After Martin Bryant murdered 35 people at Port Arthur, Prime Minister John Howard got new laws enacted that banned semi-automatic long guns. At the time, I felt that was bad policy. Since almost all most gun killings involve just one death it didn’t matter whether whether the gun was a semi-auto or not, writing:

I think that the new laws are mostly a stupid waste of money.

In the years that followed, homicide, firearm homicide, suicide and firearm suicide all fell, but a paper published last year by Baker and McPhedran argued that the decline was not statistically significant. However, Andrew Leigh found a flaw in their analysis. Look at this graph of the gun homicide rate (the solid line):


For the reduction to be significant, the rate would have had to have dropped below the bottom dotted line. In other words, to below zero. There is no possible outcome that their method would have found as significant.

Now he’s done some more analysis and written a short paper that shows that if you use more data than Baker and McPhedran or change their method slightly, you find significant effects.

There is never going to be conclusive evidence in a case like this because it is possible that the decline was coincidental, but it seems likely that the laws were at least partly responsible, so I must admit that I was wrong when I predicted that they would have no effect.

There is discussion at Kalimna, John Quiggin and a response from Baker and McPhedran that I don’t find persuasive.


  1. #1 JB
    April 25, 2007

    Perhaps the recent massacre here in the US at Virginia Tech will spur a similar ban on semi-automatic firearms, though I somehow doubt it.

    The NRA basically has all the officials and candidates for public office here licking their boots. It’s all rather pathetic.

    The worst part of the Virginia Tech case was that the shooter had been deemed mentally ill and was nonetheless able to legally purchase the guns at a gun shop.

    Actually the most imbiggest single
    Their theory is “the more guns the merrier”

  2. #2 Bill O'Slatter
    April 26, 2007

    No you were right the first time :I would say that it would be hard given the data to prove that the gun laws hve has an effect. There effect hopefully is to decrease the frequency of a rare but catastrophic event: the running amok massacre. All the above graph shows is a long term trend. Homework : Use StatXact to tighten the confidence interval

  3. #3 John
    April 26, 2007

    Hmmmm…not sure I agree with that argument about the CIs. It looks like the observed rate could have fallen below the lower CI at any time pre-2004, which allows the validity of the test to stand. Just because it “didn’t” happen doesn’t mean that it “couldn’t” happen. Two completely different matters.

  4. #4 Andrew Leigh
    April 26, 2007

    Tim, thanks for the plug. I should point out that the first author on the study is Christine Neill, who did most of the statistical heavy lifting.

    And you’re quite right to say that time series analysis isn’t the best way of answering the question, but our main purpose was to address what we thought were fragile findings by Baker & McPhedran.

  5. #5 Charles Stewart
    April 26, 2007

    Isn’t it likely that shootings involving semi-automatic weapons are more likely to result in a fatality than those involving non-automatic weapons?

  6. #6 nerdwithabow
    April 26, 2007

    “Look at this graph of the gun homicide rate (the solid line)”

    …but what trend does the graph of all homicides show?

  7. #7 nerdwithabow
    April 26, 2007

    In fact, the drop in all homicides hardly seems significant:


    Nothing has been accomplished if people just choose other means of murder.

  8. #8 Ahcuah
    April 26, 2007

    The end of the abstract says:

    The high variability in the data and the fragility of the results with respect to different specifications suggest that time series analysis cannot conclusively answer the question of whether the NFA led to lower gun deaths. Drawing strong conclusions from simple time series analysis is not warranted, but to the extent that this evidence points anywhere, it is towards the firearms buyback reducing gun deaths.

    I guess I’m having a hard time seeing how this expressed uncertainty relates to having significant effects (unless it is just referring to statistically significant, but not strongly statistically significant). It just does not sound all that strong to me, or that clear conclusions can be drawn.

    By the way, the paper is here.

  9. #9 JC
    April 26, 2007


    I think your conclusion doesn’t make sense.

    If you allow the possibility the fall is a coincidence -it seems it was seeing the Monash University student in 2001 was tackled to the ground before he could kill a bigger number than the previous record as he had the ammo to keep shooting- you can’t then claim your first impressions of the laws were wrong. It seems to be the wrong conclusion if you think there is the possibility they are coincidental.

