22 June 09



It will be nearly impossible to know whether any changes that occurred shortly after 1997 were caused by the buyback (reduction of firearm prevalence) on the one hand or by one part or another of the various laws and regulations (not limited to gun policies) passed as part of the overall program.  This is because all those things began at virtually the same time throughout the country as a consequence of instituting a comprehensive nationwide program.  If one wants to evaluate the effectiveness of a policy change, it is almost imperative to make no other policy change at or near the same time that might also impact what one wishes to impact with the subject policy change.

The effects of the new laws and ban/buyback on crime and suicide rates cannot be certain for another reason.  It is entirely likely that any changes observed in rate trends could be results of things unrelated to the laws and ban/buyback—and there is no shortage of things that are widely known to affect suicide and crime rates.  At least the most significant of such things would have to be identified and accounted for in order to separate their effects from the effects of the laws and ban/buyback.  Still, we can check for presence of a possible impact.

Because of these problems, it would not be legitimate to suggest that the ban or buyback (or other changes that occurred at about the same time) may have had an impact unless the impact is observable immediately after the ban/buyback, starting in either '97 or '98.  The change we must check for in the case of each statistic of interest is a step change in the value of the statistic (for example, a rate going up more than the usual year-to-year variation), or a change in the rate at which the statistic changes from year to year.


Before the new laws/ban/buyback began, estimates of the number of firearms in Australia were around 3.5 to 4 million, based on who-knows-what information available to police officials (like numbers of registered guns).  The "gun lobby" estimated much higher, and had estimates from importers, etc. of hundreds of thousands each just for some individual types of firearms (like SKS variants, military surplus and Ruger 10/22s).  It is apparent that the stock of firearms was much higher than authorities thought.

At the end of the buyback (31 Aug 98), Australia had paid $320M (Australian) for 643,726 firearms—only about 1/3 of the 1.5 million authorities expected to collect of their estimated 3.5 million total stock of firearms (which was only a small part of the actual stock).  Don't forget the additional administrative costs of running the program.  There were also some reported irregularities—like officials being cooperative in identifying firearms as prohibited ones when they weren't, and police departments selling their old firearms even multiple times.

Many nonprohibited firearms were turned in (supposedly not bought or included in the 643,726 tally) but these numbers were not nearly as high as the numbers of prohibited firearms, and were also insignificant in comparison to the total firearm stock.  Many people used the buyback as an opportunity to get paid for some essentially worthless firearms.

Although most jurisdictions released no details, Victoria did; their proportions should be close to typical.  Although "military style" centerfire semiauto rifles were the firearms that were supposedly of primary concern to the gun control advocates, those types having been used in the massacres, only 3.2% of the firearms Victoria bought were centerfire (probably semiautos), other than shotguns.  47.5% were .22 rimfire rifles ("pea shooters") and 47.8% were pump or semiauto shotguns.

Gun controllers and bureaucrats proclaimed the program a great success simply because some firearms were taken out of circulation.  They publicly proclaimed that compliance was high because the commonwealth had used the peoples' money to buy a telephone survey that concluded that compliance was almost universal.  The only problem was that the survey was only of people who had registered a gun in the past.  Can you imagine the stupidity of people who would think that a person that had registered a gun with "the government" might tell someone doing a government survey that he or she had surrendered a gun when he/she hadn't really done so?  Naturally, people who had registered a firearm in the past will almost certainly surrender at least the registered firearms if authorities order such surrender.  To do otherwise would invite arrest.

During the year ending 30 June 97, more than 83,000 rifles and shotguns were legally imported.  Between 1 July 97 and Mar 98 (midway through the buyback) 20,493 were imported.  Hence, approximately 100,000 of the 643,726 that were surrendered in the buyback were offset during almost the same period by people buying new guns.  [Australia had essentially zero domestic production of firearms, so all new firearms bought had to be through legal or illegal importation.]

Victoria had another "amnesty" from August '98 through February '99 for people to turn in their guns.  Western Australia has had an unlimited (open) amnesty, and gun controllers were apparently changing their position in August '99 to support one for NSW.


