- showed that agreement about whether partners did not vary with the per- tional Family Violence Survey (NFVS; specific abusive behaviors had hap- petrator’s gender or with the type of n=397)8 and respondents under age 24 pened was poor, as has been sug- abusive behavior. in the 1983 National Youth Survey gested by previous research. Study (NYS; n=477).9 members and their partners did not These findings suggest that the data agree about whether, for example, one gathered may confidently be used for Rates from all three surveys were cal- of them had tried to strangle the other. research on the correlates and conse- culated using the same CTS interview However, agreement improved dra- quences of partner abuse. The result- questions, which measured physical matically when the individual items ing high level of confidence in the data violence in the past year. The Dunedin were summed into scales that counted can be attributed to the fact that inter- phase 21 rates are shown first for co- the variety of different abuse behav- views were conducted in a setting in habiting and married study members iors performed in the past year. Al- which participants knew there was no only (n=250) to provide direct com- though members of a couple may not risk of prosecution or requirement to parison with NFVS and NYS, which recall exactly the same acts, they can participate in a treatment program if included cohabiting and married agree on whether or not abuse took they revealed abuse. couples only. Shown fourth are place and on the extent of the abuse. Dunedin rates for married, cohabiting, Agreement was even stronger when How prevalent is partner vio- and dating individuals combined random measurement errors were lence in the Dunedin sample?7 (n=861, excluding 80 study members removed statistically.6 This agreement who had not gone out with anyone in Between one-fifth and one-third of all the past year). Exhibit 2 shows that reveals that disagreement between part- Dunedin study members reported they ners is due to random forgetfulness; Dunedin prevalence rates are similar had experienced one or more of the to the other two national samples. neither partner was deliberately mis- behaviors on the CTS physical abuse

Exhibit 2. Rates of physical partner violence in three studies

Perpetration Victimization 60 60

50 50

40 40

30 30

20 20

10 10 % reporting violence any % reporting violence any 0 0 Women Men Women Men

Dunedin 1993Ð94 Dunedin 1993Ð94 NFVS 1985 NYS 1983 (Married/Cohabiting Only) (Dating/Married/Cohabiting) (n=397) (n=477) (n=250) (n=861) The three studies are the National Family Violence Survey (NFVS), the National Youth Survey (NYS), and the Dunedin Multidisciplinary Health and Development Study.

4 R e s e a r c h i n B r i e f

involved in partner violence. Exhibit 3 Exhibit 3. Rates of involvement in physical partner violence by relationship type (Dunedin males and females) shows the rates of violence for those who were involved in relationships. Given that 48 percent of cohabitating 611 Going Out 223 Cohabiting 27 Married partners and 21 percent of those who were dating experienced partner violence, the colloquial expression “The marriage license is a hitting license” appears to be outdated; the Dunedin study shows that violence cuts across

21% 48% 41% all types of relationships.

The blackened areas indicate the percentage who have perpetrated any physical violence against a partner during the past year. Is physical abuse strongly Note: Eighty are not shown because they had not dated and had no opportunity for partner violence. Although 992 sample linked with mental disorders?10 members participated in phase 21, 51 are not represented here because they either were interviewed in the field where the Conflict Tactics Scales could not be given or they refused the interview about partners. In 1994, the American Psychiatric Association first recognized “physical abuse of an adult” as a “focus of clinical When the Dunedin study members violence data for a representative attention.” An analysis was conducted to were followed up at age 21, they were sample of contemporary unmarried determine whether physical abuse was found to be involved in several types couples who were living together. Most often “comorbid” with mental disorders of relationships. Some study members study members were “going out” with among Dunedin study members. were married, but more were cohabit- someone: Some were in an exclusive (Comorbidity means that a patient suffers ing without marriage, which has be- relationship; others were “playing the from two or more disorders or problem- come a common practice for young field.” About 8 percent had not gone atic conditions at once.) Comorbidity adults in the 1990s. This offered an out with anyone in the past 12 months between abuse and mental disorders was opportunity to report the first partner and thus had no opportunity to become examined because studies of comorbidity among mental disorders have shown that coexistence of multiple psychiatric prob- Exhibit 4. Rates of mental illness among Dunedin perpetrators and victims of lems predicts more severe life impair- severe physical abuse ment, longer duration of the problems, and poorer response to treatment.11 100 Sixty-five percent of Dunedin women who 80 were victims of severe physical abuse12 met criteria for one or more disorders listed in 60 the Diagnostic and Statistical Manual of the American Psychiatric Association

