Bruce Headey
University of Melbourne
Dorothy Scott
University of Melbourne
David de Vaus
La Trobe University

Conventional wisdom holds (i) that physical domestic violence is mainly perpetrated by men against women; (ii) that violent men, being physically stronger, inflict more pain and serious injuries than violent women; and (iii) that physical violence runs in families. To examine all three beliefs, we bring to bear nationwide sample survey data.

In investigating domestic violence, three different types of data have been used, each with limitations, each leading to somewhat different accounts. First, crime statistics focus on the extreme end of the spectrum: homicides committed by a husband, wife, or lover. There, the preponderance of male perpetrators is clear. In Australia, 3.6 times as many women as men are killed by their partners (James and Carcach 1997). The same pattern holds in North America, although the gender difference is smaller (Straus 1986).

The second major source of data on domestic violence is clinical studies. In Australia these feature medical settings and mostly women patients (for example Webster, Sweett & Stolz, 1994; Mazza, Dennerstein & Ryan, 1996). Among injury presentations positively identified as domestic violence in a large, recent study of five Victorian hospitals, women outnumbered men by nearly 5 to 1 (Monash University Accident Research Centre, 1994). However the disproportion in serious injuries was less extreme, with 24% of the men and 13% of the women requiring hospital admission. Issues of labelling, misreporting by patients, and selectivity in willingness to seek help make it difficult to generalise from clinical studies.

The third major source of data on domestic violence is sample surveys. They have the advantage of covering the full range of domestic violence, not just the extremes revealed in homicide statistics or clinical studies. One important limitation is that they may under-report extreme violence(1), and previous studies in Australia also suffer from having just sampled women(2), even though two National Family Violence Surveys in the United States a decade ago showed no significant difference between physical assault rates experienced by male and female partners (Gelles & Straus 1988; Straus & Gelles 1986, 1990)(3).

Is women’s violence towards men best understood as self-defence? Conventional wisdom might say yes. But, reflecting on US studies, Straus (1993) concludes that "research on who hit first does not support the hypothesis that assaults by wives are primarily acts of self-defence or retaliation".(4)

Most have concluded that domestic violence is intergenerational (Chappell & Heiner 1990; Straus, Gelles and Steinmetz 1980).(5) Many argue that women who witnessed violence by their fathers against their mothers became more likely than other women to expect or tolerate it in their own marriages (Gelles 1987).

The hypotheses we test are thus based on ‘majority’ professional and public opinion: (6)


The International Social Science Survey/ Australia 1996/97

The "Family Interaction" module was developed to examine patterns of parental interaction, including domestic violence, and their intergenerational impact (Headey, Funder, Scott, Kelley, and Evans 1996). In the sample of 2151, 1643 respondents (804 men and 839 women) had been partnered during the last year, and hence were asked questions about domestic violence. Only one respondent was obtained from each household, so the male and female respondents, while both representative sub-samples, did not include people married or partnered to each other. For details on the IsssA surveys see page XXX and Kelley and Evans (1999).


In order to get an accurate measure of the annual incidence of physical domestic violence, we asked about the frequency of various kinds of assault (see below):

Never in the last 12 months

Once in the last 12 months

Twice in the last 12 months

Three to five times in the last 12 months

Six or more times in the last 12 months

So few respondents had committed (or suffered) multiple assaults, that we dichotomized the scale into "no assault " versus "any assault".(7)

Domestic violence -- both given and received -- is best assessed by asking about specific acts. To allow respondents to feel more comfortable reporting negative things about themselves and their spouses, we first asked about a series of positive acts (for example, "...bought a present for your husband or wife? Did they buy one for you?").(8) Then, on domestic violence, we asked:

  I did Spouse did
j. You slap, shake or scratch them? They do to you? No 1 2 4 6+ No 1 2 4 6+
k. Hit with fist or with something held in the hand, or thrown -- you do it? They do it? No 1 2 4 6+ No 1 2 4 6+
l. Kicked? No 1 2 4 6+ No 1 2 4 6+

Directly following these questions, respondents were asked whether "On any of these occasions, did you injure them -- so that they needed first aid? They injure you?" Next, they were asked whether "they needed treatment by a doctor or nurse? They injure you that much?" Finally, respondents were asked whether they had reported an assault to the police or other government authorities. Levels of missing data on these questions were at about the normal level for the survey as a whole, indicating that people were not especially reluctant to answer them."

