Domestic Violence

Domestic violence is constantly in the news these days which can lead to the impression that the problem is increasing. To the extent that scrutiny and public discussion shines light in dark places, we might have expected the real underlying rates to be tapering.

So I was more than surprised when The Age reported figures from Victorian Police Family Violence statistics in the left section of the table below, along with the headline “Family Violence Epidemic.” They specifically highlighted the increases from 2011-2012 to 2012-2013. (Warning; This is a long post so set aside some time!)

table 1

Victorian Totals are taken from Police Crime Statistics.

Even the most unsophisticated punter might question whether family violence homicides more than tripled from 2011-2012 to 2012-2013. Especially when you see the much larger totals of all homicides and rapes in Victoria (extracted from Police crime statistics reports in the right section of the table) were very stable from over the same period, it seems more likely that the inflated counts for family violence in 2012-2013 are artifact.

So how did the 13 to 45 jump happen? Unsurprisingly, it turns out to be partly a reporting error. Homicides in the family violence category mistakenly included culpable driving and several other categories and stripping these out leaves 29. The following day, the same journos did report this as well as the fact that recent reporting practices “now reflected a greater focus on whether family violence played a part in a death”.  So many of the 29 homicides were classified as family homicide in 2012-2013 but may not have been in previous years. The ambiguity is, I guess, in classifying when someone is your partner.

The journalists did not admit that their previous article might be misleading and published the new information under the headline “Number of family-violence related homicides understated: police”. This was not what the police said at all. They are actually over-reported in this case. More important, did the journos admit that actual family homicide rates had not increased three fold? No they did not. It’s not only Current Affair that indulges in fear mongering.

Once a meme like this gets going though, it is hard to kill. Three weeks later the Age reported that a Labor state government would have a royal commission into family violence. The journalist, Farrah Tomazin, mentioned that “Labor’s push to tackle the crisis comes amid a massive increase in domestic violence homicides, with 13 people arrested in 2011-12 rising to 45 last financial year.” Without making a judgement on whether a Royal commission is appropriate, I think we would all hope that Daniel Andrews did not make this commitment on the basis of the “13 to 45” statistic. I emailed Tomazin letting her know that the figures she quoted could not possibly be trusted but received no reply.

There are better ways to look at underlying trends in violence than snooping police reports for the most sensational (and often misleading) proportional change you can find.

There are three quality surveys on violence in the Australian context, all carried out by the ABS. The Women’s safety survey in 1996 (which only covered female victims) followed by the Personal Safety Surveys in 2005 and 2012 which cover male and female victims. Between 15,000 and 18,000 respondents were confidentially interviewed and were asked of their experience of various forms of violence. Obviously, victims of homicide are not covered by such a survey.  Details of the ABS methodology are at the end of this article.

The table below shows the rates of violence for both genders across the three surveys. The “rate of violence” here is the percentage of respondents who report having experienced violence in the last 12 months. This is a standard measure of risk in other sociological contexts and is easier to interpret than lifetime risk which depends on age of respondent.

table 2

Some important information is missing here such as the seriousness of the assault. While many may be at the lower end, reported assaults were only recorded by the interviewer if they were based on “actions which would be considered as offences under State and Territory criminal law”.

The clearest feature of these figures is that the rate of all forms of violence is decreasing (though there are some issues of comparability between the WSS and PSS figures for the threat categories and the smallest figures in the table have a rather high relative sampling error marked with by *).

Bearing in mind the amount of public discussion of the issues of sexual violence over the past twenty years, it would be surprising if sexual assault figures in particular were not falling. So there is some good news right there. I doubt this will make the headlines though. Moreover, keep in mind that personal thresholds of what constitutes physical and sexual assault have been reducing over the same period. So the figures in the table probably underestimate the reduction i.e. respondents in the 2012 survey might be more likely to report sexual assault than respondents in earlier surveys.

The other clear conclusion is that the rates of total and physical violence are higher for males but rates of sexual violence are higher for females. I suspect that few people would be surprised by this. Later on I will look at the perpetrators where there are perhaps a few surprises.

