Weak Tests and Strong Conclusions

Christine Neill and I have a paper out today, re-analysing a much-publicised 2006 study on the effect of the gun buyback on gun deaths. The original paper – by Jeanine Baker and Samara McPhedran – isn’t available online. But here’s how they summarised it on the Sporting Shooters’ Association of Australia website* last year: 

“The reforms did not affect rates of firearm homicide in Australia. 
The reforms could not be shown to alter rates of firearm suicide, because rates of suicide using other methods also began to decline in the late 1990s. … 
It must be concluded that the gun buyback and restrictive legislative changes had no influence on firearm homicide in Australia. 
The lack of effect of a massive buyback and associated legislative changes in the requirements for obtaining a firearm licence or legally possessing a firearm has significant implications for public and justice policy, not only for Australia, but internationally.”  

Our paper questions those conclusions. Here’s our abstract:

Weak Tests and Strong Conclusions: A Re-Analysis of Gun Deaths and the Australian Firearms Buyback
Christine Neill and Andrew Leigh
Using time series analysis on data from 1979-2004, Baker and McPhedran (2006) argue that the stricter gun laws introduced in the National Firearms Agreement (NFA) post-1996 did not affect firearm homicide rates, and may not have had an impact on the rate of gun suicide or accidental death by shooting. We revisit their analysis, and find that their results are not robust to: (a) using a longer time series; or (b) using the log of the rate rather than the level (to take account of the fact that the rate cannot fall below zero).  We also show that claims that the authors had allowed both for method substitution and for underlying trends in suicide or homicide rates are misleading. 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.

Doing time series analysis (ARIMA 1,1,1 to be precise) to test the impact of the gun buyback on deaths isn’t our favourite way of doing things. Statistical tests consistently reject the hypothesis that homicide rates follow a non-stationary process. But even when we use this approach, we find that all it takes is a couple of small tweaks in methodology to get much bigger coefficients on the effect of the gun buyback.

Overall, our paper estimates that 128-282 lives have been saved every year by the gun buyback. In other words, 1000-2500 Australians who are alive today would not be here if it hadn’t been for the buyback. This is broadly in line with two other Australian papers (Ozanne-Smith et al 2004 and Chapman et al 2006), and with what I consider to be the best US study (Mark Duggan’s Journal of Political Economy paper More Guns, More Crime).

The standard Australian estimate of the value of a statistical life (from a recent paper by Peter Abelson) puts it at $2.5 million. Since the gun buyback cost $500 million, this indicates that even if we use the lowest reasonable estimates of its impact on homicide and suicide (128 lives saved per year), it paid for itself in just two years.

Most paper have an odd genesis, but there are a few unusual aspects to this one. The first is that I never would have gone to read Baker & McPhedran’s paper if it wasn’t for my father picking up the phone to say ‘you should have a look at this study and see if it’s right’. That led to a couple of blog entries, after which commenter Christine Neill and I (egged on by Justin Wolfers, who has done some great work re-analysing dodgy death penalty research), decided to write up a brief comment paper. Without appendices, it’s only 12 double-spaced pages, so do read it if you have time.

* The SSAA website seems to have been redesigned since then, and this research summary no longer appears there.

This entry was posted in Health economics. Bookmark the permalink.

107 Responses to Weak Tests and Strong Conclusions

  1. Sinclair Davidson says:

    No appendix? Some people will do anything to get out of a seminar. :)

    Seriously, two points: first you probably assume those 1000 – 2500 Australians alive wouldn’t have been shot while committing a crime, and second have you offset the non-deaths by the deaths of crime victims who might otherwise have defended themselves? So while I’m happy to believe fewer gun deaths have occured, I suspect you might be over-estimating the impact of the buy-back.

  2. Damien Eldridge says:

    Nice work Andrew and Christine!!!

  3. Andrew Leigh says:

    Sinc, our approach (and that of BM) is just to look at total gun deaths. So that addresses the self-defence question, unless you think that a whole lot of people who were killed with a knife would have saved their lives by owning a gun.

    And your opening quip suggests a scarily good recollection of my appalling puns. Perhaps this means we both need to spend less time on the Net….

  4. Simon Chapman says:

    Andrew & Christine. Nice job. Your Google ratings are about to skyrocket with all the rage this will cause on gun lobby blogs!

    One thing that must be emphasised though, is that the 1996 gun law reforms were precipitated by Port Arthur, the proverbial back-breaking straw that came in the wake of a decade of mass shootings. The centrepiece of the reforms was the outlawing from civilian ownership of semi-automatics via the buyback. These are the favoured weapons on anyone intent on killing lots of people quickly, without the bother of reloading after each shot. The Virginia killer last week used semi-automatic handguns, as you’ll find most of these incidents do.

    Howard put tighter controls on the remaining single shot weapons (principally gun registration and denial of self-defense as a legitimate reason to have a gun license), but significantly, he did not ban these sorts of guns, although a good many were handed in in the buyback and in other amnesties.

    Around 80% of gun deaths are suicides, about 15% homicides and the rest accidentals. As you need only one shot to kill yourself, removal of semi-automatics would not be something that you would hypothesise would lead to reductions in gun suicides: a single shot weapon will suffice, and many gun owners had both.

    Most gun homicides are single victim killings, often where the victim is known to the murderer. Again, there’s no special reason to predict that removing semi-automatics would reduce gun homicides, although a semi-automatic of course gives you more opportunities to kill your victim(s) if you missed with the first shot.

    Where you might hope that policy intent matched subsequent outcome, is that removal of semi-automatics would reduce or prevent mass killings like Port Arthur, Virginia, Dunblane, Hoddle St etc. And of course that’s what’s happened, and what Baker & McPhedran simply could bring themselves to even consider in their self-serving paper (both are gun lobbyists).

    If you want to read some really scary stuff, take a while to browse gun lobby blogs and websites. They are full of post-Virginia frothing about how the dumbarse authorities in Virginia wouldn’t allow kids and teachers to pack heat in the school. They fantasise about arming our communities to the teeth, Baghdad style, so that all these pesky malcontents can meet deadly force, with armed kids blasting away in the classroom.

    Currently, the United States has 14.5 times Australia’s population, 102 times its total firearm deaths, and 173 times its number of firearm homicides. That’s the road these people seriously want us to take.

    Simon

  5. Sinclair Davidson says:

    you think that a whole lot of people who were killed with a knife would have saved their lives by owning a gun.

    Maybe. I heard you on ABC radio saying that a lot of gun suicides had been prevented. I am uncomfortable though with equating the life of a murder victim (involuntary death) with a suicide victim (voluntary death). Although I’m also aware that many suicides are mentally ill, so death isn’t really ‘voluntary’. These are difficult trade-offs. Although if the policy was to reduce gun deaths, then it has succeeded (on you evidence).

  6. Andrew Leigh says:

    Simon, I take your points, but I still think that the success of the buyback may have had an effect that it was not explicitly intended to have. Massacres make the front page, but in terms of lives saved, reducing the gun suicide rate matters much more. Same goes for gun homicide, where the typical victim is a one-off domestic violence killing, not someone murdered in a mass shooting.

    I think Sinclair’s point raises some really tough moral issues, which I’m glad to be able to sidestep by virtue of the fact that the buyback reduced both gun homicides and gun suicides.

  7. Sinclair Davidson says:

    The gun buyback would certainly have prevented (most) opportunistic gun homicide and suicide. People who are serious about killing someone else or themselves will generally find a way to do so. In that measurable sense we are better off, the deeper issues are much harder to evaluate and become very speculative very quickly. The question to ask is this: Even if Andrew and Chritine had found no impact would there be much of a constituency for a relaxation of the gun laws? I think not.

  8. harry clarke says:

    Its obviously an interesting result.

    Can I make a macabre point about your comment above rather than your paper? What is the value of a human life for someone who wishes to commit suicide? Economic models of suicide (Hamermesh & Soss) say that a person who kills themselves does so when their lifetime future consumption urtilities fall below a reservation level. While this reservation level might not be zero it might be a lot less than the $2.5 million that you are attaching to the average value of life of a murder victim.

    With a zero valuation on the value of life of a suicide the cost benefit conclusion you propound above would be much weaker since 80% of the deaths you save are suicides.

  9. Andrew Leigh says:

    Harry, it’s an important point, but I think H&S assumes rational suiciders. That will be true in some cases (eg. cancer sufferers), but my guess is that many suicides involve mental illness.

  10. Uncle Milton says:

    Why is it Neill and Leigh not Leigh and Neill?

  11. mugwump says:

    The effect of the buyback on total gun deaths, while interesting, surely is irrelevant when determining how many lives were saved by the policy. You have to look at the overall death rate, since presumably those suicides and murderers who can no longer lay their hands on semi-automatic weapons have other ways of killing themselves or each other.

    Can you determine that the overall death rate has reduced as a result of the buyback?

    Furthermore, surely the people who were previously killed by semi-automatic weapons were more likely to be criminals and/or welfare dependents, hence their deaths were probably a net saving to society, not a $2.5M loss.

  12. rossco says:

    I heard Newt Gingrich on the ABC this morning (AM?) saying gun deaths had increased in Australia and elsewhere as a result of tighter gun controls. If students at Virginia Tech had been allowed to carry guns they would have shot Cho thus saving lives. But who could believe anyone with a name like Newt Gingrich? He is probably angling for support from the NRA for his run at President.

  13. mugwump says:

    Newt has a point. In a society where semi-automatic weapons are readily obtainable by anyone, failing to enforce on-campus bans against weapons is a death sentence.

    Either ensure nobody can carry weapons into the classroom, or allow people to defend themselves.

  14. Damien Eldridge says:

    Perhaps someone should send Newt a copy of Christine and Andrew’s paper!!! On the issue of suicide, I agree that most cases are likely to involve mental illness. Furthermore, if the person contemplating suicide has time inconsistent preferences, then that raises some difficult issues in welfare valuation. On top of this, there are almost certain to be Pareto relevant negative externalities involved with suicide. So preventing it is likely to be a good thing!!!

  15. Andrew Leigh says:

    Mugwump, we find no evidence of method substitution.

    Uncle Milton, it’s because Christine did more work on the paper than me!

  16. David Berry says:

    Hi Andrew,

    Did either study address the number of gun related injuries?

    Surely the pro-gun lobby is happy to frame the debate purely in relation to the number of gun deaths. As this discounts the cumulative damage from guns which must include the number of people are injured by guns (physically, let alone even gun victims threatened with injury and the associated significant psychological damage of gun trauma) as well as those people actually shot dead.

    A pure analysis of gun deaths over time would also need to factor in medical improvements such as emergency response times and improved surgery survival rates due to medical advances.

    Finally, the pro-gun lobby probably has a valid point in that the guns initially confiscated were taken predominately from ow-risk gun owners. Violent people and criminals clearly being less likely to relinquish their weapons for obvious reasons. This introduces a significant time lag in removing firearms intended for criminal and/or violent person-person usage. That is, the best results from gun control are probably yet to come – fertile ground for research.

  17. mugwump says:

    In which case, there should have been an almost discontinuous reduction in the death rate from suicide and homicide (by all means) after buyback. What was it?

    If any lives were indeed saved, do you have evidence supporting your claim that each one was worth on average $2.5M to society?

  18. Mugwump, he’s probably using a standard estimate of what a life is worth, which is an extensively studied problem as I understand it.

  19. mugwump says:

    But that’s my point: it is unlikely the lives saved (if any, since my suspicion is that the number of homicides/suicides by semi-automatic weapons is too small a fraction of the total number of homicides/suicides to be able to pick that signal from the noise) are worth a “standard life”, given that they are more likely to have been criminals and/or welfare dependents.

    The claim that the buyback paid for itself in 2 years is unsubstantiated at best, and probably the opposite of the true cost at worst.

  20. Andrew Leigh says:

    Mugwump, if you were worried about method substitution, you’d want to first look at gun deaths, and then if you found they were down, you’d want to look at non-gun deaths to check for method substitution. That’s exactly what we do. As it turns out, your method would produce the same result.

    But I don’t want to sound too much like a defender for time series econometrics. Our study was first and foremost a response to BM. At some point Christine and I may look at other econometric strategies for answering the question.

    As to $2.5 million per life, we didn’t do any calculations to get that number – we merely cited the best estimate in the Australian literature. For those interested in the methodology, the Abelson paper isn’t online, but clicking on the relevant link above will take you to the citation, and most libraries should have the journal.

  21. Pingback: Club Troppo » Missing Link

  22. Peter W. says:

    Contrary to the findings of the your ANU report, many others who have studied this topic and promote gun ownership, tend to disagree.

    Checking the data from ABS it is clear that suicides from non-firearms actually INCREASED dramatically in the years after 1996. This is highlighted in the Alpers/Chapman report; average in the years pre-1996 was 10/100,000, and it spiked to 12.9 in 1997, 13.09 in 1998, 11.75 in 1999…and didn’t drop back to 10.71 until 2002. (an estimated 1,200 extra deaths because of the gun buy back). Most gun owners knew of people who became depressed and committed suicide, after having to hand in a precious family heirloom or treasured hunting gun and watch it crushed, or who became suicidal at all the media vilification, or being treated as a potential mass murderer by our Prime Minister.