To discover whether results from our female respondents parallel those from the Australian Bureau of Statistics, much later in the questionnaire we replicated two of their questions (ABS 1996). Because the ABS is well known for high response rates and good samples, a close match between their results and ours would enhance confidence in the data provided by both our female and male respondents.

Finally, we asked respondents about violence by their parents towards each other. They were asked whether, when they were around 10 or 11 years old, their mother had ‘slapped or scratched their father in anger’, and whether their father had done these things to their mother. A subsequent question asked about ‘hitting with the fist or something held in the hand, or thrown’. By asking about violence at a specific time in children’s lives, we hoped to make a rough estimate of the prevalence of violence in the previous generation, as well as to assess the extent to which domestic violence runs in families. Clearly, however, respondents’ memories of their parents’ behaviour would not be completely accurate.


Male and Female Assault Rates

Our first hypothesis is that men are significantly more likely to physically assault their partners than vice-versa. Table 1 gives the percentage of respondents who report that they were assaulted in each of several ways in the last twelve months.

Men and women report approximately equal rates of being assaulted by their partner, for all three types of assault we asked about. These results are in line with American data, which also show no significant differences.(9)

Moreover, the summary measure of experiencing any of these forms of assault also fails to reveal a preponderance of assaults on women: 4.7% of the sample reported being assaulted in some way during the last 12 months; 5.7% of men and 3.7% of women. This remains an unacceptably high rate of domestic violence, although it is not quite the "War on Women" referred to in the media (e.g. The Age, June 4, 1993).

In addition to asking about actual violence, the survey also asked about threats and feelings of intimidation. Similar percentages of men and women—5.7% and 6.0%—reported that their partner had threatened "to slap, hit or attack" them, but more women (7.6%) than men (4.0%) said they felt "frightened and intimidated". This latter difference was significant at the .05 level, and indeed was the only statistically significant gender difference in domestic violence.

Table 1: Percent Assaulted By Their Partnera: Self-Reports
by Men (N=804) and Women (N=839) on Assaults In Last 12 Months.
Australia, 1996/97.

% Experiencing this type of assault:
Type of Assault Men Women Gender difference significant?b
Slap, shake or scratch 5.1 3.2 no
Hit with fist, or with something held in hand or thrown 4.1 2.5 no
Kicked 2.1 1.4 no
Any physical assault? (victim of one or more types of assault shown above) 5.7 3.7 no

a Includes husbands, wives, and de facto spouses
b Significance at .05 level of phi (2 x 2 table).
Source: IsssA 1996/97.

Violence Runs in Couples

An important but unanticipated finding was that violence runs in couples. 54% of respondents who reported that they had been assaulted, also admitted that they had assaulted their partners.

An important point is that the couples who assault each other are the second largest group (2.5% of the sample). Rather small minorities claim to be assaulted without striking back (2.1%), or admit being violent while reporting that their partner does not hit back (1.0%).

Do Men’s and Women’s Responses Corroborate Each Other?

Many of the results reported here run counter to conventional wisdom, so it is especially important to establish that the estimates are accurate.

First, how do the IsssA results compare to women’s reports in previous violence surveys? In Table 2 we juxtapose women’s answers to two questions in the Australian Women’s Safety Survey (1996) with their replications in the IsssA data. The ABS is well known for the high quality of its samples and high response rates, so this survey provides a convenient benchmark.


Table 2. Women’s Reports of Domestic Violence in Two Surveys: Percent.
Australia 1996-1997.

  IsssA 1996/97
Australian Women’s Safety
Survey (N=6,300),1996
Hit or physically attacked by your husband or de facto in last 12 months 1.6 2.2b
Have you ever been hit or physically attacked by your husband or de facto 8.6 7.2c

a These questions were only put to currently married or de facto respondents in our survey. Therefore comparisons with the ABS survey relate only to assaults by current partners. Threats of assault are not included.
b Calculated from Australian Bureau of Statistics, Women’s Safety Australia (1996) p.9,19.
c Australian Bureau of Statistics, Australian Women’s Year Book (1997) pp. 159-60.