Percent 40 (“DSM–III–R”). Eighty-eight percent of Dunedin men who were perpetrators of 20 severe physical abuse met DSM–III–R criteria (see exhibit 4). Abused Dunedin 0 women were three times more likely to Female Female Male Male suffer a mental illness than nonabused Victims Nonvictims Perpetrators Nonperpetrators women. The male perpetrators were 13 times more likely to be mentally ill than Rates of mental illness among female victims of severe physical abuse, female nonvictims, male perpetrators of severe physical abuse, and male nonperpetrators. The rate for female perpetrators was nonperpetrators. The types of mental virtually identical to the rate for female victims, and the rate for male victims was nearly identical to illnesses among perpetrators varied; they the rate for male perpetrators. included anxiety disorders, depression,

5 R e s e a r c h i n B r i e f

Risk Predictors From the Dunedin Study W hen the study members were age were ages 7, 9, and 15. All of the measures represents whether or not the sample mem- 21, the Dunedin researchers examined 24 were highly reliable. (Previous reports about ber lived with both biological parents. Family individual and family characteristics they research on each risk factor in the Dunedin structure measured at birth (taken from hos- had previously measured during the study study have been published and may be pital records) indicates whether or not the members’ childhood and hypothesized obtained from the researchers—see “Full child was born to a married mother. these characteristics would predict partner Reports From the Dunedin Multidisciplinary abuse. Early childhood characteristics were Health and Development Study.”) There were seven measures of family rela- measured at study members’ birth and at tions. Negative mother-child interaction ages 3 and 5. Middle childhood character- There were six measures of family socio- was assessed when the sample members istics were measured when study members economic resources. Social class (the socio- were age 3. An observer assessed eight were ages 7 and 9. Adolescent characteris- economic status of the parents’ occupation aspects of parenting; for example, if the tics were measured when study members on a six-point scale designed for New mother’s expression was consistently harsh, were age 15. In addition, study members’ Zealand) was measured at sample members’ if her evaluation of the child was constantly * mothers answered questions about their birth and when they were ages 7/9 and 15. critical or derogatory, or if she was rough own mental health when the members Family structure measured at ages 9 and 15 or inconsiderate when handling the child. Family conflict was measured at ages 7/9 and 15 with the conflict subscale of the Correlations of physical abuse with selected childhood and adolescent Moos Family Relations Index, completed by characteristics (at age 21) mothers of the sample members. The con- Male Male Female Female flict subscale contains items such as: “In Perpetrator Victim Perpetrator Victim our family, we believe you don’t ever get Family Socioeconomic Resources Social class at birth Y Y N N anywhere by raising your voice,” and Social class at ages 7/9 Y Y N N “Family members sometimes hit each Social class at age 15 Y N N N other.” A measure of harsh discipline at Born to unmarried mother NNNN ages 7/9 was constructed from a checklist One parent absent at age 9 Y YY N N One parent absent at age 15 YYYY of disciplinary behaviors. Parents were Composite, with other risk factors controlled NNNN asked to indicate if they engaged in 10 be- Family Relations haviors, e.g., “smack your child or hit him/ Negative mother-child interaction at age 3 N N N Y her with something,” “try to frighten your Family conflict at ages 7/9 N Y Y Y child with someone like his/her father or a Family conflict at age 15 N N Y Y Harsh discipline at ages 7/9 N N Y Y policeman,” and “threaten to smack or to Parent-child attachment at age 15 Y Y YY Y deprive your child of something.” Parent- Mother’s mental health problems at ages 7/9 N Y N N child attachment was measured when the Mother’s mental health problems at age 15 N Y N N sample members were age 15 with a 12- Composite, with other risk factors controlled N N Y N item self-report measure from the Inventory Educational Achievements of Parent Attachment. The items measure Stanford-Binet IQ at age 5 N N N Y WISC-R IQ at ages 7/9 Y Y N N the adolescents’ trust, communication, and Reading achievement at ages 7/9 Y Y N N alienation in their relationships with their Reading achievement at age 15 Y Y N N parents. Mother’s mental health problems Age at leaving secondary school YY YY Y Y were measured with a 24-item question- Composite, with other risk factors controlled Y N N N naire that sampled a variety of common Problem Behaviors Difficult temperament at ages 3/5 N N N Y symptoms of emotional disturbances that Conduct problems at ages 7/9 (teacher-parent) Y N N N was completed by sample members’ moth- Conduct problems at age 15 (parent) YYYY ers when the sample members were ages Aggressive delinquency at age 15 (self) YY YY YY YY 7/9 and 15. Juvenile police contact Y N Y YY Substance abuse at age 15 (self) YY YY YY YY Composite, with other risk factors controlled Y YY YY YY There were five measures of educational achievements. At age 5, intelligence quo- Y = yes, a significant risk factor YY = significant and strong risk factor N = no risk