Multiplying the above rates by the Victorian population of 2.88 million males and 2.94 million females, the PSS 2012 figures translate into the following absolute numbers. Again the figures for males sexual crimes marked with an asterisk have very high relative standard error because of the small number of males targeted in the survey and the relative rarity of this crime. Again, recall that these are numbers of people who experienced an incident in the past 12 months.  Below the dark line are the numbers of offenses recorded in Police Statistics for 2012/2013.

table 3

I suggest you spend a little time looking at these figures. The number of reported crimes (below the line) is a tiny fraction of the actual crimes revealed by the PSS (above the line). The charge and report rates appear lower for male victims than for female victims.

I think there are two big lessons from these figures. The public policy lesson is that the vast majority of criminal assaults, both physical and sexual, are not reported to police. They should be. The lesson for researchers and information providers like the MSM is that police reports vastly under-estimate the number of sexual assaults and (non-fatal) physical assaults.

When you ring alarm bells at increased rape statistics in police reports you may be perversely reacting to a better level of law enforcement. When you draw attention to increased family violence charges you may be reacting to improved support services that lead victims to file charges. So I seriously question whether this recent Age Editorial is being responsible in claiming a 14.4% increase in domestic violence on the basis of police figures. It is surely possible to argue for better public policy solutions without suggesting that a problem is getting worse.

The main issue we have not addressed so far is gender of the perpetrator. While police reports give this data, they apparently do not cross tabulate with gender of the victim. Again, the PSS surveys have quite detailed information on this critical dimension of the issue. The figures below apply to all physical assaults which, as we have seen in the previous table, would number around a quarter of a million in Victoria.

table 4

Clearly the vast majority of perpetrators are male though for female victims around 25% of perpetrators are female. Unfortunately, the PSS 2012 report does not seem to include this level of detail. In order to compare the 2005 and 2012 figures, I have combined some of the perpetrator categories in the table below. It should be noted that the percentages are now of all assaults whereas those in the previous table were for male and female victims separately (so columns do not total 100%).

table 5

Looking at these figures it is hard to see why, in an interview about one-punch assaults, Leigh Sales would have asked Barry O’Farrell this loaded question:

Given that random violence is so much less of a threat to people than domestic violence, why is the government’s priority not on domestic violence?

If stranger violence is random then the table shows the conditional is false. Even reported crime statistics contradict it (see the last comment at the above link). It’s not too much to ask the ABC to do a little fact checking.

Anyway, there are several aspects of the numbers in the above table that might defy common folklore.

  1. Male victims out-number females roughly 2 to 1. While some of these assaults might well be at the less serious end of the spectrum, male victims of homicide consistently outnumber female victims by about the same factor (2:1) which suggests otherwise.
  2. The vast majority of assaults on female victims are not perpetrated by a partner or previous partner, in other words most violence against females does not follow the domestic stereotype.
  3. In 2012, physical assault by male partners/former partners on women is a factor 9.6/3.8=2.56 higher than by female partners/former partners on men.
  4. Assaults on males by their female partners have quadrupled from 2005 to 2012!

This last point refers to a comparison of the 0.6% and 2.4% figures.  I cannot discover an explanation for the 2.4% figure in the report. A plausible partial explanation could be a change in how victims chose to classify their relationship with the perpetrator. Girlfriends might be classified as partners or not. Abusive partners might be classified as partners or previous partners. I find it very hard to accept that assaults by female partners on males have really quadrupled.

Nevertheless, those with a particular viewpoint to push might rush to make the sensational claim in (4) above, not completely different to those who made the sensational claim that domestic homicides in Victoria had tripled. I, for one, would not make either claim.

Looking at point 3, it appears that 28% of violent assaults by partners are committed by the female (up from 22% in 2005). The level of injury would typically be less but this is irrelevant to the cause.  For these domestic assaults at least, male attitudes of entitlement cannot be the cause. There must be other factors at play in this 28% – like the stress of a long term dysfunctional relationship, generally poor self-control, effects of drugs and alcohol not to mention opportunity from the time partners spend together. If there are the same numbers of male perpetrators who succumb to the same causes, then many of the 72% are immediately caused by something other than male attitudes of entitlement. Why would we ignore this?