    Does you Andrew actually believe that it is somehow “better” for someone to have killed themselves by rope or poison or car exhaust, than by firearm? Are they “less dead”?

    Do you actually understand that the buyback involved rapid-fire, multiple shot firearms, but that a suicide only involves a single shot? For that reason alone, the buyback could not have influenced gun suicides. (Perhaps Dr. Leigh should have studied ballistics, not economics!)

    After a few years, agreed, the suicide rates with both firearm and non-firearm did start to decrease, but that is more a factor of programs such as “Beyond Blue”.

    To claim otherwise is to insult the efforts of workers in the many suicide prevention programs.

    From ABS figures we also find that assaults have been increasing at five times the population increase, since 1996. The fact that many of those victims of crime were saved, rather than becoming deaths, is a credit to the medical profession.

    We can only wonder how many more lives would have been saved if the money wasted on gun crushing had been invested in Doctors and Hospitals.

    Another factor in reducing suicides (and to a certain extent homicide) is the historically low unemployment rate. As an economist, one would have thought Dr. Leigh would have pointed to that reason as having “saved lives”.

    With Government funding to Universities being increasingly subject to academic results, could this report be a case of playing the tune requested by the Howard Government?

    With 40,000 alcohol related deaths annually, we can only wonder how many lives could have been saved by banning alcohol!

    But that wouldn’t happen, since John Howard enjoys a drink or two, but happens to hate guns.

  23. Peter W. says:

    The excellent paper by McPhedran and Baker, is available on http://www.class.org.au, as is the Alpers/Chapman effort, with critical comments.

  24. harry clarke says:

    As Peter W points out the paper by McPhedran and Baker argues that the suicide rate has declined overall a few years after 1996. If this is true it is quite damaging to your claim that mortality fell after 1996 as a consequence of the gun buyback. Its a point worth investigating. A reduction in the non-gun related csuicide rate cannot presumably be attributed to gun control.

    By the way I embellished my earlier macabre comment on your paper here.

  25. mugwump says:

    Andrew, I skimmed your paper before my original post and did not find it convincing. I will do a more thorough analysis to determine a better estimate. But in any case, it is intellectually dishonest to claim you are primarily attacking BM when you draw conclusions like this:

    “Since the gun buyback cost $500 million, this indicates that even if we use the lowest reasonable estimates of its impact on homicide and suicide (128 lives saved per year), it paid for itself in just two years.”

    And it is equally dishonest to quote an offline paper on the value of a life without the obvious attendant caveats that it is presumably the value of an average life and that you have no evidence to support that the lives saved (if any) would be drawn from the same distribution used to arrive at the $2.5M figure.

  26. mugwump says:

    Ok, I have read the paper more thoroughly. There seems to be some obvious methodological errors.

    Firstly, although not 100% clear from the text, it seems you use the entire series (either 1979-2004 or 1915-2004) to estimate the coefficients of the ARIMA(1,1,1) model, and then use the model to estimate the likelihood of the in-sample years 1997-2004. Shouldn’t you be estimating the coefficients from the out-of-sample data only (ie, up to 1996)? (BTW, the effect of this will most likely reduce further the likelihood of the 1997-2004 rate under the model, ie strengthen your conclusion).

    Secondly, and more importantly, how well does the ARIMA(1,1,1) model predict other parts of the time series. For example, what does a 5-fold cross-validation look like (split the series into 5 equal-length chunks and do 5 estimates leaving out one of the chunks and then measure the predictive power on the left-out chunk)? My suspicion: the model is equally crappy predicting the death rate for any part of the series, hence no conclusions at all can be drawn from this type of analysis.

  27. Christine says:

    Thanks everyone for all the interesting and helpful comments. I missed the real time discussion, so here’s a compendium of comments.

    I am certainly not an expert on all facets of gun laws, nor on ballistics, nor on public health interventions in Australia in the past decade. I am pretty good at statistical analysis, though.

    The main point of the paper we’ve produced is to try to point out that the findings of Baker and McPhedran are unsupportable. It is categorically NOT an excellent paper in terms of its statistical analysis or conclusions, although it does have some very interesting and useful discussion of the law changes.

    Unfortunately, our paper does have an appendix, and a rather long one at that (though if you’d like, you can call it my appendix, which remains intact), which in some detail explains why the method that BM used (and that we used too in this paper) is not particularly helpful in coming up with reliable estimates of ‘lives saved’: because of the impossibility of accounting for method substitution, underlying trends, and weird stuff like the jump in non-firearm suicides between 1998 and 2000 or so (Peter W: this is concerning, but I cannot see how you can possibly attribute all of that jump to the firearm buyback).

    We’d like to have a better approach to estimating the effects of the NFA, but figuring that out is extremely difficult. At the moment, the best we can do is point out that drawing extremely strong conclusions is not a terribly good idea. But if you want to draw some conclusion from this type of analysis, it should be that overall gun deaths – both suicides and homicides – fell post 97.

    In other words: Baker and McPhedran’s conclusion that the gun laws had no effect is simply wrong, even if we use their own preferred methodology.

    On the question of valuation of a suicide death: it’s a bit unusual for people to argue that economists place too high a value on a human life. :) But there is an interesting issue of how to value a voluntary vs involuntary death. A fairly large portion of the valuation on a life lost is from the effects on family members, rather than the individual themselves. As for the homicides (though they’re a much smaller component of lives saved), the valuation would likely be higher, because so many victims are relatively young. We don’t pretend this is a perfect estimate – we went with the standard to try to avoid accusations of fiddling the numbers – but you should note that even if you cut the valuation to $250,000, and use 120 ‘lives saved’, you’re still looking at a (non-discounted) ‘saving’ of $600million over 10 years.

    On the effect of unemployment rates on suicide: I actually ran those numbers, and tried to add in periods of drought as well (in case rural suicides might increase in times of drought, and in case those are more likely to be with firearms than are uban suicides). I found pretty much no relationship at the aggregate level. This sort of surprised me. I do know of a Canadian study (not published) that finds not much effect of unemployment rates on overall suicide, but some effect on men. But this effect is small. I gather this is fairly typical of suicide studies, and for studies of homicide.

    Finally, Peter W.: “With Government funding to Universities being increasingly subject to academic results, could this report be a case of playing the tune requested by the Howard Government?”

    For disclosure purposes: I’m currently completely unfunded by the Australian government, though I used to work for them once upon a time a long time ago. I can attest that no government official or politician has ever suggested that I write anything for them. Andrew’s description of the genesis of this paper is quite accurate, which is pretty easy to see from the blog archives. No need for conspiracy theories here.

  28. Christine says:

    mugwump: we exactly followed Baker/McPhedran, who estimated the model on data up to 1996, then used that to project out from 1997-2004. Sorry if it wasn’t clear from the paper. This wouldn’t be the standard economists’ approach to this btw: we’d normally do something a lot more like the Chapman/Alpers and look for structural breaks. But we wanted to show that using the same methodology, the results can vary quite a bit with only a few minor changes in specification (esp on firearm homicides).

    You’re right: the ARIMA model is crappy (note: standard tests of null hypothesis that the series were integrated tend to reject). Using it from 1979 on is a particularly bad idea, because 1979 was a local maximum year for firearm suicides and firearm homicides, so the I(1) component gives you an estimated downward drift. Combine that with a model specified in levels, and you get predictions of death rates that go below zero, which is obviously not a good thing.

    Basic conclusion: Baker and McPhedran’s use of the ARIMA(1,1,1) model (with, mind you, absolutely no reporting of the test statistics, nor any apparent testing of the I(1) assumption) is not a good idea. But if you do want to use it, the only reasonable conclusion you can draw is that there was a decline in firearm homicides and suicides (but not non-firearm homicides or suicides) post 1997.

    BTW: All the data and code used to run the statistical analysis is available on the web, so anyone else who wants to play with the data can go for it. The predictions were done using the R statistical package, which is also freely downloadable. If you plan to do this, you should note that the homicide (deaths by assault) data are so volatile that different statistical packages are likely to give you different results.

  29. JC says:

    Andrew
    Please explain why gun control didn’t prevent the Monash shootings which if it not for the heroic attempts of a teacher would have been another Port Arthur and a prelude to Virginia Tech.

    Maybe you need to rethink your study and explain just how gun control prevents mass murder which of course was the reason we had gun control in the first place.

  30. JC: why?

    It’s a perfectly sensible thing for economists to look at the consequences of a policy, even if those consequences are not precisely correlated with the stated aims.

    For instance, environmental regulations are not really about employment, but we’d all like economists to study the economic impact of such regulation.

    Whether the changes to gun laws have reduced the number of shooting sprees is an interesting question, but Leigh and Neil aren’t under any obligation as academics to provide an answer in this paper.

  31. mugwump says:

    Christine, I still don’t think you or Andrew are being honest. Andrew’s claim above is very strong:

    “even if we use the lowest reasonable estimates of its impact on homicide and suicide (128 lives saved per year), it [the gun buyback] paid for itself in just two years.”

    No caveats there. This conclusion has been picked up by others, eg John Quiggin, who describes it as “the best estimate” and that you got the “stats right”.

    You have pretty much granted that the model has no skill, although you have not reported cross-validation results so we don’t know for sure. Therefore, the “lowest reasonable estimate” of the effect of the gun buyback by your methodology is surely zero, no?

    BTW, 120 deaths per year multiplied by 10 years by $250,000 is $300M not $600M. And $250,000 is overly generous for a criminal or welfare-dependent, who’s net value to society is likely to be in the negative (not positive) millions.

  32. Franco says:

    Peter W claims that there are 40,000 alcohol related deaths annually in Australia. Table 6.3 at http://www.aihw.gov.au/publications/phe/qdcmma98/qdcmma98-c06.pdf shows that when you factor in lives saved (from cardiovascular risk reduction caused by low to moderate alcohol consumption) there are net lived saved as a result of alcohol use in Australia. If you discount that small problem, and only look at deaths caused, then there are 2524 deaths attributable to alcohol in the study year, only out by a factor of 15.8 from Peter W’s figure. Quite amusing.
    Franco

  33. Paul O'Grady says:

    Your claim that up to 282 lives were saved by the gun buyback is somewhat laughable. How many of these lives were terminated using alternative means of suicide? All your research can be interpreted to say is that there may have been a reduction in the number of gun-related deaths, not that lives were finally saved. Where can I get a copy of your research paper, I’d be interested to read it in its entirety.

  34. john coochey says:

    The paper is arrant nonsense and clearly rebutted by empirical data The conclusions that some 220 lives are saved each year could only be true if that number of people were dying from that king of gun before they were bought back by the Government. I cannot find any figure in the AIC’s figures that approach that figure. In which years were 220 people being killed by such weapons? Also how can a weapon which no longer exists cause death rates to drop over time? Surely any benefit would have been a one off hit? For example if you banned tobacco people would still continue to die for some time from the long term effects. However if you banned all horses accident rates would drop once and once alone, they would not continue to decline unless there were other factors at work.

    It is absolutely innane to dismiss a downward trend because would eventually lead to a negative figure, if you did so you would also have to dismiss the current paper too!

  35. TH says:

    Mugwump:

    Can I suggest, firstly, that you’re being a bit rude: you’ve basically said that you think the authors are lying about something. Not being honest is quite different to making errors in analysis, and you’re accusing them of the former. You should probably be careful when doing so.

    Second, supposing you’re right that “no conclusions at all can be drawn from this type of analysis” when “the model is equally crappy predicting the death rate for any part of the series” (this is your guess as to it’s predictive power). Even if no conclusions can be drawn about the effect on gun deaths, one broad conclusion can be drawn: that those who attempt to draw conclusions about the effect of the policy on gun deaths with such a model are making a mistake :) You’d agree – you already have – and this argument seemed to be the central point of the authors paper, that the BM study isn’t reliable. You’re furiously insisting the authors are wrong while restating the abstract of their paper…

    Third, although I’m no expert, is it not the case that valuations of human lives are concerned with quantifying, for purposes of policymaking, the intrinsic value of a life, rather than the instrumental value (the “net value to society”, as you put it)? If that were the case then a gun victim’s death would still be intrinsically worth a positive amount, say $250K, rather than a negative amount. Also, the statement that most gun victims are criminals or welfare dependent seems like a leap of faith. Finally, lets take your own logic and resubmit it to the policy implication machine – since the gun death of a welfare dependent or criminal adds to the net value of society, there are some interesting and gruesome implications for how we should treat these people (arsenic in the prison food, guillotines at centrelink, etc etc).

    Fourth, suppose that you don’t like the particular figure used to value life, and wouldn’t support the gun policy because you refuse to accept a policy valuation that uses a dollar figure on life. Nonetheless you should still be able to be convinced that the gun buyback was a good idea, by considering the cost and lives saved as compared to a particular hospital intervention. Choose all those medical interventions you believe are justified, compare each in terms of cost and deaths to the gun buyback, and as long as one intervention falls short you cannot logically favour that but not favour the gun buyback (at least on economic grounds). I’m only suggesting this as a thought experiment, of course.

  36. Andrew Leigh says:

    Where can I get a copy of your research paper, I’d be interested to read it in its entirety.

    Try clicking on the title of the paper in the post above.