6 R e s e a r c h i n B r i e f

Risk Predictors From the Dunedin Study (continued) tient (IQ) was assessed with the Stanford- the ability to persist in problem solving. Chil- building, hit a parent, fought in the street or Binet Intelligence Scale. IQ was measured dren who scored high on this factor were other public place, struggled to escape from again when the sample members were ages emotionally unstable, irritable, negative, a policeman, used force or threats to extort 7/9 with the Wechsler Intelligence Scale for rough, inattentive, and had difficulty concen- money, or used a weapon in a fight. Juve- Children-Revised. Reading achievement was trating. The measure of conduct problems at nile police contact from when sample mem- measured when the sample members were ages 7/9 was based on combined parent and bers were between ages 10 and 17 was ages 7/9 and 15 by the Burt Word Reading teacher ratings of items from the “antisocial” based on records of police contacts that Test, normed for New Zealand children. Age and “hyperactivity” subscales of the Rutter were obtained from police departments at leaving secondary school was the age at Child Scales. When sample members were throughout New Zealand. The number of which the study member left high school. age 15, conduct problems were measured police contacts in this sample ranged from (Education was compulsory until age 15 in with the Conduct Disorder subscale of the zero to 18. Substance abuse was measured New Zealand.) Quay and Peterson Revised Behavior Problem when the sample members were age 15 Checklist, which was completed by their par- with a “variety” score based on self-reports There were six measures of problem ents. The items in this subscale reflect aggres- of buying alcohol while underage, being behaviors. Difficult temperament was as- sive and interpersonally alienated behaviors drunk in a public place, smoking marijuana, sessed when the sample members were such as bullying, quarreling, disobeying, and sniffing glue, and using other drugs. ages 3/5 by psychological examiners who teasing others. Aggressive delinquency was observed each child in a testing session in- measured when the sample members were * When the same measurement instrument was volving cognitive and motor tasks. Following age 15 with study members’ self-reports of used repeatedly at adjacent assessment ages, the the testing session, examiners rated each researchers were able to combine the two scores aggressive behavior that were obtained in to produce a more reliable and accurate compos- child’s behaviors. Based on the ratings, the private, individual, structured interviews de- ite measure. In this sidebar, the use of such com- researchers identified a dimension that re- veloped for use in New Zealand. Items for posite measures is denoted by a “/” (e.g., 7/9 flected individual differences in reactions to the scale of aggressive behaviors inquired indicates that two equivalent measures of study stress and challenge, impulse control, and whether the subject ever had set fire to a members or their families from assessment ages 7 and 9 were combined). alcohol and drug dependence, antisocial offer prevention as an alternative to current behavior. Second, many previ- personality disorder, and schizophrenia. prosecution. ous studies have focused on only one risk factor—childhood exposure to par- Research shows that emergency room and What are the risk factors in ents’ violence—while neglecting other general practice medical professionals childhood and adolescence for important influences on youngsters’ de- need to be alert to partner violence. In ad- 13 velopment, such as poverty or schooling. dition, the Dunedin study findings reveal partner abuse? that more than one-third of candidates for The unique prospective longitudinal In the early years of the Dunedin treatment from mental health profession- database resources of the Dunedin study, risk factors were measured long als are involved in domestic violence. study provided the necessary means to before either the study families or the Whereas emergency room and general conduct one of the few prospective researchers knew that partner violence practice physicians usually encounter studies of risk factors for partner abuse. would be examined, thereby avoiding victims, and only after an injury, mental Risk factors were tested to determine any potential bias. The risk measures health practitioners have the opportunity whether they were present before were grouped into four broad domains: to identify and help victims before they Dunedin participants’ abusive behavior family socioeconomic resources, fam- are injured if questions about domestic began. Previous research on risk factors ily relations, educational achievement, violence are made a routine part of intake suffered from two flaws. First, child- and problem behaviors. (See “Risk assessment. Moreover, mental health hood factors have been measured pri- Predictors From the Dunedin Study.”) practitioners encounter not only women marily by perpetrators’ recall of past Measures were used that had been taken victims but large numbers of men who are family life, which has proved faulty. during three developmental periods: at risk of being perpetrators. If practition- Adults involved in violence often early childhood (ages 3–5), middle ers were trained to screen for partner vio- “remember” their childhoods in ways childhood (ages 7–9), and adolescence lence risk, the mental health system might that provide self-justification for their