History of Accidental Gun Deaths
Fig. 1. History of Gun Accident Deaths in Australia

One would expect that accidental deaths would fall after reduction of firearm prevalence, although there is a very small degree of substitution of method for accidental deaths (because accidents largely happen to people who are accident prone).  The rates had somewhat leveled off just prior to '96 (3 years).  Surprisingly, the rates headed upward for the several years since 1996.  It was too soon for accidents to start rising from people not getting safety training because they weren't supposed to have guns.  Did a lot of untrained people quickly acquire unregistered guns illegally while they still could, in anticipation of the time when they wouldn't be able to?  Did hiding guns lead to accidents?

It is curious how the rate jumped in 2000, then fell dramatically the next year, followed by what appears to be a trend upward from the 2001 level.  It looks a bit like there was one trend from '97 to 2000, then an entirely different trend starting in '01.  Did accident rates really do this, or did something change in the reporting of accidental deaths?  It has long been known that some suicides are reported as accidents because of the stigma associated with suicide.  A small change in that factor could cause a large change in accident reporting.  And, it is possible that the curious shape of the curve is just the result of a relatively rare random variation.

Another curious thing about the plot is that it appears to show rather steady, but changing, trends with occasional large deviations from the trend (specifically, a one-year drop in '81, another in '89, and a one-year rise in '96.  It seems likely that these years had some strange reporting irregularities.


Suicide is a phenomenon for which at least a very large portion of substitution of method occurs.  If a person wants to die and his/her "preferred" means is not available, he/she can quite easily determine another quite satisfactory means.  A good thing about means other than guns, when it comes to trying to kill oneself, is that most other means are quieter—so it's likely your intent won't be discovered if you are unsuccessful, leaving you free to try again until you are successful.  The "success" rate for the average attempt with a gun may be higher than for other means, but a suicidal person can use some of the other means repeatedly until successful.

History of Australia Suicide Rates, Guns and Non-guns
Fig. 2 History of Gun & Nongun Suicide in Australia

The gun suicide rate averaged about 3.45/100k from '79 to '87, then dropped unevenly at about .158/100k (average) per year from '87 to '96.  This ('97 & '98) is where gun controllers would think the suicide rate should drop immediately or start dropping at a higher rate (i.e., higher than .158/100k).  Instead, the average drop per year was virtually identical (.156/100k) from '96 to '04.

The truth is that the semiauto and pump firearms that were the target of the ban/buyback have no more value for suicide than any of the allowed firearms.  If one shoots oneself in the head with a rapid fire gun, one does not then take another shot.  It takes only one shot to commit suicide.  So one should not expect the ban (that hardly reduced the stock of long arms and didn't reduce the stock of handguns at all) to have had any impact on suicide.

It is also inconceivable that any other part of the new gun restrictions would have a perceptible impact on the gun suicide rate.  Attempts to identify people who are likely to try suicide (except for those who have already done so) have been woefully inadequate.  And those contemplating suicide are likely to hide weapons rather than surrender them all to their government.

Short Term Gun Suicide History
Fig. 3 Short Term History of Gun Suicide in Australia

One may wonder (from looking at the graph) if a trend in gun suicide had started from '93 to '96, in which case it could be argued that gun suicide dropped abruptly in '97 and '98, interrupting the trend.  It is very possible.  The '93-'96 points lie very close to a straight line.  Another year or two of data points along the line would have made it quite probably a trend.  A trend of only three years cannot be well proven because the possibility of such patterns occurring from chance alone is too great.  If '93 to '96 was a trend, the post-'96 trend still appears to be heading towards getting back on the '93 to '96 trend line in a few years.  Note too that the values from '01 to '04 look much like a nonlinear trend, largely because of the small departures from the curve.

Unfortunately, we can't be very certain of conclusions drawn from such short-term trends even if the trends are very smooth.  On the other hand, long term trends are not really good bases for detecting a change at a particular point in time if the fact is ignored that there really are many true short-term trends.  A long-term trend needs to be stable to be useful.  Suicide and crime rate trends result from many things, some consistent over the long term and some only over short terms.

The nongun suicide rate averaged about 7.859/100k from '79 to '85, then jumped up where it averaged about 10.086/100k from '87 to '94.  It increased at a rapid average rate of about .873/100k per year from '93 and '94 to '97, then dropped at a high, consistent rate (.585/100k per year, average) from '98 to '04.  Note that the '96 rate was significantly below a trend line through the '93, '94, '95 and '97 points.  Remember this when we get into the discussion of assault and robbery rates.