    Regarding the tone of the comments, I should advise that I’ll start censoring soon. I’m more tolerant of abuse directed to me than towards other commenters, but it’s unnecessary, and doesn’t make your argument any stronger.

  37. Christine says:

    John: nope – to get around the ‘resurrection problem’, we suggest the simple fix of using the logs specification, which by definition won’t let the estimated death rate go below zero. Almost anyone who studies death rates where the numbers of deaths are really really small – like here! – uses either logs or a Poisson/negative binomial framework to get around just this problem. Please, read the paper before criticising it.

    Mugwump: OK typo in comments on a blog! Mea culpa, thanks for the pointer. If you want to continue to argue about the cost benefit analysis, you can, but it’s really all in the two assumptions of (1) $2.5m per life saved, which isn’t a very profitable line of argument, I think; or (2) the 128 figure, which I think is worth arguing about.

    It does come from some statistical analysis we’ve done, and we think it’s a decent ballpark first estimate, although as mentioned the statistical procedure is problematic. Note, though, that any other statistical procedure using the same data is going to find substantial non-predicted reductions in firearm death rates, but not non-firearm death rates, post-1997. And if you throw our results out, you have to throw Baker/McPhedran out too. In fact, you’d have to throw them out first.

    We do in fact show some basic sensitivity analysis and we have reported our key statistical results in the paper (there were none of these in the paper on which we were commenting, btw), and we’ve put up all the data and the statistical commands to get the predictions.

    The data’s all online, play away and report back on any results you get. You may disagree with a couple of lines of interpretation, and with the media spin, but I’d be extremely interested in any criticism of the actual paper.

    Have a look at some graphs of the basic results while you’re at it: http://www.wlu.ca/documents/22258/FinalPredictions.xls

    Not trying to hide anything or pull the wool over anyone’s eyes.

  38. Puddleduck says:

    “And if you throw our results out, you have to throw Baker/McPhedran out too. In fact, you’d have to throw them out first.”

    Chapman et al, as well. They found no significant change in the decline in firearm homicide after 96, same as B/M, but using neg binomial. Seems your paper (which journal has it been published in, btw? I’d like to download the formatted copy) contradicts BM, C et al, and Ozanne-Smith, too.

  39. john coochey says:

    The paper only holds water if a number of criteria are met.

    1 Gun deaths decreased after the Buyback

    2 It can be traced to the that event and not due to other causes

    3 The number of lives saved in your paper are equal or less than those killed by that type of weapon before it was banned. If I have a cure for prostate cancer which I say will save 20,000 lives a year but only half that number die from the malady then there is obviously a problem. (At no time in our history were there 2200 deaths from automatic weapons unless yo count WW1 and WW11)

    4 That there was a distinct drop in such deaths after the implementation of the new policy. Clearly not the case, even a glance at the AIC website shows that if you did not know about the Buyback you would not pick it up from the statisitics. THERE WAS NO CHANGE!

    And once again how can a drop in deaths from 2006 to 2007 be connected to a once off event which occurred in 1996? I could understand an assertion that someone could claim 2006 lower than 1996 but how does this have more than a once off effect? Your paper assumes or asserts that all drops post 1996 were due to the Buyback, this is clearly shown by reference to actual AIC and ABS data

    The paper clearly does not met any of these criteria

    I had a similar correspondence with the authors of the NHMRC paper before they were force do withdraw it in disgrace. And that had withstood the full might of the Catholic Church

  40. harry clarke says:

    I tried asking this before but don’t think it got addressed.

    Did the suicide rate decline a few years after 1996 for non-gun-related suicides as well as gun-related suicides or is this factually wrong? If it did decline then the strength of Andrew and Christine’s results are reduced. The reduction in non-gun suicides cannot be attributed to the gun buyback but needs to be related to something else. There is then no reason that this explanation should not apply to the gun-related suicides as well. Or have I misunderstood something?

    Mugwump, Paul, John, The paper is obviously very interesting and important otherwise we wouldn’t be analysing it. I agree with Andrew – the language you are using is unhelpful.

  41. Uncle Milton says:

    “how can a drop in deaths from 2006 to 2007 be connected to a once off event which occurred in 1996?”

    Because if semi-automatics had no been banned in 1996, they would have been around in every year since, and people would have been killed by them in every year since.

  42. john coochey says:

    Yes that may affect the total overall deaths but not the rate of decline in subsequent years. Which has actually occurred, and furthermore was occuring at a higher absolute rate before the buy back. Unless you think that somehow the Buyback cut back death rates for over ten years before it was implemented, (time travel a la Star Trek?)do your own research I will try and put it as simply as I can. If angle grinders cause two thousand injuries per year the maximum number of injuries that can be avoided by their elimination is two thousand. If a study claims that banning them reduced injuries by four thousand it is obviously flawed. Check? Some more injuries would be expected from substitutes such as hack saws. If eleven years down the track industrial injuries are still declining on a relative or absolute basis then that cannot be due to the ellimination of angle grinders because they have already been eliminated, ten years ago and the accident rate dropped accordingly. Any further drop from the preceding year cannot be due to their elimination, it has already happened some years ago. So any recent falls in accidents or must be due to some other factors. If smallpox is eliminated in 2000 any decline in death rates in 2007 vis a vi 2006 cannot be due to the elimination of smallpox because there have not been any sufferers for five or more years. If you cannot understand that I cannot make it any simpler. Beam me up Scotty there is no intelligent life down here

  43. Paul O'Grady says:

    Mr Clarke – I don’t think there’s anything wrong with the language I used. I’m sure the comments to which you refer were addressed to all bloggers, just happened to be made in the same posting as my question was answered and couldn’t have possibly reflected upon me. In any case, apologies accepted.

    John – just love your comments, especially regarding angle grinder deaths. Said it all for me. A Tardis might be a better device for time travel though.

    Looking forward to reading this paper tonight. Hope you lot haven’t spoilt it for me.

  44. Paul O'Grady says:

    Andrew – was my question about the source of the quotation “Lies, damned lies and statistics” censored or did it not just not come through?

  45. Andrew Leigh says:

    Harry, your point is an important one. This is how we attempted to address it in our paper.

    BM state: “The inclusion of suicide and homicide by methods other than firearm provided a control against which the political, social and economic culture into which additional legislative requirements for civilian firearm ownership occurred could be evaluated, as well as determining the level of method substitution within homicide and suicide.” It is clear, however, that the potential for method substitution invalidates the use of the non-firearm death rate as a control for the firearm death rate. A good control group must be unaffected by the policy change. If the gun buyback caused an increase in non-gun homicides, the non-gun homicide rate cannot be a good control group for the gun homicide rate. For a more formal discussion of the problems that arise when attempting to use non-firearm deaths to look at both substitution effects and as a control group, see the Appendix.

  46. john coochey says:

    Yes that is why we use intertemporal analysis which shows yet again that rates were falling completely independantly of the Buyback. Just go to the ABS and AIC web pages, A Queensland study of about fifteen years ago showed that overall suicide rates actually increased slightly the year firearm restrictions were introduced, plummeted the following year and quickly got back to the original rate with a bit more thrown in to “make up the backlog” which was quickly made up. However the authors tried to disguise this by bracketing years together in groups of two. All goes to show “all the fact that fit we print”.

  47. harry clarke says:

    Andrew, I still don’t get it. What happened to non-gun related suicides after 1996. Did they eventually fall off a lot. If they did your story has problems.

    I am not talking about substitutions but just whether or not other factors drove down the suicide rate.

  48. mugwump says:

    Christine:

    “Have a look at some graphs of the basic results while you’re at it: http://www.wlu.ca/documents/22258/FinalPredictions.xls

    Very helpful, thankyou. I recommend anyone commenting here take a look at that spreadsheet, particularly sheet fsuicide(2), which plots firearm suicide rates since 1915.

    You don’t need to run anything through R to see that the model has captured none of the post-1980 downward trend in firearm suicides. In fact, it is a classic regression towards the mean (the model just wants to shoot back up). The plot is not annotated, but am I correct in guessing that the dependent variable in the model is log death rate?

    BM’s basic conclusion (ignoring their methodological errors) seems to at least be consistent with the data: post 1996 firearm suicides continued an existing downward trend. Certainly that’s a lot more consistent than what the model suggests.

    It also seems clear from fsuicide(2) that no amount of statistical jiggery-pokery is going to be able to resolve the basic question of whether the decline in firearm suicides 1980-1996 was a real trend or just a random fluctuation.

    That said, the two periods on the plot with comparable declines coincide with WWI and WWII respectively, times in which a large fraction of the population likely to commit suicide by firearm were pointing their weapons at someone else, providing some evidence that such steep declines are more likely due to exogenous factors than random fluctuations.

  49. the following is a comment that I posted on a thread over at Kalimna (Harry Clarke’s blog):

    “Actually, that issue about the impact of the buyback on the stock of guns poses an interesting question about its impact on gun suicides.

    I think the buyback only applied to ceretain classes of weapons. I am guessing that these included fully automatic weapons and at least some semi-autoimatic weapons. I don’t think it applied to bolt action rifles, some types of shotguns and some types of pistols. While it probably reduced the overall size of the gun stock in Australia, I suspect that it did so by less than the number of guns that were “bought back” by the Commoinwealth Government. The reason for this is that at least some of those shooters probably replaced the banned guns with alternative guns that were still legal. In addition to affecting the size of the gun stock, the gun buy-back probably also altered its composition.

    I suspect that most (or at least many) gun related suicides do not involve the type of guns that were banned. As such, would we expect a large impact on gun-related suicides as a result of the gun buy-back?”

    The thread can be found here:

    http://kalimna.blogspot.com/2007/04/gun-control-murder-suicide-rates.html .

    The following is a slight elaboration of the above comment. It is difficult to believe that restrictions on what seem to be fairly extreme types of weapons would have altered substantially altered most peoples preferences about ownership of a legal gun. Maybe the decline in gun owenership was a response to a general decrease in the desire of the poublic to own a gun that was caused by the horrible events at Port Arthur?

  50. Peter W. says:

    Franco. Sorry mate, I can’t find the source of my “40,000″, but it was on ABC Radio health Report sometime ago. Perhaps it was an accumulated figure of the past 10 years? Cheers!
    But back to guns. It is well accepted that not all semi-automatic rifles and shotguns were handed in during the buyback. From Customs Import figures (and my memory) there are around 700,000 semi-auto centre-fire rifles imported from China, which still have not been handed in. I still have a s/a shotgun from pre-1996, while the old s/a rifle I handed in was replaced by a new s/a rifle. Therefore the “removal” of all semi-automatic firearms didn’t happen, so can’t be given any credit for any change in firearms deaths.
    Now another point:
    An important factor in crime reduction is keeping criminals locked up. “NSW has 3000 more prisoners in its jails than 10 years ago. Authorities predict the figure will reach 10,000 by this time next year.”(Source: Sunday Telegraph April 15th 2007). The ANU report does not take that factor into account.

  51. There is one point that nobody has mentioned so far.

    Semi-automatic rifles and shotguns only became widely available in Australia from the mid 1970s. I read somewhere that it followed a Customs decision but I don’t have details. Nonetheless, it is true that prior to that time there were mostly only bolt actions available.

    The “buy-back” only applied to semi-automatics (and pump action shotguns). Bolt actions were not affected.

    So why would the data series going back to 1915 be relevant?

  52. Christine says:

    Apologies in advance that I keep doing mega-posts.

    Harry: a picture is worth a thousand words. Have a look at the charts. Roughly speaking, I’d say there’s not terribly much reason to think that non-firearm suicides have tracked firearm suicides in the recent past.

    It’s possible that people see what they want to see in the charts. You could point to the uptick in non-gun suicides in the three years post-97 as method substitution, or gun owners who suicided because they were forced to turn in their family heirlooms, and the later fall (back to where it was in the early 90s) as an underlying downward trend that affected all suicides, even though prior to ’97 firearm suicides were going in the opposite direction from non-firearm suicides. As we pointed out in the paper, no-one can really rebut this story because we have no real control group here, but I think that twisting the data that much is not very helpful.

    Perhaps the best way I can answer the underlying question you have is as follows: if you do a regression with stacked firearm and non-firearm suicide rates, and include a dummy for post-97 and a ‘post-97 interacted with firearm’ dummy, then the former is not statistically significant, and the latter is negative and statistically significant. This is true whether you use a levels or a changes specification, and whether you use ARIMA or OLS, and whether you start from 1915 or from 1979. The evidence on suicides suggests firearm suicides fell substantially more than non-firearm suicides post-97. (This is closer to the approach taken in Chapman et al than in Baker/McPhedran, btw.)

    mugwump: the predictions are all from the models described in the paper, viz ARIMA(1,1,1) models. There’s no imposition of a regression to mean, here, since the I(1) assumption allows for a downward trend. It’s just that unless you use data starting in the late 70s or early 80s, you won’t find a downward trend in either firearm suicides or homicides. Doesn’t work using data from 1915, but it doesn’t even work if you use data from 1969 (see paper). Give me a reason why there should be a structural break in firearm suicides or homicides at that point causing a downward trend, and I’ll think about that. Even then, though, the data points to a decline in firearm homicides and suicides after 1997 that did not occur in non-firearm homicides or suicides – just not as big. So you’ll need a story of something else that changed around 1997 as well that caused firearm homicides and suicides to decline but NOT non-firearm homicides and suicides. (Note: as everyone agrees here, the data is much clearer for suicide than homicide.)