7 R e s e a r c h i n B r i e f

(age 15).14 The pattern of results shows In terms of prevention policy, the finding were found to pose risks for partner that male perpetrators’ backgrounds in- that partner abuse in adulthood is pre- abuse. This underscores the importance clude primarily poverty and poor school dictable from certain characteristics of prevention programs that involve both achievement. In contrast, female perpe- during—and even before—adolescence parents and schools. trators’ backgrounds include primarily suggests that primary prevention of part- disturbed family relationships, espe- ner violence should begin as early as Is partner violence closely cially weak attachment, harsh discipline, youngsters develop an interest in the op- linked with other kinds of and conflict between parents. Poverty posite sex. One clear implication is that violence?15 and school failure were less important. children in secondary school (ages 12 to Perpetrators of both sexes have a long 17) are not too young to learn healthier Interestingly, each of the risk factors history of aggressive behavior problems. ways to handle conflicts with partners. found for partner violence also posed a For male and female perpetrators, the Violence education may become as im- risk for other kinds of criminal offend- strongest risk factor is a record of physi- portant as sex education for developing ing by Dunedin study members such cally aggressive delinquent offending healthy relationships. In addition, expe- as drug, property, and theft offenses before age 15. However, physically ag- riences in different settings (e.g., at and violence against nonpartners. In gressive delinquent offending before age home and at school) and in different fact, some researchers and law en- 15 is also the most significant risk factor behavioral domains (e.g., academic forcement personnel question whether for victims. achievement and behavior problems) there are any risk factors specific to partner violence, as opposed to crimi- Exhibit 5. Rates of partner violence perpetrated by Dunedin males nal offending in general. This issue convicted of any crime should be resolved before it becomes clear whether a unique theory for part- No Convictions Any Conviction ner violence and specialized interventions for its perpetrators are needed.

In the absence of a knowledge base, popular opinion holds that batterers pose less danger to the general public than other violent offenders because their violence stays within the family.16 18% 38% As such, law enforcement resource n=376 n=104 allocation and judicial decisionmaking Blackened areas show those who perpetrated physical violence against a partner. reflect competition between ensuring the public’s safety from street crimi- nals and ensuring the private safety of Exhibit 6. Rates of partner violence perpetrated by Dunedin males battered wives. By studying associa- convicted of a violent crime tions between partner violence and

No Violent Convictions Convicted of Violent Offense violence against other victims who are not intimate partners, the researchers hope to inform policymakers about whether they should think of most partner violence as a special problem arising from the intimate relationship between two adults or as part of a pattern of repeated aggression toward others by the perpetrator. If the latter is 20% 51% n=442 n=38 true, targeting batterers for priority intervention could improve both spouse Blackened areas show those who perpetrated physical violence against a partner. safety and public safety at once.