    The Monash attempt at mass murder is proof enough that laws do zippo to prevent murder by hand guns seeing the Howard laws were well established at the time.

  10. #10 ben
    April 26, 2007

    Isn’t it likely that shootings involving semi-automatic weapons are more likely to result in a fatality than those involving non-automatic weapons?

    Maybe, but first tell us why you think that, then we might be able to answer.

    In reality, no. Especially suicides. Tim says that firearm suicide fell. They type of gun there certainly could not have made a difference, at least not at the time the trigger was pulled. The ban could have made a difference if folks who might have had a semi-auto chose to have no gun instead of a non-semi gun. Other studies have shown that for most groups of people, they simply find an alternative method for ending their lives when the guns are taken away.

    If you can actually rid the place in question, and maybe this worked well in Australia (being an island and all), then you can probably lower the lethality of gun rampages like our recent one at VTech.

    And yes, VTech did highlight the need for mental health data to be immediately and thoroughly reported to the background check system. The background check should also be made available to all of us citizens who sell guns privately.

  11. #11 P
    April 26, 2007

    Call me crazy, but I see a rate that was, over time, falling since 1980, fell quite far and fast in the early 90’s, had a small uptick the year before the massacre and a spike the year of….then kept falling as it had been before.

  12. #12 Robert S.
    April 26, 2007

    I agree that besides the chart not conclusivly proving anything, plus, if the overall murder rate stays the same, there’s just substitution going on, so what does it matter, overall?

    As far as suicides (with or without guns) Japan has none and has a high suicide rate, so that tends to say it’s a social issue.

    I’m sure there’s a number of people in Australia that either didn’t give up their semi-automatic rifles or bought illegal ons smuggled in. So it would still be possible for a deranged mass killer to have or get one (although of course more difficult).

    Inherently, no, bolt action rifles are no less lethal, at least not per shot. They just typically hold fewer rounds, are slower to fire, and take longer to reload if they are not magazine fed. Non semi-automatic pistols like revolvers aren’t less lethal either, just hold fewer rounds and can take longer to reload than magazine fed weapons. Overall, it’s the caliber per shot, and a number of other factors if you’re talking about rate of fire.

    The Virgina Tech shooter had been in a mental facility in the past, but he was not currently “insane” or otherwise tagged as ineligable to buy weapons, nobody ever did that.

    The NRA is a lobbying organization with a great many people, so it is a political force as powerful per person as any other (say AARP or AFL-CIO or etc) It’s a group of people with a common interest that band together to get their views known.

  13. #13 Robert S.
    April 26, 2007

    I got part of that wrong before, he hadn’t been in a mental facility (or “involuntarily committed to a mental institution”) it was counselling on an outpatinent basis. So at least according to the law in VA, he wasn’t disqualified.

    “A judge declared him a threat to himself and ordered him to seek outpatient mental health counseling.”
    “The form [handgun purchase] asks, ‘Have you ever been adjudicated legally incompetent, mentally incapacitated or been involuntarily committed to a mental institution?'”

    So the answer to the question in Cho’s case was “No” because he hadn’t. A danger to yourself and ordered to attend outpatient counselling is none of those three.

    The writer of the below (where I got the above information) thinks that’s a loophole that needs to be closed.

    Close the loophole Cho sneaked through

  14. #14 Ian Gould
    April 26, 2007

    Robert: “Close the loophole Cho sneaked through”

    I think there needs to be some consideration of what “closing the loophole” would mean. Can a form of word be drafted that would have caught Cho but that wouldn’t unfairly penalise millions of other Americans with minor psychiatric problems?

    (While I tend to think tighter handgun laws in the US would probably be a good idea, I’m not in favor of effectively introducing such laws by stealth and in a discriminatory fashion.)

    But leaving that caveat aside, I agree with Robert that closing the loophole Cho exploited seems like a good idea.

    Wouldn’t it be nice if the people with polarised opinions on both sides of the US gun debate put them aside long enough to deal with that issue.