    On the graphs: each graph shows the actual data (red) and two predictions: black line is from the model estimated in logs (predictions tranformed to levels for ease of comparison), blue from the model estimated in levels, and the dotted lines are the 95% confidence intervals.

    Everyone should note that we (and B/M, and Chapman et al) are relying solely on time series evidence. We can’t isolate the effects of the buyback from the effects of the other aspects of the NFA, nor from any general cultural revulsion to guns caused by the Port Arthur killings. I don’t think anyone will ever be able to do that. But if you’ve got a brilliant idea for it, let us know.

  53. I’m a theorist Christine. Why would I want to get my hands dirty with data? Despite what one of my econometrics lecturers used to say, I am not convinced that data are really our friends. I suspect that data might actually be better counted among our enemies!!! Their habbit of pouring cold water on nice stories is particularly frustrating. Or it would be if we actually paid them any heed!!! ;)

    Seriously though, I was not implying that you should have tested for cultural changes arising out of the terrible events at Port Arthur. I realise that that would be a very difficult, if not imposasible task. My comment was just expressing some thoughts that occured to me as a result of the discussion on this thread and the one over at Kalimna. For the record, I think thast Andrew and you have done a very nice, and much needed, job with this paper. Highlighting the fragility of the results from these types of models to sensible respecifications is an important contribution to debate.

  54. mugwump says:

    Christine: “Give me a reason why there should be a structural break in firearm suicides or homicides at that point causing a downward trend, and I’ll think about that.”

    I don’t need to give you a reason for a structural break. On the contrary, the onus is on you to show that the model has predictive power, since you are making the claim that the probabilities assigned by the model to the firearm suicide rates post-1996 are significant.

    It seems clear from fsuicide(2) that the model cannot capture long-term trends. What do the model predictions for 1987-1996 look like? I expect, like post-1996, the model assigns very low probability to that portion of the series. Likewise the downward trends from 1940-1945 and 1915-1922.

    Unless you can show that model predicts well out-of-sample, its predictions for post-1996 have minimal relevance for rejecting the null hypothesis: “the reduction in firearm suicides post-1996 had little to do with the buyback and simply continued an already existing trend”.

  55. Christine says:

    Damien: there is a point with type of firearm, but there’s not very much data that helps us figure that point out, sadly. And thanks for actually getting the biggest point of the comment.

    mugwump: this is starting to get tedious.

    1. You do need a reason. Otherwise more data is preferred to less. As I’ve said, it doesn’t make enough difference anyway if you start in 1979 (at least not for firearm suicides).

    2. I can’t give you a model prediction for 1987-96 that’s out of sample because it’s in-sample. Unless, that is, you want to estimate an ARIMA model from 7 years of data (79-86). I would advise against.

    3. No pure time series model is going to adequately predict the up and down swings. Given that this was a comment on a paper that used only TS methods, we thought it appropriate to stick to those. To have any hope of explaining the swings, you’d need some co-variates. As mentioned, I did look for some anyway, and found nothing much interesting (this includes droughts, unemp rates, WWI and WWII and GDP growth, by the way).

    4. The model gives precisely equal weight to all observations. But don’t take my word for it. Do your own regressions with the data, or check the R code that’s posted at:

    http://www.wlu.ca/page.php?grp_id=1418&p=8153

    5. The link below has some predictions using the 1915-96, 1969-96 and 1979-96 models in log and level terms, with start points each decade from 1919. Which one would you say is a more accurate representation of long run trends in firearm suicides? (Note: I know this isn’t an entirely fair comparison, and I know eyeballing isn’t exactly a reliable statistical test, but since most of the complaints are based on eyeballing thought I’d help out with a bit more.)

    http://www.wlu.ca/documents/23851/STATA_predict_rfs_level.xls

    6. “The reduction in firearm suicides post-96 … simply continued an already existing trend” is a poor null hypothesis if you’re talking levels. But whether in logs or levels or neg binomials, that hypothesis has been refuted – by BM, by Chapman et al, and by us. That’s not really up for discussion any more. The only thing up for discussion there is the underlying reasons for the decline post-97.

    BM’s statistical analysis of firearm suicide death rates showed a greater than expected decline in firearm suicides after 1997, but not in non-firearm suicides. From this they concluded that it was impossible to say that firearm suicides fell after 1997 because non-firearm suicides had fallen too. This conclusion may be consistent with what you want the data to show, but it is certainly not consistent with their own statistical analysis.

    While I’m happy to do my best to answer serious questions, I am neither your teacher nor your research assistant, and I have students who are clamouring for their term grades. But by all means, do some work of your own, get your facts straight, and then we can talk.

  56. mugwump says:

    Christine, your tone is patronizing and offensive. Obviously, as an anonymous poster I can claim no authority, but for the record I have a PhD in mathematical statistics, and an extensive publication record on the subject (with an H-index of 18 if you care), so I think I might know what I am talking about.

    As to your points:

    1. I don’t need a reason for a plausible null hypothesis: “the trend continued”.

    2. You can give model predictions for 1987-96 by training the model on 1915-1985.

    3. Exactly my point: the model doesn’t model the data, so how can you draw any conclusions from it?

    4. I never questioned the weights given to observations.

    5. Models trained from 69 or 79 are not interesting: way too noisy (as you have pointed out). The ones trained from 1915 show exactly what eyeballing the data tells you: the unusual downward trend began in 1987, not 1997.

    6. I just don’t know how you can make this claim. Look at the data from 1915 onwards. A steep decline began in the late 1980s and has continued since. The decline did accelerate marginally in 1997, but not significantly.

    Maybe what you are doing is standard practice in your field, but that still doesn’t make it right. The simplest refutation of your argument is to compare the predictions made by a model trained up until the late 1980s and one trained until 1997. If the model trained to the late 1980s assigns very low probability to 1987-1997, how can you draw any conclusions from the probabilities assigned to 1997 onwards?

  57. mugwump says:

    I should add: my comments above are directed at the firearm suicide rate, a 1915-2004 plot of which is here:

    http://www.wlu.ca/documents/22258/FinalPredictions.xls

    (sheet fsuicide(2))

    Here’s a simpler way to cast my point. Suppose I told you I was modeling the data with a very simple predictor: P(death-rate = x) is proportional to
    exp( -100 * (x – 1.0) * (x – 1.0) ) if x is greater than 0, and p(death-rate = x) = 0 otherwise.

    Such a predictor would assign vastly higher probabilities to the fsuicide(2) series post-1997 (where the death rate is close to 1) than it would to any other part of the series. But you would quite rightly reject any conclusions drawn from my predictor, precisely because of its lack of predictive power on other parts of the series.

    For the same reason, I don’t see how you can draw strong conclusions from your ARIMA(1,1,1) predictor unless you can show that it is able to predict significant features of the time series (which of course it cannot).

  58. ChrisPer says:

    I reiterate the point Jeanine Baker made, that the only reason that Baker and McPhedran had to be written was the pathetic analysis which had been done until then.

    It is really sad that having pointed out the weaknesses of the method used, Andrew made an extreme claim of benefits that cannot be justified on his data. Note that the costs of the Government in the buyback are only about half the costs the community suffered.

    And where is the analysis of what actually caused any changes? What motivational mechanism, what individual utility/cost analysis has re-balanced the equation for mass killers? What exact mechanism in banning rabbit rifles like the Ruger 10/22 affects mass killers?

    It is a few days short of the anniversary of Port Arthur, and we just had a horrific mass killing in the US which led to an unbelievably bad torrent of media incitement to further crimes. If the behaviour models that seem to have most power in this are correct, we can expect more killings to come soon.

  59. Christine says:

    mugwump: be fair, your tone hasn’t been particularly pleasant at all times. And I was obviously not arguing from authority: I said ‘teacher or research assistant’, thus allowing you the potential status of my student to my boss. I was arguing from having actually done some work with the data. And I’m getting irritated because you seem to be almost deliberately misunderstanding me.

    On the evidence of your comments, I can only conclude that you haven’t read the paper itself. It is very very largely a comment on Baker and McPhedran. Let me quickly restate the points we made, though:

    1. Starting in 1979 is an awful idea. In your comment above, you seem to agree.

    2. ARIMA(1,1,1) is a poor modelling strategy, and in this case is likely biased in favour of finding no post-97 change, esp when combined with (1). Safe to say you agree?

    3. Using levels instead of logs or some other transformation that ensures you don’t get predictions of negative death rates is quite frankly just plain stupid. You haven’t commented on this, but if you’re a mathematical statistician I can only assume you’re fine with it?

    4. With a pure time series methodology you simply can’t separate out method substitution and underlying trends, although BM claim that they can. You seem to agree there too.

    5. BM’s statistical results are quite frankly very sensitive to changing 1 and 3 (we didn’t change 2 because we thought it wasn’t fair to them, but it’s also very sensitive to that). Their conclusions as they stated them simply do not hold given the data. I think you agree with this?

    6. Given small changes to the most obvious flaws, and BM’s ARIMA(1,1,1) model (which makes it harder to find any effect of the gun laws on firearm homicides or suicides), the best estimates are a reduction in death rates that would be equivalent to on average 120ish – 280ish lives not lost per year (note: point estimates only! Lower bound is not lower end of 95% CI, though that’s still positive – still, there’s a possible gripe for you). You seem unhappy we put any numbers on it at all.

    7. Our overall conclusions were: “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.” Your last statement explicitly agrees with the first part, though not with the second.

    So: you don’t like point 6. Is that all? If so, then I don’t see why you’re arguing with me. Go argue with Baker and McPhedran, and tell them you agree with Leigh and Neill on their substantive points. But good luck to you in getting any useful response from that quarter. We emailed Dr Baker asking for her data, and she refused to provide it, though we didn’t press her too much. We got the data via Jenny Mouzos at the AIC (and many thanks to her for that).

    On your comments to my earlier numbered points:

    1. Which trend? In logs or levels? Why are you using levels? Accounting for the fact that death rates were abnormally high in the early 80s or not? It’s actually not that easy.

    2. Yes, I can do that. Thankyou for making that clearer. Obviously randomly picking sub-periods and seeing how well a model estimated on that sub-period fits the next is a particularly bad idea in the case of a series with a significant AR component (definitely the case with firearm suicides) and a possible I component (cannot be rejected by the Dickey-Fuller tests) is not particularly helpful ESPECIALLY if you estimate the model for a sub-period starting at a local maximum or minimum point, during which the trend is always up or down. Any ARIMA(1,1,1) model is going to predict a continuation of the trend in that case, which is obviously completely unlikely to occur (see graphs). Estimating the model over the longer series, with a couple of ups and downs, on the other hand, will give you a model that predicts pretty much what happens with the series (see graphs). In other words: I don’t think doing this would be particularly valuable.

    3. Argh! See points above.

    4. “I expect, like post-1996, the model assigns very low probability to that portion of the series.” Perhaps I misinterpreted you? Normally I’d have said that “the model assigns low probability” to something means that it attaches a low probability weight to the data over that period. Your clarifying post seems to suggest that this interpretation is correct. If so, I must say, huh? x is never smaller than zero, if x is a death rate. I did try to figure out your point there, but I’m afraid I couldn’t quite. And again, we didn’t do any spooky transforms along the lines you suggest. I directed you to the code as evidence of this.

    However, it is also possible that you meant “the model would have failed to predict … “. If so, I can say the 1915 model and only the 1915 model, would predict a decline in firearm suicides any time there has previously been an increase in firearm suicides. Check out the 1915 graphs, btw: every one of them shows the increase in firearm suicides in the early 80s was a ‘surprise’ on the upside, and ‘predicted’ some decline in firearm suicides following that upside surprise.

    5 and 6. I disagree. The predictions from the 1915 model show an unusual uptick in firearm suicides in the early 80s followed by a reversion to more normal levels until the mid 1990s. That was followed by a drop in firearm suicides to unprecedentedly low levels, which has not been reversed. The longest previous drop in firearm suicides was from 1940-1945. The firearm suicide rate has been considerably below its long-run level from 1996 to 2004. But I suspect we’re going to see whatever we want to see in the graphs – which is why we tend to rely more on statistical analysis than eyeballing. Here’s one more set of pictures for you, though, this time just the raw data in log terms (this was one of the first things I did when I looked at the data last November or so). What does eyeballing this tell you?

    http://www.wlu.ca/documents/23859/Graphs_logs.xls

    Again: Our paper, Baker and McPhedran, and Chapman, Alpers, Agho and Jones all find a statistically significant drop in firearm suicides post-97. McPhedran appears to have made recent media statements to this effect too. You seem to be the only person fighting over this point, and you’re doing it on the basis of eyeballing. If you really wanted to pick on a weak point, you’d pick on firearm homicides. Or question the weirdness in non-firearm suicides around the time of the NFA and just after. Or talk about what the underlying causes of the decline were, rather than just focus on the data. Really, really, not the statistical evidence on the change in firearm suicides after 97.