8 R e s e a r c h i n B r i e f

The Dunedin study research points to Gender and partner violence: ner done any of them to you?” When strong links between violence against Do men and women hit for the data were analyzed, victimized a partner and a history of violence the same reasons?17 women were 10 times more likely to against other victims. As noted earlier be perpetrators than other women and As shown in exhibit 2, women report in “What are the risk factors in child- male perpetrators also were 19 times perpetrating partner violence more fre- hood and adolescence for partner more likely to be victims than other quently than men. This was true in the abuse?” the strongest predictor of men. The data do not include who Dunedin study and in both the National partner violence among the many risk started each incident or if some of the Family Violence and National Youth factors in childhood and adolescence acts were in self-defense, but it is clear surveys. Exhibit 2 also shows that male in the Dunedin study database is a that in most cases of partner violence victims’ reports corroborate this finding. history of aggressive delinquency be- in this age group, the parties are fore age 15. (Aggressive delinquency Such findings about gender similarities involved in mutual violence. was measured by the study members’ in partner abuse have been contested. Other studies have shown that although self-reports of assaults at age 15, their However, one of the first lessons partner violence behaviors are similar parents’ reports of their aggressive learned in the Dunedin study is that across genders, consequences differ. behavior problems at age 15, and the there are no tidy and distinct groups of Women are much more likely to be presence of “333” incident forms— victims or perpetrators. Interviewers physically injured by men than men are containing charges of assaults by the first asked study members, “Have you to be physically harmed by women.18 study member—in the Youth Aid done any of these things to your part- The Dunedin study findings show that offices of the New Zealand Police). ner?” Next they asked, “Has your part- Researchers checked whether Exhibit 7. Rates of victimization reported by Dunedin women Dunedin study members who were already known to the courts by age 21 21-Year-Old Women With 21-Year-Old Women were also likely to be perpetrators of No Children (n=409) With Children (n=52) partner violence. A search of the New Zealand Police computer files (with each study member’s written permis- sion) revealed that by age 21, 141 study members (14 percent) had been convicted of one or more criminal offenses. Among those convicted, 60 percent were repeat offenders. Al- 26% 53% though most of the convictions were for Blackened areas represent women experiencing physical abuse by their partners. property crimes, 38 men and 8 women had 113 violence convictions for incit- ing violence, manual assault, assault Exhibit 8. Rates of perpetration reported by Dunedin men with a deadly weapon, rape, robbery, homicide, and threatening with an 21-Year-Old Men With No Children (n=454) 21-Year-Old Men With Children (n=25) offensive weapon. Partner violence scores of convicted male study members were compared with those of male study members who were not known to the courts. Exhibits 5 and 6 show that the police and courts already know many of the perpetrators of partner violence because they have been suc- 17% 60% cessfully prosecuted for other crimes. Blackened areas show those men who reported being perpetrators.

9 R e s e a r c h i n B r i e f although women report perpetrating partners, women’s behavior is generally Are young parents more physical violence, the personal charac- not accompanied by multiple problems in likely to be involved in a teristics of male perpetrators are much other areas. The researchers speculate violent relationship than more deviant. Dunedin study male per- that knowledge about the consequences of young adults who have petrators of severe physical violence partner violence might explain this differ- not had children?19 had extreme levels of polydrug abuse, ence. Most men know that if they hit their antisocial personality disorder, drop- partner, she is likely to be injured, the po- One of the most worrisome findings from ping out of school, chronic unemploy- lice may be called, and the police are now the Dunedin study is that young adults ment, poor social support, and violence likely to act swiftly against male perpetra- most likely to be involved in partner vio- against victims outside the family. tors. As a result, young men whose self- lence are also most likely to be parents. Among men who severely assaulted control is compromised by enormous Ten percent, or 52, of the study women their partners, 72 percent had used two social stress, mental illness, or intoxica- had a baby by age 21. Five percent, or 25, or more illicit drugs, 56 percent had left tion will be most likely to risk the conse- of the study men were fathers. Of those secondary school early without any quences of hitting their partner. However, who were parents, 13 percent, or 6, of the formal certificates or qualifications, women know that they are unlikely to in- women and 7 percent, or 2, of the men 51 percent had assaulted someone else jure their partner, he is unlikely to call for were married, although not necessarily to in addition to their partner in the past help, and the police are unlikely to inter- the person who was the mother or father year, and on average they had been un- vene. Thus, there is little to deter an angry of their child(ren). Exhibit 7 shows that employed for 20 months since leaving young woman from hitting her partner. young mothers were twice as likely as school. These extreme social and per- As such, women of all sorts may be apt to other young women to be physically sonal problems were not found for hit their partners, not just women whose abused by their male partners. Exhibit 8 Dunedin study female perpetrators. judgment is clouded by stress, mental indicates that those who were fathers were illness, or intoxication. Further research more than three times as likely as those The Dunedin study findings suggest that should be conducted to confirm this pos- who were not fathers to report being per- although women do report assaulting their sible explanation. petrators of abuse. All of the study women