Of course, it is almost a tautology to say that many men who assault their partners will have (at least at the moment of the offense) a lack of respect for her autonomous rights. Some may well generalise this to all times and all women. But this is like saying that those who commit thefts lack respect for private property, or that criminals in general lack a sense of civic duty and work ethic. It’s so obviously true that it’s fatuous. Is a renewed emphasis on the protestant work ethic in school really going to banish criminality? Unlike the causes of burglary however, many bloggers seem to assume that males attitudes are not only the main causative factor of DV but that the same factors are carried by all men – you know the 99% who do not assault their partners. This distracts attention from policy options targeting likely offenders, such as early IVO’s.

The final issue I will briefly mention is alcohol and drugs and their association with domestic violence. There are some commentators and researchers in this area who would seek to minimise their role. The table below is based on Figure 10.4 in the 2011-2012 Police report. The total number of reported family violence incidents was 50382 and for 18573 of these a charge was laid. Clearly alcohol and drugs are heavily involved. Nationwide, between 2000 and 2006, 44 percent of intimate-partner homicides were alcohol related.

Of course, in many cases alcohol and drugs will only be the proximate cause – the trigger – rather than the underlying cause. To ignore the interaction between alcohol and other causes is to ignore an observable cause which could potentially be managed. It is as dumb as arguing that guns have no role in US crime. That is not to make it an excuse, let alone a defense, and if any perpetrator thinks they can claim diminished responsibility for DV or any other crime on the basis of alcohol impairment they will find they are out of luck.

figure 1

So where does this leave us?

I have not suggested any policies for reducing DV because it is not my area of expertise. What I can conclude is that we should not be rushing to new policy initiatives on the basis that family violence is increasing. It isn’t. We should not be focusing solely on general male attitudes to women at the expense of other claimed risk factors such as alcohol abuse, economic stress and mental health.

We should always keep in mind that public policy involves trade-offs. Focus more on say traffic violations and you will likely focus less on say the “ice epidemic”. For instance the Age recently reported

The latest figures released by police show a spike in drug crimes (up 42.7 per cent), car theft (up 39 per cent) and property crimes (up 17 per cent) …, as domestic violence, alcohol-fueled violence and traffic offences take priority.

Since the PSS data suggest that domestic violence is a modest proportion of all violence, should we be focusing on domestic violence at all? Not on the basis of its relative prevalence, distorted perceptions of which seem to be driving public debate and maybe even policy.

The key issue should always be where we think interventions will be most effective. To the extent that domestic violence is often the end of a long observable process, and those at risk are often well known to community services, I think that intervention in the domestic violence process may well be more effective than other crimes. But I would like to see the evidence.

Unfortunately, the Royal Commission will not tell us this because it will not be asked to make the trade-off between DV and other crime prevention options. That’s the problem with single issue politically motivated inquiries.


Details on PSS 2012 survey. Individuals over the age of 18 were sampled at the household level. There were 41350 private dwellings approached with pre-assigned intention of a female respondent for 31650 of these. Response rates for males and females were almost identical (57%). Responses were obtained by specially trained interviewers, in a private setting, without the knowledge of other household members. The definitions of violence used were based on actions which would be considered as offences under State and Territory criminal law. Appropriate wording was based on a survey advisory group including experts with legal and crime research background.








This entry was posted in Economics and public policy, Gender, Politics - national, Uncategorized. Bookmark the permalink.
Notify of

Newest Most Voted
Inline Feedbacks
View all comments
9 years ago

Thanks, that was interesting — if the interview figures are correct, then perhaps the royal commission should be into why there is so much violence in general in the Aus community.

Also, you state:
The public policy lesson is that the vast majority of criminal assaults, both physical and sexual, are not reported to police. They should be.

The obvious reason why people often (indeed generally) don’t, of which the Man Honis case was an especially good example, is that most people expect the police will do more or less nothing tangible even if they do, and most of the time, they will be correct in this assumption.