The two surveys yield very similar results. The ABS survey gives a slightly higher estimate of violence "in the last 12 months", while the IsssA data provides a slightly higher estimate of "ever" having experienced violence. This close match is not surprising, since prior research has found that the IsssA samples closely match ABS population estimates on objective characteristics (Sikora 1997).

A second approach (see Table 3) involves comparing men’s and women’s responses within the IsssA.(10) If the women’s reports are to be believed (as they have been in previous work), and if men and women report similar rates of violence, then the men’s reports probably have to be believed too. This is quite a stringent test since some previous research on sexual behaviour has found large discrepancies between men’s and women’s reports (Lewontin 1995). The IsssA male and female respondents were not married to each other, but they are both nationally representative sub-samples. So if both men and women are being truthful, the percent of women who say they were assaulted should agree with the percent of men who admit assault (within the limits of sampling error).

Table 3. Do Men and Women’s Reports Agree in the IsssA 1996/97?

  Women who say they were assaulted
Men who admit assault
Sig. At .05 level?

Men who say they were assaulted
Women who admit assault
Sig. At .05 level?

Slap, shake or scratch 3.2 3.1 no 5.1 2.7 Yes
Slap, shake or scratch 3.2 3.1 no 5.1 2.7 yes
Hit with first, or with something held in hard, or thrown 2.5 2.2 no 4.1 2.8 no
Kicked 1.4 1.6 no 2.1 1.3 no
Any physical assault? 3.7 3.4 no 5.7 3.6 yes

For assaults on women, there is virtual agreement between the sexes: almost exactly the same percentage of men admit assault as there are women who report being assaulted. 3.4% of men admit violence and 3.7% of women say they were assaulted. Statistical tests confirm that the difference is not significant.

For assaults on men, the results are quite different. More men claim to be assaulted, than there are women who admit assault. In terms of overall numbers (row 4 of Table 3) the difference is just statistically significant at the .05 level. It is hard to interpret the finding.(11)

The key finding here is that men’s and women’s reports corroborate each other in the case of assaults on women, and thus partially validate each other. That encourages confidence in the truthfulness and accuracy of the responses.

Injury, Pain and Reporting Violence to Authorities

Our second hypothesis is that male assailants inflict more serious injuries than female assailants. We strove to avoid the issue of gender differences in willingness to seek help by focusing on the occurrence of injury and pain rather than on the action respondent took about them.

Table 4. Injury and Pain Due to Assaults (N=1643). Australia 1996/97. a

% Victims
Men Women Sig. at the .05 level?b
Injured, needed first aid 1.8 1.2 no
Needed treatment by a doctor or nurse 1.5 1.1 no
Pain as bad as hitting thumb with a hammer, or worse 1.9 1.9 no
Called the police or other government authority 1.3 1.7 no

a All results in Table 4 are reports by ‘victims’ of assault.
b Significance of phi (2 x 2 table).
Source: IsssA 1996/97.

Men are at least as likely as women to be victims of domestic assaults that lead to injury and pain (Table 4). Consistent with victimization rates (Table 1), the results here suggest that women inflict serious injuries at least as frequently as men. The evidence in Table 4 needs treating with caution because it runs counter, not just to conventional belief, but also to medical and police records. Clearly, established beliefs cannot be overturned by one set of findings. These issues need further research.

Is domestic violence intergenerational?

The explanation most frequently offered for domestic violence is that men who had violent fathers are violent towards their own wives. And some would argue that daughters of violent parents are likely to be in violent relationships themselves.

Table 5. Relationships Between Parental Violence and Own Domestic Violence, Australia 1996/97: Pearson Correlations

Male Female
Father Violenta Mother Violenta Father Violenta Mother Violenta
Respondent Violent .13** .11** .06 ns .08*
Respondent Victim of Violence .09* ..08* .10** .05 ns

ns = not significant at the .05 level. * p<.05. **p<.01
a A parent who had either slapped, scratched or shaken the other parent, or hit him/her when the son/daughter was around 10 or 11 years old is classified as violent. Retrospective report by son/daughter.

Source: IsssA 1996/97.