Full Reports From the Dunedin Multidisciplinary Health and Development Study D etailed reports may be obtained logical Study,” American Journal of Psychia- Magdol, L., T.E. Moffitt, A. Caspi, and P.A. from the authors at the following ad- try 155 (1998): 131–133. (J94) Silva, “Hitting Without a License: Testing dress: Professor Terrie E. Moffitt, Institute Explanations for Differences in Partner of Psychiatry, Social, Genetic and Devel- Krueger, R.F., T.E. Moffitt, A. Caspi, and A. Abuse Between Young Adult Daters and opmental Psychiatry Research Centre, Bleske, “Assortative Mating for Antisocial Cohabitors,” Journal of Marriage and the 111 Denmark Hill, London SE5 8AF Behavior: Developmental and Methodologi- Family 60 (1998): 41–55. (J96) England. Please cite the order letters/ cal Implications,” Behavior Genetics (1998): numbers in bold below. 173–186. (J99) Moffitt, T.E., and A. Caspi, “Annotation: Implications of Violence Between Intimate Bardone, A., T.E. Moffitt, A. Caspi, N. Magdol, L., T.E. Moffitt, A. Caspi, D. Partners for Child Psychologists and Psy- Dickson, and P.A. Silva, “Adult Mental Newman, J. Fagan, and P.A. Silva, “Gender chiatrists,” Journal of Child Psychology and Health and Social Outcomes of Adoles- Differences in Partner Violence in a Birth Psychiatry 39 (1998): 137–144. (J93) cent Girls With Depression and Conduct Cohort of 21-Year-Olds: Bridging the Gap Disorder,” Development & Psychopathol- Between Clinical and Epidemiological Ap- Moffitt, T.E., A. Caspi, R. Krueger, L. ogy 8 (1996): 811–819. (J81) proaches,” Journal of Clinical & Consulting Magdol, G. Margolin, P.A. Silva, and R. Psychology 65 (1997): 68–78. (J83) Sydney, “Do Partners Agree About Abuse in Danielson, K.K., T.E. Moffitt, A. Caspi, Their Relationship? A Psychometric Evalua- and P.A. Silva, “Comorbidity Between Magdol, L., T.E. Moffitt, A. Caspi, and P.A. Silva, tion of Interpartner Agreement,” Psycho- Abuse of an Adult and DSM–III–R Mental “Developmental Antecedents of Partner Violence: logical Assessment 9 (1997): 47–56. (J86) Disorders: Evidence From an Epidemio- A Prospective Longitudinal Study,” Journal of Ab- normal Psychology 107 (1998): 375–389. (J105)