If the figures above are correct, the reason the police/justice system do very little is also obvious — if everyone was reported, there wouldn’t be enough space in our jails to put the bad offendors or enough police to do anything about violated AVOs etc. . So not much gets done because that is the most you can do in the short term.

Perhaps this type of data should give people some idea about the trade-offs we make about who goes to jail. At present, it appears we have any number of violent offendors wandering around getting away with it, yet drug-offendors, who mainly do harm to themselves or willing others, are filling up our already full jails (Aus has a relatively high rate of incarceration). I know who I’d rather bump into.

9 years ago
Reply to  conrad

Yes the war on drugs is closer to a crime against humanity (particularly in its indirect effects!) than a crime prevention program.

derrida derider
derrida derider
9 years ago

4.Assaults on males by their female partners have quadrupled from 2005 to 2012!
This last point refers to a comparison of the 0.6% and 2.4% figures. I cannot discover an explanation for the 2.4% figure in the report

Old hands always mistrust movements in small proportions between sample surveys. The cell sizes involved mean this “quadrupling” could easily be statistical noise – did you look at your SEs? Even if it isn’t pure noise it can take only a minor change in such things as the wording of the question or the sample selection to generate a 1.8 percentage point movement.

Moz in Oz
Moz in Oz
9 years ago

with 15,000-18,000 people who can only choose from two genders the quantum is about 1/9000, or 0.01%. So the change from 0.6% to 2.4% represents 180 extra reports. Not a small number, but with 9000 from 20M your 95% confidence level is about 1% – so that 1.8% swing is likely to be significant.

It could also very easily be a classification change – we have seen an awful lot of anti-assault campaigns that focus on persuading people to classify things as assault (I agree with the campaigns, but this could simply show that they’re effective). The swing from “getting someone drunk so they’ll have sex with you” is a normal, accepted part of life to it’s unquestionably rape is still in progress (for example). I suspect “she slapped me” is making the same transition, at least I’m seeing less “friendly” female-on-male violence over the last few years.

Chris Lloyd
Chris Lloyd
9 years ago

Indeed that was the point of my following commentary. I do not for a minute believe there has been a four fold increase. It will be a combination of statistical noise and flexible definitions.

Chris Lloyd
Chris Lloyd
9 years ago

Quite right DD: which is why I said that I would not make any strong claim about trend based on these figures. The relative SEs are indeed pretty high.

David Walker
9 years ago

Terrific post, Chris

John walker
9 years ago
Reply to  David Walker

Agreed terrific post!
Can only add the following (anecdotal )comment- going off my friends and contacts in the Canberra Goulburn Anglican church and related community groups- the levels of violence are in some places really growing. And in particular The figures suggesting a possibly growing randomness , big mental health problems, and links to drug use (above all to the use of ice ) all ring as very true, but I admit this is anecdotal .

Chris Lloyd
Chris Lloyd
9 years ago
Reply to  John walker

Well the increase has not appeared in the PSS yet (there are state breakdowns as well). Interesting anecdotal evidence though and nothing wrong with using anecdotes to generate hypotheses. Market researchers call them ‘focus groups’!

derrida derider
derrida derider
9 years ago
Reply to  John walker

John, Canberra used to have a serious heroin problem which when the local wholesaler was busted (its a small city) morphed into a serious meth problem. Amphetamines breed a lot more violence than opiates; I’d be cautious about generalising beyond Canberra.

John walker
9 years ago

DD while I agree re generalising, the Canberra/ Goulburn diossian region covers a lot of country , not just Canberra – I fact far from it.

derrida derider
derrida derider
9 years ago

Oh, one more thing – despite my quibble above I agree this was an interesting and enlightening post.

Nicholas Gruen
9 years ago

Indeed – thanks for the post Chris.

(Which reminds me, the number of interesting posts on Troppo in the last two weeks has quadrupled which is a very interesting trend ;)

Jasmin Newman
9 years ago

As a mere mortal ( and not great with maths) trying to interpret similar data from ABS and AIC I find this somewhat of a relief.

You have exposed the ineptitude of both the media in reporting and the government in providing readily accessible real data.

How can we prevent anything in society when we are not aware of the real causes (and effects)

Thank you!!