    (BTW: just checked, and found that cause of death stats for 2005 are out. Don’t have guns figures, but the overall suicide rate fell slightly from 10.44/100,000 to 10.34. Suicide by hanging rose slightly, so suicide by methods other than hanging fell from 5.47 to 5.08, definitely their lowest levels in the past decade. Death by assault rose slightly from 0.82 to 0.98, but this is still lower than any other year on record, I think. Not much room for an uptick in gun-related deaths just yet. I gather the data aren’t exactly perfect, but I’d say that adds another data point on post-97 exceptionalism).

    I’ve done my best to answer your questions. I’ve done my best to figure out where we disagree and explain why I think we disagree. All the data and code is available for you to replicate (in the narrow sense of the word, or in the broader sense of the word) what we’ve done.

    All this without having the foggiest idea who you are. If you wanted to remain anonymous on the web, you could of course have emailed me and let me know privately who you are. All my contact details are quite publicly available.

    Are you happy yet? And would it be too much to ask for a brief thankyou for having spent several hours trying to satisfy your curiosity?

  60. john coochey says:

    My two critical questions still remains unanswered.
    1)Why was there not a sudden vertical drop, in absolute terms, in gun related deaths in 1997?

    2) How can a reduction in deaths in subsequent years, as opposed to continuing at the new “lower level” or continuing previous trends from this new “lower level” have been caused by an one off removal of guns in 96?

    Unless these questions are answered then the rest is irrelevant and the study worthless.

  61. mugwump says:

    Christine, my beef is and always has been with Andrew’s statement:

    “Overall, our paper estimates that 128-282 lives have been saved every year by the gun buyback. In other words, 1000-2500 Australians who are alive today would not be here if it hadn’t been for the buyback.”

    I claim this conclusion is baseless. You and Andrew admit ARIMA modelling is dubious (which I completely agree with, including BM’s approach), but then you draw strong conclusions from it anyway.

    (BTW, using the lower limit on a confidence interval rather than the mean or an upper limit does not increase the likelihood that your conclusion is correct. If the model is bad then none of the numbers coming out of it are believable. If you disagree with this statement, what is the difference between your ARIMA model and my silly exponential model above?)

    You suggest I have not read your paper. That’s not true. I admitted early on to having only skimmed it, but since then I have read the entire thing, I have downloaded and played with original source data from the ABS, and since my last post I have read the Chapman paper.

    FWIW, the Chapman statistical analysis seems sound. Their approach of measuring trend rates seems right to me, as the data seems to be telling us that there was a structural shift in trends post-buyback.

    However, the Chapman analysis also raises its fair share of questions which they do not address. Observations/issues:

    1. Figure 1C shows that firearm homicide excluding mass murders continued its pre-existing trend post-buyback. This is exactly what I would have expected (and of course, it is a very good thing that we’ve expunged mass-murders from our society. Whether that was worth the price is a matter of opinion).

    2. Figures 1E and 1F show that a downward blip in firearm suicides in the 2 years post buyback were offset by an upward blip in nonfirearm suicides in those years, suggesting substitution.

    3. After the initial upward blip, non-firearm suicides reversed their pre-buyback trend and started declining. What’s the explanation?

    4. I have only dug out data back to 1994, but it seems there may have been substitution of firearm suicide by hanging (this is not discussed in any of the studies I have read).

  62. mugwump says:

    test

  63. mugwump says:

    Christine, I posted a response but it has been deleted, so I guess there is no point continuing this conversation here.

  64. Andrew Leigh says:

    Mugwump, the Akismet spam filter caught your comment, and I just retrieved it. I have so far deleted one rude/abusive comment in this thread, but I’m certainly not trying to stop polite debate.

  65. Jeanine says:

    I have been informed Dr Christine Neill claims she asked for the data Dr McPhedran and I used in our BJC published paper and this request was refused. I have enclosed a copy of my response to Dr Leigh on this subject, in which I trust I made it clear that, because we did not use the historical data apart from to provide a historical context, our preference would be that these data be sourced from the original source (ABS or AIC). Given that these are publically available data I believed, and still do, that it was pure courtesy to the orignal collator of the data that we should redirect anyone with an interest in obtaining it to the original source. The raw data used in our analysis was contained within the paper. I have not received any communication, written or verbal from Dr Neill, at any time. Nor did I receive any communication that either Drs Leigh or Neill were having trouble obtaining the historical data, which probably would have led us to supply it. It certaily appears obvious that they had no trouble obtaining it from the original source. :)

    As to the rest – let me know when Neill and Leigh have undergone peer-review and had their work published.

    > —–Original Message—–
    > From: Andrew Leigh [mailto:andrew.leigh@anu.edu.au]
    > Sent: Friday, 27 October 2006 5:57 AM
    > To: jb@ssaa.org.au
    > Subject: Your British Journal of Criminology paper
    >
    > Dear Dr Baker,
    > I am an economist at the Australian National University, interested in
    > the results in your recent paper in the British Journal of Criminology.
    > I am writing to ask whether you would be willing to share the full
    > dataset with me. Table 1 of the paper shows the data for 1979-2004, but
    > I would be most grateful if you could also send me the data for
    > 1910-1979, as shown in Figures 1-2.
    > Thank you in advance for your assistance.
    > Yours sincerely,
    > Andrew Leigh.
    > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    > Dr Andrew Leigh
    > email andrew.leigh@anu.edu.au
    > address SPEAR, RSSS, HC Coombs Bldng, Australian National University ACT
    > 0200
    > phone 02 6125 1374
    > fax 02 6125 0182
    > mobile 0431 706 600
    >
    >
    >
    > —–Original Message—–
    > From: Jeanine Baker [mailto:jb@ssaa.org.au]
    > Sent: Sunday, 29 October 2006 7:17 PM
    > To: jb_ssaa@internode.on.net
    > Subject: Re: Your British Journal of Criminology paper
    > Dear Mr Leigh
    >
    > Thank you for your request for the full data set used in our paper. As
    > you would be aware, figures 1 and 2, provide the historical context only
    > and were not used in our analysis. The raw data used in our analysis are
    > included in the paper.
    >
    > The data that provided the historical trends, like the data used in the
    > analysis (which represents the lifetime of the National Homicide
    > Monitoring Program), are all publicly available from the Australian Bureau
    > of Statistics or could possibly be sourced from the Australian Institute
    > of Criminology.
    >
    > Regards
    >
    > Jeanine Baker

  66. Christine says:

    ChrisPer: I assume that if the only reason Jeanine Baker wrote the paper in the first place was to get something out there, and that she doesn’t think too much of her own work that she’s admitting to the flaws now? Oh, no, wait, she’s not:

    http://www.ic-wish.org/Neill%20and%20Leigh's%20Weak%20Tests%20and%20Strong%20Conclusions%20Revisited.pdf

    And earlier:

    http://www.ssaasa.org.au/pdf/Baker-McPhedran_critique_of_chapman_et_al.pdf

    Got do admit, they’re fast: I’m very impressed, though I do think they would have done better to sit back and think about it. I know taking a bit of time (and getting some advice from some very wise people) improved our comment a lot, and helped to ensure we weren’t being unreasonably critical of BM. They don’t seem to have bothered with that. To say that “N&L may prefer to conclude that the gun laws had an unquestionable effect than to admit the possibility that the laws may not have had an impact … ” is, I think, a bit rich. Consistently saying we give no justification for claims that are documented with statistical results in the paper, and with the data and code on line, is also a tad unreasonable. (I’m getting a little annoyed with the suggestion that we should have included more specification tests and statistical results in our comment paper than were included in the original paper.) Calling our arguments “disingenuous”, saying we “tellingly overlook” stuff or take an “evasive position” and saying that omitting a discussion of mass murders “raises serious questions about the object and intent of the authors” (!) is actually offensive.

    Perrhaps I’m naive, but I do actually take this academic objectivity and evaluating the evidence carefully stuff quite seriously, and worked and am working bloody hard to be as open and honest about all of our analysis as possible. I truly am not sure what else I can do on that front.

    Here’s a thought: perhaps one reason no relatively neutral person wants to get involved in this issue is that you get hit by stuff like that, and some of the comments above? (Not to mention extraordinarily rude emails from John Coochey.)

    OK, so that was a bit of a rant. I would like to say, though, ChrisPer, that I’ve always appreciated your comments. And I hope you’re enjoying the debate, even if you’re not that keen on my position so far. Again, hopefully we all come out of this having learned something we didn’t know beforehand.

    mugwump, you can appreciate that I’m not going to comment substantially on Chapman et al. I too prefer their technique to BM (that would be why we haven’t taken a hatchet to it – no mysteries there!), though there are some dilemmas, including allowing both trend and levels changes in 1997. It’s asking a lot of the data we’ve got at the moment to separately identify these things. Doing that works against finding any effect, btw.

    I must also point out that there’s a lot of dispute on whether we should be looking for a structural break in level or change terms. John Coochey (above) seems to be very unhappy in attributing a change in the rate of trend decline to the NFA. Others say that unless there’s a change in the trend rate of decline then the NFA didn’t do its job. Hard to satisfy everyone there.

    The lower CI thing was actually to point out another possible point of criticism of our numbers. Showing all the cards.

    All of the studies mention the possibility of method substitution (and switching to hanging would be most logical because it’s the most commonly used method). That’s one reason for checking on the non-firearm suicide data.

    I talked about the up-blip in non-firearm suicides in an earlier post.

    Alright, I think that deals with your latest questions.

    Now then: I’ve written a lot (I type fast) trying to answer your questions, but you don’t seem to ever get around to directly answering my questions. Do you agree or disagree with points 1-7 in the post above? What is the point of your exponential model? What do you think about the pictures in log form?

    I am sorry to say this, but I haven’t learned very much from you. Other commenters have been helpful even if critical (ChrisPer, Sinclair Davidson, Harry Clarke, David Berry, Robert Merkel and Damien Eldridge (!!!!), eg). But honestly, you don’t really seem interested in my responses, or in an actual discussion. I’m really going to have to stop responding to you if it’s just a one-way street.

    Right, that’s enough, off to mark again.

  67. Ren says:

    Good heavens – if Christine got it so very wrong about contacting J.Baker for the data, what else is she getting wrong? Perhaps that explains her over the top defensiveness?

  68. Andrew Leigh says:

    Ren, how on earth did Christine get anything wrong? She wrote:

    We emailed Dr Baker asking for her data, and she refused to provide it, though we didn’t press her too much.

    Jeanine Baker evidently misread “we” as “I”.

    And as Jeanine Baker’s publication of our email exchange makes clear, her answer to me was a clear “no, we won’t give you the data”. It’s wonderfully convenient for her to now say “oh, if only you’d asked twice”, but that doesn’t change history.

  69. Pingback: Deltoid

  70. john coochey says:

    I would be happy to publish the two emails that I sent to Christine but they were in the same tones and covered the same, still unanswered, questions as are on this blog. How can anybody claim that comparing two years data in neither of which automatic firearms were present that any decline in deaths between these two years was due to the absence of such technology? It is a simple as that. A public health campaign for example anti smoking or driver training may well bring about effects many years after they have ceased but you cannot be killed in 1998, or whenever ,by a weapon which was destroyed in 1996 so any subsequent drop in death rates cannot be due to the absence of automatic weapons. They were also absent in any earlier periods post 1996.

    It has to be due to other factors.

    If Gun Control Australia is correct and one in four weapons in Australia were destroyed, and such weapons were responsible for a large number of deaths, then basic logic states that there should have been a marked one off reduction in such deaths. It did not occur.

    In the years when large numbers of Chicom weapons were flooding in from China gun related deaths should have gone up. In fact they went down.

    Don’t get offended by simple questions, simply answer them, if you have an answer

  71. Andrew Leigh says:

    John, Christine has addressed your point here and here.

    Also, I can’t speak for Christine, but you’re welcome to publish the email exchange that you and I had on Monday. If you do so, please post all our emails (you sent me 4, I sent you 4), without any changes.

  72. john coochey says:

    Yes but where? I have made specific questions, illustrated them with metaphors which are understandable to everyone I have spoken to and nowhere has Chistine responded to them. In the name of all that is holy how can a gun that has been destroyed in 1996 kill someone in 1997 but not in 1998. If the gun was destroyed it cannot kill anyone anymore in any future time period. How can weapons which have been destroyed continue to kill people, it is the time machine problem again. If death were reduced from 96 to 97 then that is it! No more. If it were decided to remove heavy vehicles from the road in 96 to reduce accidents then any reduction from that action would occur in 97. Any reductions from 2000 to 2001 could not be due to the disapearanc of heavy truck because they have already been removed. They do not exist in either of the time periods. It is like saying deat rates in Darwin have been reduced from 2000 to 2001 because the Japanese are no longer bombing the city. They did not bomb the city in either year. Since 96 there have been no legal semi autos in the community so how can any reduction in death rates be due to the reduction in such weapons? What information have you presented that there has in fact been any reduction in the availability of such weapons since 96? So far as I can see, non!

    Where is the massive reduction in gun related deaths that should have occured in 97. It is not in the references you give and could not be because it is not in the raw data.

    How many time do I have to ask the question?