  15. #15 oconnellc
    April 26, 2007

    Does anyone really think that any wording of the question would have caused Cho to answer in such a way that would have caused him to be denied in his attempt to purchase a gun? I’m sorry, but the thought of rewording the question and somehow preventing more of these sorts of incidents is crazier than Cho was…

  16. #16 SG
    April 26, 2007

    Ben, before you get your holster in a twist you should bear in mind that the authors of this study don`t actually support the type of statistical analysis used, and were doing it in response to another quite poorly executed paper. If you go to Andrew Leigh`s website and read the comments, I think you`ll find they preferred the method used by Chapman et al, which found no statistical evidence of an effect due to the banning and gun buyback (I`m saying this from memory so don`t sue me if I`m wrong, Mr. Leigh). Leigh and Neill have been faultlessly polite throughout the ordeal on Andrew Leigh`s website, too, answering critics patiently and thoroughly and putting up many graphs (as well as all their data and methods, in case David Kane is reading).

    But as regards the suicides, I think it is reasonable to suppose that there would be a decline if you accept, as you said it, that

    [some] folks who might have had a semi-auto chose to have no gun instead of a non-semi gun

    Also, presumably for some proportion of these people, the semi-auto might have been the most powerful gun in their arsenal, and having gotten rid of it they were left with a (presumably only marginally) less powerful weapon with which to take the final trip to the Great Hunting Grounds. This would presumably have reduced the chance of successfully topping themselves (although I`m sure Ben that these shooters are all highly trained and would never miss – I`m not trying to impugn their manhood).

    So presumably, depending on what the proportion of these people is, removing these autos would have reduced suicide rates.

    But from an empirical point of view it doesn`t matter. The hypothesis: less guns= less gun deaths. The experiment: less guns. The evidence: less gun deaths. Hypothesis confirmed (except Leigh and Neill argue that this experiment wasn`t measured properly).

  17. #17 Ken
    April 26, 2007

    Given a long overall trend of lowering homicide rates in Australia, given that the number of murders are relatively few (under 300pa) and that the number that were from guns is much smaller and quite variable (eg 53 victims in 2003-04, 40 victims in 2004-05) and that a single mass killing incident will skew numbers significantly, I’d say it’s hard to infer too much using recent statistics.

    That said, a more general decline in gun ownership probably is a significant factor in that trend – as a kid in the 60’s in a provincial town, going hunting was more popular than now, Dad taking the boys (almost always boys, not girls) shooting was normal, the boys taking the guns themselves to go shooting, quite normal. Lots of households had guns and they weren’t locked away. No doubt they were more likely to result in death by suicide and accident than murder. In any case I think it’s reasonable to think that limiting access to tools designed to make killing people easy might make killing people more difficult. I’ve never heard of a drive-by Knifing. Someone inclined to commit random murder is more likely to carry it out when the available weapon allows distance between perpetrator and victim.

    I have to say I’ve never been convinced that carrying weapons makes people safer – whilst there are situations where it may save the day, in the (relatively more common) case of being in the vicinity of someone brandishing a gun (who probably plans not to actually use it) I’d think trying to draw a gun would increase the likelihood you get shot, not reduce it.

    On a lighter note, the issue of the right to bear arms in the US context always causes me to remember the Chaser crew a couple of years back asking Americans on the streets whether the Iraqi constitution ought to include the right to bear arms. The dumbfounded looks were hilarious. So would Baghdad be a safer place now if citizens were entitled to carry guns and use them against any regime they perceived to be tyrannical – or against foreign occupiers? Mmm … seems that’s what they are doing.

  18. #18 davidp
    April 27, 2007

    A declining trend before an intervention can be a result of previous interventions – weren’t Australian gun laws being tightened up prior to 1996? It’s like saying road deaths in Victoria dropped from 1024 in 1970 to about 500 in 1985, so interventions since then have all failed because road deaths are still over 250 in 2006. In fact it took interventions to get from 1024 to 500 and further interventions to get from 500 to 300.

  19. #19 Ian Gould
    April 27, 2007

    “Does anyone really think that any wording of the question would have caused Cho to answer in such a way that would have caused him to be denied in his attempt to purchase a gun?”