History of Total Suicide Rate
Fig. 4 History of Total Suicide in Australia

It is noteworthy that the rates for total and non-gun suicides dropped fast and consistently from '98 to '04.  There is no way these impacts could relate to firearms.  So something else has been actually working and reducing suicide rates.  Maybe it was good economy.  Maybe Australian governments have actually been doing something worthwhile.  Whatever the cause, it had no impact on gun suicide rates, which simply continued to drop at essentially the same rate as between '91 and '97.


It appears that the suicide rate and non-gun suicide rate started dropping dramatically in '98, but the gun suicide rate dropped no faster than pre-'96.  Because the gun suicide rate did not improve, it could not be argued that the gun ban/buyback had anything to do with the suicide reduction that began in '98.  However, some other policy change that was begun via the National Committee on Violence may have had something to do with the overall suicide reduction.


Supporters of the right to private possession of firearms have pointed at increases of some crime rates, since the buyback began, as evidence that it has increased the rates of those crimes.  The Australian government and other gun controllers have cried "foul" and argued back in terms only of "firearm" deaths—which is not a valid basis for evaluating a policy.

Let's examine what has happened to rates of some types of Australian crime since the buyback, and some other changes that went into effect at the same time, but also for a few years before it began.  First, it has been reported that the rate of assault has increased dramatically since the buyback.


Australia Assault Rate History
Fig. 5. Assault Rate History for Australia

Looking at the graph of the Australian assault rates from 1991 through 2004, it appears that something bad was at work in Australia until near the end of that period.  The assault rate increased every year at a high rate.  (And, in fact, for many years preceding.)  Note, too, that the rates do not vary much from the basic trend.  This means that the data are not affected by chance as much as it is by underlying causes.  The relationship seen is a reliable one.

It is true, as claimed by gun possession advocates, that the assault rate has increased a lot since the ban/buyback.  But the ban/buyback cannot be blamed for this since the rate was increasing at essentially the same rate before the ban/buyback (and the other changes) began.  The rate did not rise as fast for '99 and 2000, but then rose even faster than before for '01 and '02.  The ban/buyback did not have a definite effect of any kind on the assault rate—neither improving it nor making it worse.  It is possible that the small declines for '99 and 2000 were caused by the ban/buyback and other changes made at the same time.  But, if so, something must have completely counteracted that in '01 and '02.

As shown in the assault rate graph, virtually a straight line can be drawn through the points for 93/94, 95/96, & 97.  Leaving the 1996 point below the line is justified because the 96 rate was artificially lowered as a result of a reduction for the 2/3 of 1996 following the Port Arthur incident.  This reduction immediately after the incident can be calculated from the rates shown in the graph of assaults, by month, over a period of several years in Facts and Figures, 1999 (pdf) by the AIC* (after accounting for the seasonal variation apparent in the graph).

The same effect is obvious (without calculation) in the graphs in the same document for murder, robbery and armed robbery.  One could hypothesize several possible reasons to explain why murder and robbery frequencies reported by police dropped dramatically, and assault frequency dropped moderately, for a while after the incident.

Because the assault rate is artificially low for '96, it is not legitimate to compare post-ban rates with the '96 rate.

*One should be cautious about bias in information or conclusions in documents produced by the AIC.  The organization has a definite bias toward promoting gun control on the basis of biased "evidence," like the biased advocacy of USA medical researchers.

Note that the available figures include some (the first 5) that are for years beginning in April and ending the next March.  So, in the graph, the points for these data pairs are centered on Sep 30 (Oct 1).

One AIC paper suggested that crime hasn't really increased over the years as one would think from looking at crime stats (other than homicide) reported by police.  The theory is that the increases are just because the police are recording more of what crime occurs, and that crime rates resulting from the victimization surveys are more real.  Easily said.  What about the assault rates from victimization surveys?