10 R e s e a r c h i n B r i e f were rearing their children. A few of the or therapy requires sound knowledge own partner violence. Thirteen questions about partner violence were asked, e.g., “Have you men were rearing their children, and about how and why individuals become hit your partner?” and “Have you used a knife other abusive study men were living with perpetrators and victims. In the Dunedin or gun on your partner?” Each question in the a new partner and her children. Presum- research, the goal was to improve predic- interview was assumed to have “random flaws” because the respondent’s attention might wan- ably some of those children have been tion, understanding, and treatment of der, they might mis-hear the interviewer, mis- exposed to violence between their parent partner violence by continuing to study understand a word, forget something, circle and partner. This finding underscores the developmental experiences, personal true when they meant to circle false, etc. These the importance of services for high-risk characteristics, and situational circum- random flaws, however, are distinct from delib- erate bias—which is not random—such as ly- adolescents that integrate issues of family stances that lead individuals into partner ing to conceal the violence or exaggerating planning, parenting, and partner violence violence. violence to get attention. to break the cycle of violence transmis- The confirmatory factor analysis program es- sion to the next generation. Notes sentially triangulates among the different an- swers to the 13 questions to come up with a 1. Greenfeld, L.A., M.R. Rand, D. Craven, P.A. best estimate of the perpetrator’s “true score.” Klaus, C.A. Perkins, C. Ringel, G. Warchol, C. Conclusion The same program was then run on the victim’s Maston, and J.A. Fox, Violence by Intimates: reports about the perpetrator’s behavior to get a Analysis of Data on Crimes by Current or The Dunedin study findings to date “true score” for the perpetrator’s abuse from Former Spouses, Boyfriends, and Girlfriends, the victim’s perspective. Finally, the correla- have demonstrated three aspects of Bureau of Justice Statistics Factbook, Wash- tion between the two resulting “true scores” violence between partners. First, ington, DC: U.S. Department of Justice, Bureau was calculated. Before a confirmatory factor of Justice Statistics, March 1998, NCJ 167237. young people who become involved in analysis was used to remove random measure- a violent relationship tend to come 2. The study’s history is detailed in Silva, P.A., ment errors from the perpetrators’ and victims’ from backgrounds that include family and W.R. Stanton, eds., From Child to Adult: scores, they were correlated at 0.59. After the adversity, dropping out of school, and The Dunedin Multidisciplinary Health and scores that had been cleaned of random errors Development Study, Auckland, New Zealand: were used, they were correlated at 0.83. This violent juvenile crime. Second, the Oxford University Press, 1996. difference shows that much of the “disagree- most violent relationships are found ment” between partners that researchers wor- 3. Straus, M.A., “Measuring Intrafamily Conflict ried about is understandable random error. It among young parents of small chil- and Violence: The Conflict Tactics Scales,” Jour- does not indicate that partners are deliberately dren, especially parents who are un- nal of Marriage and the Family 41 (1979): 75–88. trying to misrepresent their violence to researchers. They actually agree very well, if rea- married. Third, partner violence is 4. See Straus, M.A., and R.J. Gelles, Physical complicated by other problem behav- sonable human error is allowed for. This and Violence in American Families, chap. 4, London, the way the program works are explained in iors, especially long-term unemploy- England: Transaction Publishers, 1992, for a dis- detail in Moffitt, T.E., A. Caspi, R. Krueger, cussion of the methodological issues that were L. Magdol, G. Margolin, P.A. Silva, and R. ment, mental illness, drug abuse, and initially raised by the Conflict Tactics Scales. violence against nonfamily victims. Sydney, “Do Partners Agree About Abuse in 5. For the full report, see Moffitt, T.E., A. Their Relationship? These findings demonstrate a need for Caspi, R. Krueger, L. Magdol, G. Margolin, P.A. Silva, and R. Sydney, “Do Partners Agree 7. For the full report, see Magdol, L., T.E. three intervention approaches. First, early About Abuse in Their Relationship? A Psycho- Moffitt, A. Caspi, D. Newman, J. Fagan, and interventions with teenagers are needed metric Evaluation of Interpartner Agreement,” P.A. Silva, “Gender Differences in Partner Psychological Assessment 9 (1997): 47–56. Violence in a Birth Cohort of 21-Year-Olds: to teach them not to use violence against Bridging the Gap Between Clinical and Epide- partners. Second, interventions with 6. For everything that is measured, there is miological Approaches,” Journal of Clinical & young parents are needed to reduce their some slippage between the score the measure- Consulting Psychology 65 (1997): 68–78. ment instrument records and the true score. If stress and protect their small children there were a perfect measurement instrument, 8. Fagan, J., and A. Browne, “Violence Be- from exposure to violence at home. Third, all flawed instruments could be tossed out. But tween Spouses and Intimates,” in Understand- perpetrators of partner violence tend to there is no perfect instrument, so the best strat- ing and Preventing Violence, vol. 3, ed. A.J. Reiss and J.A. Roth, Washington, DC: National be mentally ill and commit other violent egy is to use several different measures of the same data and then triangulate among all their Academy Press, 1994, 115–292. crimes as well, suggesting a need for results to try to find the true score. Statistical coordination among police, judicial, and theory assumes that although every measure is 9. Elliott, D.S., D. Huizinga, and B.J. Morse, The Dynamics of Delinquent Behavior: A Na- psychiatric interventions. flawed, they are each flawed in a different, random way. One score will be too high, another tional Survey Progress Report, Boulder, CO: will be too low, and the true score will lie Institute of Behavioral Sciences, University of These research findings have the poten- somewhere in the middle. Colorado, 1985. tial to inform the work of policymakers The Dunedin study used a statistical procedure 10. For the full report, see Danielson, K.K., and practitioners. Policymaking that de- called “confirmatory factor analysis” to find the T.E. Moffitt, A. Caspi, and P.A. Silva, ters batterers through arrest, prosecution, true score for perpetrators’ self-reports of their “Comorbidity Between Abuse of an Adult and