  73. Simon Chapman says:

    John, can I refer you to my post above (April 22) where I emphasise that the main change that one would have predicted (hoped) from the 1996 gun law reforms would have been a diminution in the incidence of gun massacres. The laws removed some 600,000 rapid firing semi-automatic guns from the community. It did not remove single shot weapons that were still around in abundance to be used in suicides (only one shot is required) and in the sort of homicide incidents we so often read about where husband shoots his wife or an acquaintance. There was no real reason to expect that these sorts of incidents would descrease after the 1996 reforms. Together they comprise >90% of all gun deaths.

    But there was every hope that mass killings would decrease. Someone wanting to randomly kill lots of strangers in a public place cannot do this easily with a single shot weapon, but even an inexperienced, inexpert shooter like Bryant can do it with semi-automatic weaponry. Howard said repeatedly that the introduced the laws to stop Australia going down the American road. Everyone understood that to mean we didn’t want a culture like that which repeatedly produced Virginian incidents.

    In our Injury Prevention paper, we used the anology of a government committed to reducing the road toll at large, seeing a spate of level crossing train/car smashes and legislating for barriers. In sum, these might not count for a lot of the total road toll, but most would agree that it you can prevent such incidents, then you should. The obvious test of the barriers effeciveness would be the future incidence of such events, not the impact on the total road toll.

    If that went down as well, that would be a bonus, but may well reflect other factors too — as many have been saying about the reasons for the ever-declining gun death rate in Australia.

    What happened? There have been NO mass shootings in >10 years, when there were 13 in the 18 years before. The rabid end of the gun lobby cannot bare to say this. They avoid it like the plague and keep talking about the Childers arsonist killer.

    This is like saying “don’t ban barbiturates, because people can just as easily hang themselves.” This is such a puerile argument. Guns are a particular problem because of their ultra-lethality. I’ve never heard of a drive-by stabbing. Plenty of people who swallow (non barbiturate) pills in suicide attempts survive. Very few people who shoot themselves live, and most people running amok with semi-automatic weapons kill many people before they are finished.

    The myopic passion in this present discussion about total gun deaths going up or down is a total red herring.

    Simon Chapman

  74. Simon Chapman, if I understand you correctly, you are suggesting that Neill and Leigh can only explain their purported saving of lives in terms of massacres that would otherwise have occurred.

    For what it’s worth, I see that as more rational than the argument presented by Neill and Leigh. There was no reduction in total guns (the semi-autos were replaced by legal guns), so there could not have been any impact on firearms suicide and little or none on typical firearms homicide.

    On that basis Neill and Leigh’s estimate of lives saved is quite excessive. There has never been anything like that many killed in massacres. .

  75. Christine says:

    John, do please go ahead and publish the emails and the responses. As you know, I responded only to your first email. The second email was simply rude, and I deleted it.

    I have tried to deal with your questions here, however. Again, I think I have previously answered your big concern, even though “when I eyeball the data it doesn’t look to me like there’s anything much there” is not much of an argument, and I doubt that you’re planning on changing your mind no matter what evidence I or anyone else present you with. But let me try one more time:

    You seem to want a model that allows for: a) a constant trend in death rates from, say, 1979 and on past 1997; b) a break in the level of the death rate after 1997. On that assumption, please see:

    http://www.wlu.ca/documents/23908/OLS_trends.xls

    I did this just for you, trying my very very best to figure out what model you were thinking of. For disclosure purposes: since you don’t fully specify your model, I did two things you may not like. First, I use logs. You don’t say whether you want logs or levels, but for reasons that I’ve argued about extensively, I think that the latter is unsupportable for statistical reasons, so I went with the logs. Second, I do a very simple sensitivity analysis for the start date. I start by estimating the model in 1979 and go back in one year steps until 1975.

    The results basically show a statistically significant and very large ‘one-off’ drop in firearm suicides, an economically significant, but not always statistically signifiant drop in firearm homicides, and perhaps some evidence of some drop in non-firearm homicides, although it is not statistically significant in general, and is somewhat smaller in percentage terms than the drop in firearm deaths. Non-firearm suicides just don’t budge. I’ve got some more detailed interpretation in the spreadsheet, to save people from reading more of my stuff on the blog.

    If you don’t like my modelling strategy, go try your own, and please post your results and explain why they are better at describing the data than the ones I’ve did for you.

    David, FWIW, I personally think that Simon’s point on mass murders is a strong one. I think he could back it up a bit more strongly by doing a simple calculation of the probability of no mass shootings occurring in Australia over 10 years under various assumptions of the underlying probability of a mass shooting occurring. Otherwise, if another mass shooting were to happen (let’s hope not), then I think he’d be met with howls about how that showed he was wrong, which would be extremely unfair.

    But I can’t really speak to this question directly, because we didn’t look at it. We didn’t look at it because … once more … our paper was a comment on the particularly bad statistical analysis of Baker and McPhedran. Also, of course, we didn’t have the data on mass shootings – I think the collection of that information is a really valuable contribution.

    I disagree with Simon that the ‘total deaths’ thing is a red herring, although I also think you can make the argument that the buyback alone was worth it if it stopped the mass shootings. I think that’s the sense in which you’re using the term ‘red herring’, Simon? (If you’ve been beaten up enough for only caring about victims of mass shooting and not about the deaths in ones or twos – which is pretty much what happened to Simon in a radio interview

    http://www.abc.net.au/rn/lawreport/stories/2006/1776336.htm

    - well, then I can completely understand why you’d get upset every time someone says something like “yes but it can’t have affected total deaths”.)

    But I also think that the accompanying legal changes around the NFA could have had an impact of their own on smaller scale events, and it’s useful to consider that possibility. I note that Chapman et al did indeed allow for that in their analysis, so you can’t criticise them for not having explored the possibility, even if they didn’t think there’d be much to see.

  76. cba says:

    john coochey said:


    My two critical questions still remains unanswered.
    1)Why was there not a sudden vertical drop, in absolute terms, in gun related deaths in 1997?

    2) How can a reduction in deaths in subsequent years, as opposed to continuing at the new “lower level” or continuing previous trends from this new “lower level” have been caused by an one off removal of guns in 96?

    a few points:

    1. I suspect the direct effect of the gun buyback was small.

    2. The relevant policy change was NOT the buyback of guns, but rather making several classes of guns illegal and simultaneously tightening gun ownership rules. The buyback solved the problem of how to compensate owners for once legal purchases of now illegal guns.

    3. A plausible explanation for the gradual decline in gun deaths in the data rather than a one off reduction is that the new laws encouraged gradual behavior modification. a few reasons for those changes to be gradual might be:
    a. the new laws affect new owners of guns more than past owners, so you will see their effect only as the ranks of new owners swell.
    b. % of gun owners in the population (not number of guns) is declining due to the law changes/community pressure
    c. the gun owning community increasingly self regulates its members for fear of even more regulation.

  77. mugwump says:

    Christine:

    “Alright, I think that deals with your latest questions.”

    They were more observations on Chapman – I wasn’t expecting you to comment.

    You have still failed to deal with the main question: how can you justify making a claim of a minimum 128 lives a year saved by the gun buyback with a model that utterly fails to model the data?

    That’s the point of my exponential model: it also utterly fails to model the data, hence I would not be justified drawing any conclusions from it. Likewise, you are not justified drawing any conclusions from your ARIMA model.

    There are two steps to statistical modeling: model selection and parameter estimation (well three: interpretation). If the model suck it’s “garbage-in, garbage-out”. You are not justified drawing any conclusions from a lousy model.

    I agree with Simon Chapman: at this point about the only definitive thing we can conclude is no more mass murders.

    As for your points 1-7: they are details that don’t address the main problem (the model sucks, hence all conclusions drawn from it are baseless, including your “minimum 128 lives per year saved”). However, since you seem offended that I have not responded to them, here goes:

    1. 1979 is not a great place to start but in the context of a lousy method I wouldn’t particularly single it out as awful (it’s all pretty “awful”)

    2. Agreed.

    3. Negative death rates is a bit of a red herring. It only really matters for predictions close to zero. Away from zero the model may tell you something useful, and will differ little from the predictions of a truncated model that just sets p(death rat e “In other words, 1000-2500 Australians who are alive today would not be here if it hadn’t been for the buyback.”

    The claim is baseless.

  78. mugwump says:

    Last post had a formatting error. Repost.

    Christine:

    “Alright, I think that deals with your latest questions.”

    They were more observations on Chapman – I wasn’t expecting you to comment.

    You have still failed to deal with the main question: how can you justify making a claim of a minimum 128 lives a year saved by the gun buyback with a model that utterly fails to model the data?

    That’s the point of my exponential model: it also utterly fails to model the data, hence I would not be justified drawing any conclusions from it. Likewise, you are not justified drawing any conclusions from your ARIMA model.

    There are two steps to statistical modeling: model selection and parameter estimation (well three: interpretation). If the model suck it’s “garbage-in, garbage-out”. You are not justified drawing any conclusions from a lousy model.

    I agree with Simon Chapman: at this point about the only definitive thing we can conclude is no more mass murders.

    As for your points 1-7: they are details that don’t address the main problem (the model sucks, hence all conclusions drawn from it are baseless, including your “minimum 128 lives per year saved”). However, since you seem offended that I have not responded to them, here goes:

    1. 1979 is not a great place to start but in the context of a lousy method I wouldn’t particularly single it out as awful (it’s all pretty “awful”)

    2. Agreed.

    3. Negative death rates is a bit of a red herring. It only really matters for predictions close to zero. Away from zero the model may tell you something useful, and will differ little from the predictions of a truncated model that just sets p(death rat e less than 0) = 0 but otherwise makes the same predictions.

    4. Never say never, but I agree that the Chapman et al approach has much better chance of identifying trends and substitution than ARIMA.

    5. Agreed.

    6-7. Yes, I am unhappy that you put any numbers at all on the reduction in deaths. My beef is not with BM, since they did not make your claim:

    “In other words, 1000-2500 Australians who are alive today would not be here if it hadn’t been for the buyback.”

    The claim is baseless.

  79. john coochey says:

    It does not matter what model you use. Wapon destroyed cannot kill people in subsequent years so any reduction after the immediate one cannot be due to removal of such weapons from the community. Any model that incorporates destroyed weapons killing people in earlier periods but not later ones is obviously in error. The only reason to look at figures post 97 is, given no noticeable drop in deats, would be to see if that year was somehow anomolous. There was some exogenous shock which neatly put deaths back where they were. But did not put them up above previous levels. This is not the case and trends have continued as before. I do not have the suicide figures in front of me but was there later a “make up factor” to catch up on the backlog, as happened in Queensland

    So are we talking abou “massacres” or not? If so in what year did massacres reach 128 let alone 282? In fact there was one area of Australia which did not ban such weapons, at least for some time after 96 and possibley even now, but I bet you cannot pick which area it was. There may have been no mass shootings in Australia but there have not been any in New Zealand either. That is the poblem with social analysis, the absence of a control.

    But if we are talking about mass shootings why did you not say so at the outset? But then we would be back to the problem of more people allegedly being saved by the Buyback than were actually being killed before it took place

  80. Christine says:

    mugwump, thanks. So we mostly agree.

    BM did in fact make a claim that the number of deaths ‘saved’ was 0. This is a pretty precise number too.

    My points 1-7 are actually the main points of our paper, so I guess we differ on whether they’re detail or not. But so far we’ve got:
    1. Agree (reluctantly?)
    2. Agree
    3. We agree that if the death rate is close to zero, this is a big problem. The firearm homicide rate is was 0.16 per 100000 people or around 30 individuals. The firearm suicide rate was 0.84 per 100,000 in 2004, or around 170 individuals. (The 1000-2000 fewer deaths figure is over 10 years.) These are low enough that the bounded by zero problem clearly matters. So I’m going to sum this one up with: Agree.
    4. Agree.
    5. Agree.
    6-7. Well, we had to put in some numbers, just to give some idea of the scale of the differences between various model estimates. And I’ll quote back at you again from our conclusion: “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.” If it’s of any interest, we did try to avoid headlining of the numbers, but it’s not terribly easy to manage.

    Still, we’re agreeing on every methodological point. This seems to me like rather an accomplishment.

  81. Christine says:

    cba: Nice explanation.

    John: try reading cba’s entry again – it might help a bit. And did you look at the stuff I did just for you? Do you have any comments on it? Do you think it is a decent summary of the data? Also: we are NOT talking about massacres in our paper (Chapman et al did, though, which is why the issue came up). If you haven’t figured that out yet, then I really don’t know how anything I say or do for you can help.

  82. mugwump says:

    Ok, I can agree to disagree :)

    It is only the number that I have ever objected to, because it is exactly the thing people are going to pick up on to justify their various political positions. No one (other than we geeks) gives a rat’s proverbial about the methodology.

    I believe BM are far closer to the truth. My impression from Chapman and looking at the raw data is the buyback has prevented 0 deaths other than those in mass-shootings. Is that worth it? Personally, I think so, but then I am not a shooter.

  83. Christine says:

    Yep, you got me, definitely a geek. Admittedly, if you’re going to debate Dickey-Fuller statistics and exponential models and the like, there’s no credible denial.