The assault prevalence rate (victims, not victimization incidents, per population unit) from the 1993 survey by the Australian Bureau of Statistics (ABS) was 2.5, closely 5 times that indicated in the graph for the middle (10/1/92) of the period covered by that survey.1  The rate from the next (1998) survey (Apr 97 - Mar 98) was 4.3, 6.33 times as great as shown in the graph (10/1/97).2  Finally, the survey by the AIC as part of the UN international crime survey in 2000 for victimization in '99 yielded a survey rate 11.07 times the rate indicated in the graph for '99—so the increase for '99, according to the victimization surveys, is higher than for the earlier times.

In all three surveys taken within the time period covered by the graph the survey rate was greater by an ever increasing amount than the rate shown in the graph.  The rates for incidents would be even greater (some people are victimized multiple times in the survey period).  So changes shown in the graph are conservative indicators of actual changes in assault rate for Australia, and assault truly has been rapidly increasing for years in Australia.

It is obvious from the figure that something good happened to the assault rate for the last two years.  Maybe the same thing that caused reductions in suicide rates for the last few years?  Like a good economy?  The trend came too late to be attributed to the gun policies, and one would not expect gun policies to impact assault rates anyway unless the policy related to people being widely able to protect themselves from assault.

SUMMARY:  The assault rate data is inconclusive.  Two years ('98 & '99) of the assault rate not rising as fast as it had been does not make a definite trend, especially since the rate then jumped up dramatically for the next year (2000) and rose again rapidly in 2001.  The ban/buyback had no perceptible impact on assault rates, neither increasing assault nor decreasing it.


AU Property Crime Hist
Fig. 6. Robbery and Burglary Rate Histories for Australia

Both robbery and armed robbery went up a bit in '96, then increased at a higher rate during the buyback period in '97, then again at a lower rate in '98.  The rate increase for '96 would have been larger if it were not artificially depressed as a result of robberies being extraordinarily low for the 2/3 year immediately after the Port Arthur massacre.  The robbery rates dropped a bit in '99 and 2000, but stayed higher than '97 or preceding years all the way up to '04.  The rate for overall robbery had been increasing steadily from '91 through '95, although not as fast as afterward, while the rate of armed robbery was nearly constant during this period.

Both robbery and armed robbery appear to have stabilized (about 4 years) at rates higher than they were before the Port Arthur incident and the ban/buyback, but not so much higher that they could not simply have been a continuation of the pre-'96 trend.

The burglary rate had been dropping slightly for three years until the Port Arthur massacre and buyback, then increased a bit in '97, where it appears to have continued on the downward trend barely started before Port Arthur.  (No artificially low rate of burglary was apparent in the remainder of '96 following the massacre.)  [The burglary rate for 2000 is not shown in the figure because it was not included in the statistics released by the Australian Bureau of Statistics in May 2001 because a law change in South Australia made it impossible to separate statistics for "unlawful entry with intent to take property" from statistics for other types of unlawful entry.]

Rates of burglaries of homes with occupants present would be useful to analyze, because this could indicate increased burglar willingness to confront and possibly injure defenseless occupants, but we have not yet located a source for data to make such analysis.

SUMMARY:  The rates for both robbery and armed robbery rose faster for a couple of years after '96 than they had before, then stayed higher for several years.  The burglary rate appears to have been affected only in '96, although this could easily have been a chance effect.  The fall-offs for all three after '01 were too late to be attributed to the '96 gun changes.  The chart looks a bit like a short-term increase in robbery may have occurred after '96 until about '04.  Maybe criminals were emboldened for a few years by the thought that potential victims would not be able to defend themselves.

Australia has been increasing its expenditures on "welfare"3 and their criminal justice system.4  Are things like these and reductions in problems like unemployment more likely reasons for the crime downturn starting from '99 to '01?  Expect crime rates to rise because of the '08/'09 recession.


"Homicide" as discussed here is murder plus manslaughter.

The homicide incident rate (not shown) is a better basis for evaluating impacts on homicide than the total homicide victim rate.  Even though the incident rate is based on fewer numbers than the victim rate, the incident rate is a better measure of homicide, than the victim rate, when examining whether or not homicide correlates with other events/factors.  The reason for this is that the incident rate is more stable from year to year than the victim rate, which jumps (up) randomly and sometimes radically in random years.  The victim rate does this because it is highly influenced by relatively rare multiple (mass) killings.  (The multiple-killing data alone could be good for examining effects of high firing rate weapons.)