    No but I do think that it might be possible to build a Federal database of people who had been subject to certain types of mental health problems and require gun dealers to check it before selling people guns.

  20. #20 oconnellc
    April 28, 2007

    Ian, I’m glad we agree that the sort of ‘survey’ type of gun restrictions are likely to do much to prevent the “Cho’s” of the world from getting guns. You do raise an interesting question about a federal database… Since lots of people here are from different parts of the world, I’m not sure how universal this comment is… But, what are the privacy implications of this federal database? Whose responsibility would it be to make sure that my mental health problems get reported to the government? Given the extreme sensitivity that most people (in the US, at least) seem to attribute to health records, how likely is it that we want gun sellers to have access to this kind of information? And just what sort of mental health problems would cause you to go into this database? Personally, I’m not really interested in telling big brother about any mental health issues I might be having. I’m not a mental health professional, so I could be wrong, but from what I understand, it would have to be a very liberal interpretation of ‘problem’ to cause Cho to go onto a list like this in the first place. And how many people commit crimes like this without having previously been examined by a doctor and had some sort of trackable prognosis made?

    Typical disclaimer… I’m against gun violence and I don’t want to see anyone get murdered. So, that aside, I wonder if anyone has done a serious study that compares the number of gun deaths that could actually be prevented, with, say, the number of *extra* people that have died in highway accidents since the US government relaxed restrictions on highway speed limits. That may seem like an odd way to word a question, but I consider many gun deaths to be unpreventable. For example, gangs are unlikely to be stopped by gun laws, and here in Chicago, most gun deaths are caused by gang violence. I kinda put guns into the same category as drugs. Things like marijuana, heroin, crack and cocaine are all illegal. But people who *really* want them, get them. Perhaps some readers who live in England or Australia would care to comment on how available illegal things are in those countries as well. Many suicides are committed with guns, but I have to guess that a certain percentage of those sad people would just use some other method to kill themselves.

    I don’t mean to be cavalier about this, but I read that this gun law that we are discussing ended up costing society hundreds of millions of dollars. Has anyone done a study that compared how many lives would have been saved had that money been used to fight gang crime? Or, what if that money had been spent on cancer research? Or used to buy insulin for poor kids with juv. diabetes?

  21. #21 Ian Gould
    April 28, 2007


    The cost and civil rights issues you raise are quite valid.

    There needs to bew a serious debate about whether the potential benefits of such measures outweigh the costs.

    This article says that current US Federal law requires that guns not be sold to people who have been involuntarily committed. States are required to implement that law and in Virginia that means doctors and hospitals have to notify the State police.

    It asppears the current system fell down in two ways – trhe Virginia definition of mental illness is different top that in the Federal law and the medical authorities only considered the Virginia definition. THat meant Cho was able to buy one of his guns legally in Virginia. The other gun he bought over the internet from Illinois – and that sale should have been illegal. But the gun dealer had no way of checking Cho’s mental state and relied on a simple statement from Cho.

    Creating a Federal register would mean consolidating existing state-based registers rather than creating a new register from scratch and impose a single consistent definition of mental illness. That’d close the two loopholes Cho exploited to get his guns.

    As to the issue of giving gun dealers access to the information, presumably they’d have to submit written applications from their custoemrs and would jsut get back a yes/no response. In any case, some similar system must already be in place for the existing checks for criminal records.

  22. #22 z
    April 29, 2007

    “Statistically significant” is a bit of a red herring here. As discussed elsewhere on the blog, statistically significant is not an absolute test of reality.

    The old eyeball test suggests that regression to the mean plus whatever process had been reducing the rate since the 1980s would suffice to explain the results.

  23. #23 Gaz
    August 6, 2008

    Statistical analysis of gun-related deaths that omits other plausible influences (age distrubution of population, trends in drug use, unemployment, the number and profile of people who still have guns, etc) seems unlikely to be very illuminating.

    It’s a bit like looking at changes global temperature without taking El Nino into account.

New comments have been temporarily disabled. Please check back soon.