11 DSM–III–R Mental Disorders: Evidence From an Epidemiological Study,” American Journal Terrie E. Moffitt, Ph.D., is a professor William Freeman Vilas Trust at the of Psychiatry 155 (1998): 131–133. in the department of psychology at the University of Wisconsin. 11. Newman, D., Moffitt, T.E., Caspi, A., and University of Wisconsin-Madison and in Silva, P.A., “Comorbid Mental Disorders: Im- the Social, Genetic and Developmental Full reports cited in the Notes in this plications for Clinical Treatment and Sample Psychiatry Research Centre, Institute of Research in Brief may be obtained from Selection,” Journal of Abnormal Psychology Psychiatry, at the University of London; the authors. See “Full Reports From 107 (1998): 305–311. Avshalom Caspi, Ph.D., is a professor the Dunedin Multidisciplinary Health 12. Severe physical abuse was defined as the in the department of psychology at the and Development Study” for more CTS items likely to cause injury: kicking, bit- University of Wisconsin-Madison and in information. ing, hitting with a fist, hitting with an object, the Social, Genetic and Developmental beating up, choking, strangling, threatening Authors’ acknowledgments: The au- with a knife or gun, or using a knife or gun. Psychiatry Research Centre, Institute of Psychiatry, at the University of London. thors are grateful to Dunedin Unit in- 13. For the full report, see Magdol, L., T.E. vestigators and staff and to the study Moffitt, A. Caspi, and P.A. Silva, “Develop- The Dunedin Multidisciplinary Health members and their partners. They are mental Antecedents of Partner Violence: A and Development Research Unit is sup- Prospective Longitudinal Study,” Journal of also grateful for the assistance of the Abnormal Psychology 107 (1998): 375–389. ported by the New Zealand Health Dunedin Police over the years and for Research Council. This research was 14. Bivariate and multivariate regression were the help and guidance provided by supported by grant number 94–IJ–CX– Police Officer Paul Stevenson, retired. used to test which risk factors predicted self- 0041 from the National Institute of reports of partner abuse, and for study members whose partner participated, the test was repli- Justice. Additional support was pro- Many individuals helped with this cated by predicting the partner’s reports of abuse. vided by U.S. Public Health Service research: Jim Amell, Anna Bardone, (USPHS) grant MH–45070 to Terrie E. 15. For the full report, see Moffitt, T.E., A. Kirstie Danielson, Nigel Dickson, Caspi, and P.A. Silva, in preparation. Moffitt from the Violence and Trau- matic Stress Research Branch of the Waleska Echevarria-Troche, Jeff 16. Fagan, J., and A. Browne, “Violence National Institute of Mental Health, by Fagan, Janet Gafford, HonaLee Between Spouses and Intimates.” USPHS grant MH–49414 to Avshalom Harrington, Bob Krueger, Lynn 17. For the full report, see Magdol, L., T.E. Caspi from the Personality and Social Magdol, Gayla Margolin, Marguerite Moffitt, A. Caspi, D. Newman, J. Fagan, and P.A. Processes Branch of the National Insti- McClelland, Denise Newman, Jay Silva, “Gender Differences in Partner Violence in tute of Mental Health, by the William Rodger, Phil Silva, and Ros Sydney. a Birth Cohort of 21-Year-Olds.” T. Grant Foundation, and by the 18. Fagan, J., and A. Browne, “Violence Between Spouses and Intimates.” Findings and conclusions of the research reported here are those of the authors and do 19. For the full report, see Bardone, A., T.E. The National Institute of Justice is a not necessarily reflect the official position or Moffitt, A. Caspi, N. Dickson, and P.A. Silva, component of the Office of Justice policies of the U.S. Department of Justice. “Adult Mental Health and Social Outcomes of Programs, which also includes the Adolescent Girls With Depression and Conduct Bureau of Justice Assistance, the Bureau This and other NIJ publications can be Disorder,” Development & Psychopathology 8 of Justice Statistics, the Office of Juvenile (1996): 811–819. found at and downloaded from the NIJ Justice and Delinquency Prevention, and Web site (http://www.ojp.usdoj.gov/nij). the Office for Victims of Crime. NCJ 170018

U.S. Department of Justice PRESORTED STANDARD Office of Justice Programs POSTAGE & FEES PAID National Institute of Justice DOJ/NIJ PERMIT NO. G–91 Washington, DC 20531

Official Business Penalty for Private Use $300