The homicide incident rate had been pretty steady from 90 thru 94, then went down for 95/96 (the year barely including the Port Arthur incident), then went down less each year for 96/97 & 97/98.  The drop for 95/96 was approximately the result of including two months following the Port Arthur incident, when homicide frequency was dramatically reduced apparently as some kind of national or criminal shock in the wake of the incident.  That is, don't be tempted to think that the homicide incident rate was falling before the Port Arthur incident.  However, the rate definitely dropped a little for 96/97 and 97/98, too soon to be a result of the buyback.  The rate for a later time could not be found.

Australia Homicide Rate History
Fig. 7. Homicide Rate History

The top line represents homicide victim rate history, but with some modifications.  First, the rate shown is a 3-year moving average.  Specifically, the average is of the current year plus the 2 preceding years.  Another modification is that what is shown is the rate minus 1.

The moving average is shown in order to smooth out some of the random wild swings alluded to earlier (caused by dealing with small numbers).  The average is for current and past time only so that any change that appears abruptly as time increases could be correlated with the time at which the rate is plotted.  The "1" is subtracted to reduce the height of the graph.  (Just add "1" to a value from the curve to get the actual value.)

The homicide rate had peaked about '90, then dropped to a low about '94.  The rate increased slightly after the Port Arthur incident, but it was also increasing by about the same rate for a couple of years before.  So it is not reasonable to blame the increase in homicide on the ban/buyback.

The lower line represents firearm homicide victim rates.  We had too few years of data for firearm homicide incident rates to bother plotting this, although such data would be better to use.

The firearm homicide rate appears to have dropped a bit erratically after Port Arthur and the ban/buyback.  However, it also appears to have been dropping at about the same rate since '86.  The reduction in gun homicide after the ban/buyback should therefore not be attributed to the ban/buyback.

SUMMARY:  The homicide rates provide no support for a proposition that the ban/buyback has helped.  However, they also do not indicate that the ban/buyback caused anything, good or bad.

Gun controllers are fond of claiming that gun availability to the individual often causes the individual to kill someone, and that these killers are mostly just normal people like you and I.  Since the buyback, Australia's homicide monitoring program has observed that the huge majority of gun homicides have been committed by people not legally allowed to have a gun, using unregistered (illegal) guns.  Does this mean that a similar portion of those with guns don't possess them legally?  The same program observes that a large majority of those who commit homicide had a prior criminal history, which is consistent with what criminologists have known for years.

It is intriguing that the gun homicide rate peaked a bit in '88, the year after the Hoddle St. and Queen St. shootings that first got Australians excited about the "gun problem," then peaked a bit in '92, again the year after the '91 Strathfield shooting (and two after the '90 Surrey Hills shooting).  The gun homicide rates in the years in which these incidents occurred were similar to rates for other years.  The Port Arthur massacre and the other mass shooting in '96 did noticeably increase the rate for that year.

Australia Murder Rate History
Fig. 8. Murder Rate History (short term)


The murder rate had no definite trend from '93 through 2000.  All the rates lie within a range that is about the same size as the changes from one point to another.  That is, the random variations are too large to draw any conclusion about some small trend that might exist.  The gun murder plot seems to be slightly less random, rising a little for '96 and '97.  However, the pattern is not very conclusive since the rate for '99 went back up a bit after the drop in '98 (i.e., no consecutive years heading the same direction).  So the murder and gun murder rates do not support the idea that anything about murder rates was caused by the ban/buyback.  Note that the murder rate actually dropped a little in '96 in spite of the Port Arthur massacre and another mass killing in that year.  The reason for this is that the usual peak in murders/homicides during the Australian spring didn't occur that year.



  1. S. Mukherjee, A Statistical Profile of Crime in Australia
      (Canberra, Australia: AIC, April 1993) RPP07, Table 4.5
  2. Crime and Safety part of "Australia Now" series at Australian
      Bureau of Statistics (ABS) web site, Table 11.12
  3. "Income and Welfare:  Income support programs - Dept. of
      Family & Community Services" p/o "Australia Now"
      series at ABS web site.
  4. "Crime and Justice:  Expenditure on public order and safety"
      p/o "Australia Now" series at ABS web site.