    Personally not a shooter either, but as a Queenslander with a lot of family from out back, I suspect there’s a branch of my family will not be particularly happy with this stuff going out under my name. C’est la vie.

  84. Is that worth it? Personally, I think so, but then I am not a shooter.

    I am a sporting shooter and I don’t consider it was worth it. Not that I don’t care about the possibility of massacres – I care a lot. However, the risk of being killed in a massacre is lower than the risk of a car crashing into my house and killing me, so a sense of proportion is required.

    The gun laws affected me personally because it wiped out my sport. I used to shoot “Service Rifle”, a target shooting discipline using military rifles and probably the most popular form of rifle shooting internationally. However, since military rifles were mostly semi-automatic, Service Rifle is no more in Australia.

    True, there were other shooting disciplines available that I could (and did) take up. However, that’s comparable to being forced to play soccer because the oval balls used in rugby have been banned.

    As an ordinary taxpayer I also think the use of $500 million in taxpayers funds for the possible prevention of a small number of deaths, the fact of which is disputed by some very clever people, might be just a bit unjustified. What if the same funds had been used on mental health measures?

  85. john coochey says:

    Simon’s points, or rather hypocracy, have already been dealt with on a recent Life Matters broadcast where Wetherburn described him a disingenous for now claiming that it was only “massacres” that were important when he had an article published in the Sydney Morning Herald claiming all reductions in gun related deaths were due to the buyback. The SMH article is available on the web.

    I seem to recall an incident revealed in a case by the Tobacco Institute where he had recommended concealing research which had shown that the deaths form side stream smoking should be concealed.

    Once again neither your paper nor McPhedran explains how weapons continue to kill after they have been destroyed but kill at a lower rate over time. It apears the danger of guns is inversionally proportionate to the length of time since they were destroyed. To call that illogical would be a euphemism. If they are not avialable in either of two time periods any change in mortality cannot be caused by something that did not exist in either time period. Once they are destroyed they cannot kill anybody in any of the future time periods. There is no escaping that fact.

    Incidentally a brief look at AIC data show firearm related suicides were higher at the end of 97 than before the buyback.

  86. Franco says:

    David, you are suggesting that your freedom to spray bullets about a rifle range using military weapons for your pleasure is more important than having a community where we no longer open a newspaper and routinely read about the latest disaffected person — often a licensed shooter with gun club history (Thomas Hamilton at Dunblane) mowing down a hall full of kindergarten kids.

    I’m quite sure there are many people out there who think it would be great fun to own dynamite & muck about with like minded individuals blowing up sides of mountains; or ground-to-air missile launchers that they could practice shooting targets dropped out of aircraft; or armed tanks that they’d love to hoon around in working out their General Paton fantasies.

    Do you think people should be free to have access to this sort of gear too? I’m quite sure 9999/10,000 of these people would be harmless, but what a price to pay for the 1 in 10,000 who isn’t.

    Franco

  87. ChrisPer says:

    Franco, perhaps you might contribute to the debate more if you looked at your words and removed the ones which are laden with emotional value judgements.

    The gun debate is full of people who think they are in some way better people than others because of the opinions they hold, so we get two kinds of communication mixed together. One is moral posturing which belittles the other side and helpfully displays what a wonderful person the writer is becasue of the right opinions they hold. The other actually has content, including factual observations, interpretations, reasons and objections.

    Its very hard to pull the good stuff out and restrict yourself to constructive comment, but I find it always worthwhile to try.

  88. Paul O'Grady says:

    Franco – For some reason, buybacks and amnesties don’t seem to work on getting fireams out of the hands of criminals, just the law-abiding owners who have done the right thing. Even Howard commented in 1996 that the new firearms restrictions wouldn’t stop another “massacre”. I suppose that seeing research such as this he will now act the hypocrite and take the credit, such as it may be. I would have preferred our money that was spent on the buyback to have gone into public health. Even if there are 282 lives saved, as this research purports, how many more would be alive had the millions been spent instead on hospitals?

  89. Franco says:

    ChrisPer, sorry, could you clarify which words you would like removed? David was saying he values his freedom to shoot military weapons. I thought I was asking a quite reasonable question. We don’t allow people the freedom to access dynamite, ground to air missiles or tanks. Why is that do you think? Where do we draw the line? What principles are drawn on in such judgements?

    Paul, I’m not sure what statement of Howard’s you are referring to, but I’d not be surprised if he did say that. Just as I’d not be surprised if the decision to double the number of booze buses was accompanied by a political statement that these would not “stop” all drink drivers. The question of course is whether such measures reduce these events. Nothing gurantees to eliminate them. Howard and all the states were plainly of the view that the laws would reduce massacres, as they indeed seem to have done.

    And Paul, you seem to be suggesting that gun violence is committed only by “criminals”. They are all criminals after the event, but as I’m sure you know, many people like Hamilton, Bryant and the recent Virginia shooter had no criminal history. We were all wise after the event that they were “strange” and shouldn’t have been given access. Easily said.

  90. john coochey says:

    I promised to publish the correspondence that Christine found so offensive. Here it is

    Yes Christine but what is your response to the questions I raised?

    Your paper only holds water if a number of criteria are met.

    1 Gun deaths decreased after the Buyback

    2 It can be traced to the that event and not due to other causes

    3 The number of lives saved in your paper are equal or less than those kille by that type of weapon before it was banned. If I have a cure for prostate cancer which I say will save 20,000 lives a year but only half that number die from the malady then there is obviously a problem

    4 That there was a distinct drop in such deaths after the implementation of the new policy.

    And once again how can a drop in deaths from 2006 to 2007 be connected to a once off event which occurred in 1996? I could understand an assertion that someone could claim 2006 lower than 1996 but how does this have more than a once off effect?

    You have clearly not met any of these criteria or am I missing something

    Her response to a previous communcation was made on This was sent on Tuesday the 24 April Tuesday at 2.14 PM

    Christine Neill wrote:
    Thanks for your comments. It is never unkind to criticise academic papers, and all comments on the paper itself (but please, not the media reports!) are very welcome.

    Regards,

    Christine

    That original correspondence from me read

    “I am afraid I am having some problems with your paper on the gun buyback The conclusions that some 220 (sic) lives are saved each year could only be true if that number of people were dying from that kind of gun before they were bought back by the Government. I cannot find any figure in the AIC’s figures that approach that level. In which years were 220 (sic) people being killed by such weapons? Also how can a weapon which no longer exists cause death rates to drop over time? Surely any benefit would have been a one off hit? I can understand why gun related deaths may be lower in 2006 than 1996 but why should they be lower in 2007 due to something that was banned in 1996? For example if you banned tobacco people would still continue to die for some time from the long term effects. However if you banned all horses accident rates would drop once and once alone, they would not continue to decline unless there were other factors at work. However your study asserts or assumes that all reductions in gun related deaths are due to the buyback, obviously incorrect.
    Would it be unkind to say that unless these questions are answered your paper fundamentally flawed?”

    The only editing has been my acknowledgement that I used the wrong figure for gun related deaths at “220″, I should of course have used the figure 282
    There has been no other correspondence betwem us other than on this blog. However in my haste this morning, doing two things at once I said in part concerning Simon Chapman’s objectivity

    “I seem to recall an incident revealed in a case by the Tobacco Institute where he had recommended concealing research which had shown that the deaths form side stream smoking should be concealed”

    That should have read “recommended concealing research which showed that the deaths from side stream smoking were minimal.” I am a militant anti smoker but that does not entitle me to conceal research which does not support my personal moral viewpoint. Having re read the McPhedran paper I should apologise for accusing her of making the same errors that Christine and Simon made. If in fact the on going reduction in gun related deaths is due to the general legislative package post 96 rather than the Buyback as suggested by CBA then why spend half a billion dollars on the Buyback? If we accept his (or her?) argument then that legislation has been more effective than buying some 700.000 guns and at minimal cost. Although I am baffled as to how gun clubs would induce their members not to undertake mass murders. Once again and I am perplexed I have to keep repeating this point, the disapearance of guns from the general community may or may not have had an immediate, once off, effect on gun related deaths but it defies logic that it had an ongoing effect in further reducing levels.

    To carry it to the logical conclusion, if by some unknown means all guns were removed from Australia in 96 and smelted in the sun, no one being able to secrete weapons in back paddocks or whatever, then there would we assume be a reduction in deaths. But any further reduction in later years could not then be due to a reduction in firearms because no further reduction, from zero, can be possible. My neighbour’s kid can understand that but not it apears highly paid academics

  91. Paul O'Grady says:

    Franco – Despite your reluctance to believe he could say such a thing it was said. Check NSW Hansard, 15 May 1996, where the PM – bless his soul – is quoted: “I do not pretend that these recommendations can stop another massacre”. Great political mileage comes from it though.
    I didn’t make any such suggestion about gun violence being committed only by criminals, must be your misinterpretation. My point, easily understood really, is that firearms are significantly more likely to be used illegally by criminals rather than the law-abiding yet no government of any persuasion has ever done much to remove the villains’ tools of choice. It’s not so much of a long bow to draw to assumme then that no guns in the hands of criminals equals so many fewer victims of homicide.

  92. Franco says:

    Paul, I guess this is a perfect illustration of how those with an agenda can choose to take succour from words as meaning whatever they want them to mean. Howard said ” “I do not pretend that these recommendations can stop another massacre”. Everyone knows that Howard and the other leaders introduced those gun laws in the wake of Port Arthur, as an attempt to prevent civilian access to rapid firing weapons that could kill many people like Bryant did. Anyone who doesn’t believe this is in total denial.

    As I said earlier, no piece of legislation can guarantee that it will stop every incident that it is designed to stop. If after 9/11 Howard had said of increased airport security ” “I do not pretend that these recommendations can stop another hijacking” would you rush about saying “Howard says he’s not trying to prevent another hijacking?” Please, let’s get real here …

  93. Simon Chapman says:

    This is a digression from the subject, but John Coochey has made defamatory remarks about me which I cannot let pass.

    John, I’d suggest that you get your facts straight before making potentially actionable remarks about my scientific conduct (“concealing research”). I see from googling you that you have made these remarks in public before (http://www.mensrights.com.au/page13z2.htm).

    You are recalling tobacco industry spin put on a fax I sent to other members of a National Health & Medical Research Council committee I was on in 1995. I wonder if you have ever actually read what I wrote? Here it is: http://legacy.library.ucsf.edu/tid/fvp71d00

    In my fax, I raised two concerns about an early draft chapter of a report we were writing on the health effects of passive smoking in Australia. The draft contained a table that showed estimated deaths in Australia from passive smoking exposure calculated for age and sex bands. The numbers shown were expressed as fractions of deaths. My fax argued that “fractional” annual deaths (such as 7.6) would prove difficult for journalists and the public to understand. For example, I would be surprised if many non-epidemiologists would be able to decipher what “0.5″ deaths per annum meant (it means one death every two years).

    There are many perfectly correct ways of expressing the same data in more comprehensible forms, and my fax urged that we should realise that the table as it was in the draft would cause unnecessary confusion. I subsequently argued in the committee that we should recast the data in a more understandable way. The claim that I urged that it be “concealed” is grossly offensive and wrong, as the final report revealed. It is standard procedure for all draft papers to undergo changes and editing. Often these are to improve clarity.

    In my fax I pointed out that our very conservative methodology estimated there to be some 93 annual deaths from ischaemic heart disease caused by passive smoking in Australia, whereas a recent American estimate had put the corresponding US figure at 62,000. I advised the committee — correctly – that our report would be therefore “out of step with every international review’s conclusion on this subject”. In fact, our final report included the same very conservative estimates which resulted from our only considering domestic (spousal) exposure data in people who have never smoked. We did not factor in workplace exposures, nor deaths among ex-smokers (who would have had massive passive as well as active exposure during their smoking histories).

    My fax rehearsed the sort of questions that we were likely to get from those who were familiar with the much higher US mortality estimates. If I had really wanted to conceal results why then would we have persisted in using our ultra-conservative methodology which was guaranteed to produce low estimates of deaths which would not have helped the anti-smoking agenda?

    Will you please show me in the fax where I sought to “conceal” data? I would ask you to withdraw the remark.

  94. john coochey says:

    My recollection was of an email that was passed to me via a thoracic surgeon who is also a militant anti smoker and which I should have in my records however the reference you gives clearly states

    “I am deaply concerned about the implications for the credibility of our whole report arising from the calculations on…what I am getting at here is that a reseonable calculation will be that there is any lung cancer caused by ETS in Australia will be seen as a huge joke”

    Guess that says it all. Am I correct that this document only became public after the discovery phrase of a court trial? By the way it have you worked out how guns which have been destroyed continue to kill people but do so at a lower rate over time? This seems to be a central pillar of some of your earlier work.

  95. Simon Chapman says:

    Sorry John, it does not “say it all”. What you said was that I had “concealed research”. Where have I done that? You were saying…?

  96. ChrisPer says:

    Franco,
    People can access ‘dynamite’ by doing the course and passed the police clearance for a Shotfirers’ Licence. It isn’t a big deal in the right professions.
    This is a good example where your post is rhetorically strong but factually weak.