Intergenerational transmission is weak: All the Pearson correlations in Table 5 are quite small (the largest is 0.13).(12) The vast majority of people who had violent parents do not assault their own partners. Also, the vast majority of those who admit being violent do not claim to have had violent parents.

Nonetheless, some intergenerational transmission occurs. Men who were the sons of violent fathers were significantly more likely than the sons of non-violent fathers to report being violent themselves, or to being victims of violence, or both (Table 5). Of the men who had violent fathers, 9.8% were violent themselves. By contrast the rate of physical domestic violence was only 2.5% among men with non-violent fathers. In the case of women, the findings were more complex, perhaps resulting from modeling. Women with violent mothers were somewhat more likely to be violent themselves, and women were more likely to be victims of violence if they had violent fathers.

Thus the data provide evidence of weak intergenerational transmission of domestic violence, and thus modest support for ‘family systems theory’.

We noted earlier that marriages in which the partners both hit each other appear to be more common than marriages with one violent and one non-violent partner. Respondents’ reports of their parents’ marriages tend in the same direction, although less strongly. 85.8% reported no violence, while 5.4% reported that their parents assaulted each other, 5.9% reported that the father assaulted the mother without being struck back, while the remaining 2.9% said that the mother was the only violent partner.


To sum up:

    1. Men were just as likely to report being physically assaulted by their partners as women. Further, women and men were about equally likely to admit being violent themselves.
    2. Men and women report experiencing about the same levels of pain and need for medical attention resulting from domestic violence.
    3. Violence runs in couples. In over 50% of partnerships in which violence occurred both partners struck each other.
    4. People who had violent parents were significantly more likely than others to be violent to their own partners and to be victims of violence themselves. On the other hand, a huge majority of people whose parents were violent do not assault their own partners. Moreover, the vast majority of those who are violent did not have violent parents.

The first two results run counter to conventional wisdom and to the hypotheses with which we began the paper. However, some degree of confirmation or at least plausibility derives from the fact that men’s and women’s reports on rates of domestic violence more or less agree. If the women are to be believed (as they have been by previous investigators), then so are the men. Further, the results relating to women being as violent as men are in line with some recent American research.

Of course it takes more than one survey to overturn received wisdom. It is fair to ask researchers how much confidence they have in their own findings. We are reasonably confident about the first and third results; that female and male partners assault each other about equally often and that violence runs in couples. Nor do we have reason to doubt that the offspring of violent parents are unlikely to be violent themselves, albeit at greater risk of being violent than are the children of non-violent parents. We have much less confidence in the second result, finding it hard to credit that women injure men as seriously as men injure women. We hope that our measures of the severity of injury and pain were a reasonable first attempt. Nevertheless, in future work it will be important to compare subjective assessments of severity to more reliable and objective measures.


Age, The (1993) ‘War on women’, June 4.

Australian Bureau of Statistics. 1997. Australian Women’s Year Book, Canberra: ABS

Australian Bureau of Statistics. 1996. Women’s Safety Australia, Canberra: ABS

Chappell, C. & Heiner, R. (1990) "The intergenerational transmission of family aggression" Journal of Family Violence 5(2):135-52.

Gelles, R. 1987. Family Violence. Beverly Hills: Sage.

Gelles, R. & Straus, M. 1988. Intimate Violence: The Causes and Consequences of Abuse in the American Family. New York: Simon & Schuster.

Headey, B. K. Funder, D. Scott, J. Kelley, and M.D.R. Evans. 1996. "Family Interaction Module" in J. Kelley and M.D.R. Evans (eds.) IsssA: International Social Science Survey/ Australia 1996/97: Questionnaire. Canberra and Melbourne: Australian National University and University of Melbourne.

James, M. & Carcach, C. .1997. Homicide in Australia 1989-96, Australian Institute of Criminology Research and Public Policy Series No.13.

Jurik, N. & Gregware, P. 1989. A Method for Murder: An Interactionist Analysis of Homicides by Women. Tempe: Arizona State University, School of Justice Studies.

Lewontin, R. C. 1995. "Sex, lies and social science." (review of R. T. Michael, E. O. Laumann & G. Kolata: Sex In America: A Definitive Survey. Boston: Little Brown) The New York Review of Books, April 20 1995

Kelley, J. and M.D.R. Evans 1999. "Australian and International Survey Data for Multivariate Analysis: The IsssA." Australian Economic Review 32 (September): In press.

Mazza, D., Dennerstein,L. & Ryan, V. 1996. "Physical, sexual and emotional violence against women: a general practice-based prevalence study." Medical Journal of Australia, 164: 14-17.

Monash University Accident Research Centre (1994). Domestic Violence: Patterns and Indicators, Report No.63.

Sikora, J. 1997. "International Survey of Economic Attitudes in Australia, Finland and Poland: Comparison with the Census." WwA: Worldwide Attitudes 1997-12-27:1-12.

Straus, M. 1986. "Domestic violence and homicide antecedents" Domestic Violence 62: 446-465.

Straus, M. 1993. "Physical assault by wives, a major social problem" in R.Gelles & D.Loseke (eds) Current Controversies on Family Violence. Newbury Park: Sage Publications.

Straus, M., Gelles, R. & Steinmetz, S. 1980. Behind Closed Doors: Violence in the American Family, Garden City, NY: Anchor Press/Doubleday.

Straus, M. & Gelles, R. (Eds) 1990. Physical Violence in Amercian Families: Risk Factors and Adaptations to Violence in 8,145 Families, New Brunswick, NJ: Transaction.

Webster, J., Sweett, S. & Stolz, T. 1994. "Domestic violence in pregnancy." Medical Journal of Australia, 161, 466-470.


* Dr Kathleen Funder of the Australian Institute of Family Studies was an original participant in the study but died before analysis was complete. This paper is respectfully dedicated to her memory, although it should not be assumed that she would have entirely agreed with its conclusions. For over two decades she did superb research on families and divorce. Many thanks to Dr Jonathan Kelley and Dr Mariah Evans of the IsssA (International Social Science Survey/ Australia) who invited us to contribute questions to their survey and gave excellent advice on design issues.

1 Some victims of severe violence are in refuges and so not available to surveys; perpetrators and victims of severe violence may also be less willing to admit what is going on than are people in milder situations.

2 The most recent example is a study, commissioned by The Office of the Status of Women and conducted by the Australian Bureau of Statistics, which found that 2.2% of women reported being “hit or physically attacked” by a male partner in the previous year (ABS 1996).

3 This pattern remained when, to allow for possible male underreporting, only the reports of women were used (Straus 1993)

4 He argues that this is true even with the unlikely assumption that in all cases of mutual violence women were acting in self-defence. When women kill their husbands, a common feminist response is that this reflects “the battered woman syndrome” -- desperate acts of self-defence by women subjected to a long history of abuse. However, North American data suggest such abuse, or threats of it, is present in only 21% of these cases (Jurik and Gregware 1989).

5 For example, for the U.S.: "Men who had seen parents physically attack each other were almost three times more likely to hit their own wives ... Women whose parents were violent had a much higher rate of hitting their own husbands as compared with daughters of non-violent parents ... In fact, the sons of the most violent partners have a rate of wife-beating 1000 per cent greater than those of the sons of non-violent parents. The daughters of violent parents have a husband-beating rate that is 600 per cent greater than the daughters who grew up in non-violent households." (Straus, Gelles and Steinmetz 1980)

6 It must be emphasized that this paper deals only with physical violence and not other forms of domestic violence.

7 This decision did not affect estimates of assault rates by gender

8 Other positive actions asked about were to suggest a romantic evening together; organise a party for a bitherday or special occasion; and suggest a day at the beach or bushwalking.

9 Although men report slightly higher rates of being assaulted than women, the difference is not statistically significant.

10 Ultimately, confirmation or refutation of our results will depend on replication, preferably using other methods as well as surveys.

11 There are at least three possibilities. (1) Women won’t admit to committing assaults even though they do. (2) Women do not define minor attacks on men as assaults especially perhaps if they are retaliating, although the men they attack do feel assaulted. (3) Men overstate the number of assaults made on them. With the present data it is not possible to assess the relative importance of these explanations.

12 It is possible that they are biased downwards by the extreme non-normality of the distributions, but the crucial points that the vast majority of people from violent families are nonviolent and that the vast majority of violent people are from non-violent families are independent of this bias.