    Choice of words frames your meaning beyond the basic definitions of the words. “Freedom to spray bullets around a rifle range” is hyperbolic codswallop, just the start of you framing this legitimate activity as dangerous and morally questionable. ‘Spraying bullets’ is journalistic hype, unconnected to sport shooting. Service Rifle is a normal, well-disciplined activity conducted in rigorous safety and your words are therefore a falsehood.

    Claiming that mass killers are ‘licenced shooters with gun club history’ is another falsehood. Hamilton’s history was the gun club wanting nothing to do with him, expelling him and notifying the police but a senior police official protected him from having his guns removed.

    The whole argument about tanks and explosives and ‘acting out General Patton fantasies’ is entirely unconnected from any real position of any person here except yourself. It is what we call the ‘atomic bomb fantasy’, taking a position to false and ridiculous extremes, a strawman argument misrepresenting the position of sport shooting advocates. The great-grandfather’s heirloom Winchester 1903 .22 rabbit rifle that was taken and destroyed was no a weapon of mass destruction, except in your imagination, and service rifles can at least be discussed as legitimate civilian possessions as they are in many countries – eg Switzerland and New Zealand.

    Franco, your posts seem to me framed in contempt and inflammatory meanings. I don’t know what kind of milieu you hang around in, but you need to get out more.

  97. Paul O'Grady says:

    Sorry Franco, I still don’t get your point and clearly mine are too intelligent for you to grasp. It certainly isn’t me with an agenda. What the PM said is clear aknowledgement that restrictions on the ownership of certain types of firearms will not prevent future multiple killings and calls into question his motives for the banning of semi-automatic firearms. In any case, this is supposed to be a forum to discuss the results of the research, not for debating the merits or otherwise of gun control. Let’s get back on topic.

  98. David Wilson says:

    john coochey wrote:

    Incidentally a brief look at AIC data show firearm related suicides were
    higher at the end of 97 than before the buyback.

    Eh? What data are you referring to? According to the AIC’s
    report
    no. 116
    on trends & issues in crimes and criminal justice, the number
    of firearm suicides was 347 (a rate of 2.17 per 100,000) in 1996, and 331
    (a rate of 1.79 per 100,000) in 1997. According to their
    report no.
    162
    it had dropped further to 235 (a rate of 1.25 per 100,000)
    in 1998, the first full year after the end of the buyback.

  99. David Wilson says:

    I wrote:

    … the number of firearm suicides was 347 (a rate of 2.17 per 100,000) in 1996, …

    Oops, I miscopied that number. It should have been 384. But that makes the drop in firearm suicides from 1996 to 1997 even larger.

  100. Franco says:

    Gee, Im sorry to have offended you ChrisPer. I genuinely had no idea that an expression like “spraying bullets” could give such offense to shooters. I’ll try to be more sensitive to such blasphemy in future.

    My point is a serious one though. Below is a transcript from a CNN program, describing freely available fearsome weaponry in the US, probably capable of bringing down an airliner. Would you and your associates like these to be available to law abiding citizens in Australia? If not, why not?

    Franco

    PAULA ZAHN NOW

    Aired August 26, 2005 – 20:00 ET

    PAULA ZAHN, CNN ANCHOR: Good evening, everybody. Thanks so much for joining
    us.

    ZAHN: The Fearsome 50, it’s the most powerful rifle you can buy legally.

    UNIDENTIFIED MALE: Wow, right through it.

    UNIDENTIFIED MALE: Right through it.

    UNIDENTIFIED MALE: Right through it, baby.

    UNIDENTIFIED MALE: One-inch steel plate.

    ZAHN: Just a few clicks and some cold, hard cash, no background checks, no registration. We bought one. In this age of terror, who else is in the
    market?

    - – - – - -

    ZAHN: So who outside the military needs a gun so powerful it can pierce
    armor from more than a mile away? The New York Police Department, for one, which we learned this week has spent about $20,000 on two 50 caliber rifles.
    And guess what? If you have the cash, you can get one, too. Yes, it’s
    perfectly legal to buy a rifle that fires with such force that some people
    fear it could actually bring down a plane. Don’t even think about what it
    can do to a person. Investigative reporter Drew Griffin looked into it for
    us recently.

    (BEGIN VIDEOTAPE)

    DREW GRIFFIN, CNN CORRESPONDENT: To buy a gun, even a .50 caliber gun, this huge gun, you just need to go to your computer and click on one of the biggest classified gun sites, which in our case is GunsAmerica.com. AK 47s, shotguns, pistols, all kinds of rifles. But what we wanted to buy was the biggest caliber rifle you could possibly buy, and that’s this category right here, big .50 caliber rifles. This is the gun that is now banned in
    California, and on this Web site we have about three dozen of them for sale

    But what we’re looking for is one that is not being sold by a dealer. See
    where it says federal licensed firearm dealer? We’re trying to find one
    that’s being sold by just a private citizen. This is actually the gun we
    bought. When you finally find the gun you want on this Web site and you’re
    dealing with a private party, you just give him your e-mail and you send him a note. Let’s set up a meeting. I’m paying cash. And the next thing you
    know, we’re going to buy our gun.

    (voice over): But before I shelled out $2,500 to buy this gun, I wanted to
    make sure I could buy ammunition. That turned out to be as easy as ordering flowers. With just a couple of clicks on my computer, I ordered and paid by credit card for 50 .50 caliber armor- piercing rounds. They were delivered in a week, shells as long as my hand, delivered no questions asked by UPS. I could have even ordered tracer rounds if I’d wanted. Now it was time to get the gun.

    (on camera): What we’re about to do is perfectly legal in dozens of states
    where cash and carry is the rule. A private seller, a private buyer. There
    will be no background check, no government waiting period, no government paperwork at all. In fact, the only paper that will change hands is the money we used to buy our .50 caliber rifle.

    (voice over): The transaction at a house in suburban Houston took about 20
    minutes. We walked out with a case holding the gun critics say is the
    perfect terrorist weapon. A brand new .50 caliber with scope, bipod and
    directions. We flew home. Guns are checked as baggage. And when the bags arrived for our flight, I simply picked it up and left. Ronnie Barrett, who manufactures .50 caliber rifles believes as an American it is your right to own one.

    Isn’t that particular gun in the hands of a terrorist dangerous?

    RONNIE BARRETT, BARRETT MANUFACTURING: We’re not talking about terrorists. We’re disarming here civilians. These laws have nothing to do with terrorism.

    GRIFFIN: Barrett’s company makes one of the most popular and top of the line .50 caliber rifles on the market. A semi-automatic favored by armies around the world. But Barrett says his company couldn’t survive on military orders alone and what keeps all these workers busy is its popularity among
    recreational shooters. Barrett says it may be effective on the battlefield,
    but on the target range, it’s just plain fun.

    Should there be any regulations on your guns?

    BARRETT: There should be regulations on criminals.

    GRIFFIN: Gun control advocates want a federal ban on this weapon. Their
    reasoning? Anything that can hit a target at 1,000 yards with the bullet the
    size of a small artillery shell could certainly pose a major threat to
    aircraft.

    (on camera): But the question at most of the nation’s airports is not what
    you could do with a .50 caliber gun at 1,000 yards. Here at LAX, a would-be
    terrorist could get within less than 1,000 feet. (voice over): This week at
    a police gun range, I found out what this gun could do to the emergency exit door of a Boeing 727 fired from 1,000 feet away. The gun is very heavy, not easy to maneuver, but took only a few moments to set up. The first time I fired it, I missed. After adjusting for the site, round after armor-piercing round went straight through the door. But just about any gun could pierce the thin aluminum skin of an airplane. What scares law enforcement is what else this round can do when fired from this gun. This is a one inch thick piece of steel plate. More protection than almost any armored car. The .50 caliber goes right through the aircraft door and right through one inch steel.

    UNIDENTIFIED MALE: Wow. Right through it, baby.

    GRIFFIN (on camera): That’s where it came out. That’s where it went in. One inch steel plate.

    UNIDENTIFIED MALE: Unbelievable.

    UNIDENTIFIED MALE: Just unbelievable.

    UNIDENTIFIED MALE: My goodness.

    (END VIDEOTAPE)

    ZAHN: Force, amazing. Drew Griffin reporting for us. So the question we all
    had is, could a bullet from a .50 caliber rifle actually bring down a
    commercial airliner? Probably not. But it could be enough to cripple one.
    And that was enough of a threat for California to ban sales of .50 caliber
    rifles.

    http://transcripts.cnn.com/TRANSCRIPTS/0508/26/pzn.01.html

  101. ChrisPer says:

    Franco, now that is just rude – posting whole articles is an abuse of people’s comment space. Courteous and experienced net users post an extract and a link.

    And FWIW, journalism like this is just scaremongering hype. Its main effect is to create resolve in people to own what is being criticised, and a lot of decent people are buying these long-range target rifles specifically because the moral poseurs like this journalist are trying to ban them.

    Now here is how to post an article:

    “A few months before the Port Arthur Massacre, a glamorous current affairs presenter showed Australia how easy it was to get the guns used in massacres. She demonstrated every step, emphasising that she knew nothing about guns and had never held a licence. She filmed the guns lying on the street and cut in similar scenes from after massacres. An activist showed her (and us) how to load and shoot the guns. She showed us how these guns ‘designed for killing people’ are easy to use, then blew apart a target like the head of a victim.

    Then a gun control activist offered a key to worldwide infamy: “We are going to have a massacre in Tasmania like those elsewhere!” He sold the massacre in Tasmania as certainty, almost destiny.

    After seeing this show, Allen Burrows travelled from Victoria to Tasmania, bought a gun as he had been shown by the program, and killed himself. Coroner Ian Matterson found he acted on a script provided by that current affairs segment. ”
    ….
    “This single media story did not alone cause the Port Arthur massacre, but the evidence is persuasive that it was part of the cause. The news culture probably caused not just one, but a string of massacres. Our broadcast and print journalists, with their activist sources, gave Martin Bryant the idea, showed him how and offered huge rewards – incentive to murder 35 innocent people.”

    http://www.class.org.au/ideas_kill.htm for the whole article.

  102. Franco says:

    This is now getting very silly. ChrisPer, you are plainly unwilling to engage with any uncomfortable question asked of you, preferring to correct language, guess at character, offer nettiquete instruction and digress down an off-topic path. I’ll ask one one more time: do you and your associates support the availability of guns as described above to citizens? Rocket launchers? Sporting use of explosives? If not, why not?

  103. ChrisPer says:

    Franco, you seem to me to be playing a manipulative game.

    There is a line drawn over such things and that line is set by legislation. Typically around the world it is set to restrict ‘weapons of war’ and ‘destructive devices’ above a certain calibre and of certain functional types. Machine guns are typically banned for civilians, including heavy calibre ones such as the M2 .50cal.

    However, a single shot target rifle firing the same calibre as a machine gun is still a single shot target rifle. The first shot I ever fired, aged 15, was an Australian-made heavy target rifle over almost a kilometre range, in the traditional rifle club movement known as fullbore. There is no functional difference between that and the kind your article was hyperventilating about. In the right context Australians could do well to use such a rifle, and I mean target matches or destroying large feral animals in mountainous station country. Of course, it would be better as a semi-auto than a bolt but that would be unlikely to occur.

    I recall a student when I was in uni whose job was shooting herds of buffalo out of a helicopter with an SLR; these kinds of situations do actually arise.

    Many of our restrictions are more about the movies and what makes ordinary people worried than actual dangers. Silencers were banned when they became popular in movies; same for flick knives. Silencers are legal in the UK and New Zealand, and are very functional for reducing noise impacts. Machine guns have not been used for crimes in Australia ever, even though they were legal in most states until WW2 and in Tasmania until 1996.

    And we ban semi-auto .22s. Such a piece of futility, those are the world’s most popular rabbit rifles yet hyperventilating activists and journalists got them banned, even though they almost certainly ACTUALLY CAUSED the Port Arthur massacre.

  104. ChrisPer says:

    I clarify: the activists and journalists ACTUALLY CAUSED the Port Arthur massacre, by offering rewards and instructing the mad how to do it.

  105. john coochey says:

    In response to David Wilson it is true that suicides using guns dropped by 53 from 96-97 oveall but page 4 of the AIC publication http://www.aic.gov.au/publications/tandi/ti116.pdf shows a distinct increase at the end of 97. (page 4) Which is what I stated. But once again which of the previous suicides used automatic versus still legal weapons? If they utilised non automatics then issue regarding the buyback are irrelevant. They were legal before and legal afterwards so the Buyback is irrelevant. As a comparison gun related suicides from 94-5 also fell be 51. A figure statistically insignificant from 96-7. Obviously not due to the Buyback, unless we once again confront the time travel problem .

    And once again although I never thought I would be in the same corner as Simon Chapman, (particularly given his apparent dismay when his own research shows “avoidable deaths of Australians are lower than we thought” which I thought that would be a cause for celebration, not dismay) Why do you need more than one shot to suicide successfully” But once again, the eternal question which no one has answered, How does a gun that has been destroyed kill anyone in future time periods?

  106. Andrew Leigh says:

    Since the signal to noise ratio on this thread is rapidly asymptoting towards zero, I’ve closed it.

  107. Pingback: Andrew Leigh » Blog Archive » The RPG Buyback

Comments are closed.