Thanks once again to @simonr916 for engaging in debate on this issue. I think we can both agree that the academic literature in this area is not clear cut, that those producing it are seldom impartial (on either side) and that quoting different gun research papers at each other is unlikely to move the debate forward in a way that would convince either side.
The first of the remaining issues is our disagreement about the correlation, or lack thereof, between the homicide rate and gun ownership data.
Perhaps the most obvious point that I have not previously made is that the burden of proof rests with those who wish to assert that there is a correlation and not the other way around. The base presumption must be that A & B are not correlated unless strong evidence is adduced that they are.
@Simon916 has two main arguments
1. The data that I used to demonstrate a lack of correlation is not good data
2. The fact that no correlation is demonstrated is invalid because of failure to control for other variables
I will address these points in turn:
@Simon916 attacks the 2007 small arms survey as being an imperfect guide to cross national gun ownership levels. This is of course perfectly valid, there is no such thing as perfect data for cross national gun ownership.
However the Small Arms Survey, despite its flaws seems to me the best data we have:
” The Small Arms Survey is an independent research project located at the Graduate Institute of International and Development Studies in Geneva, Switzerland. It serves as the principal international source of public information on all aspects of small arms and armed violence and as a resource for governments, policy-makers, researchers, and activists”
It is widely quoted, by non-pro gun sources, such as the Guardian and the UN
If, arguento, this data is not robust enough and there is no better data, then there is no argument with which to refute the null hypothesis that there is no relationship between gun ownership levels and homicide rates.
In the absence of reliable contrary data we would retain the null hypothesis.
If @Simon916 (or anyone else) can point me to more robust data that does show a correlation, then I would of course have to re-examine my conclusions.
@Simon916 does not raise any specific concerns about the data for gun ownership across US states, which was based on the Behavioral Risk Factor Surveillance System (BRFSS). The numbers as reported in the Washington Post can be seen here. I of course accept that this data is also not perfect, no form of survey data can be, but once again the best data I can find shows not the slightest hint of a correlation.
The burden of proof is not on those denying a correlation it is on those wishing to assert one. The alternative would be the same as somebody saying we should ban chocolate because it causes cancer and using as an argument the fact that there is no reliable data to prove that chocolate eating does not cause cancer!
Failure to Control For Other Variables
It is certainly important to control for other significant influencing variables when working out the scale of a relationship. The problem here is that how you define and select the controlling variables can have a significant influence on the outcome. If you have a political agenda in your research or are funded by a group with a political agenda it is possible to get data to say almost anything you want!
Controlling for other variables is most useful when you have a strong initial correlation between the two factors under investigation and there is a concern that your predictive variable may be overstated in importance because it is highly correlated with something that is already known to be predictive. (E.g. For homicide rates controlling for poverty seems to significantly reduce the importance of inequality)
For there to be zero initial correlation between two variables and a significant correlation after controlling for other variables would require a near perfect negative correlation between the predictive variable and the control variables, which seems a highly unlikely state of affairs!
If @Simon916 can point me to any papers in any, non-politicized, field where an initial zero correlation between two variables becomes significant after controlling for other variables then of course I would be prepared to reconsider my view on this point.
@Simon916 glosses over what I think is the strongest evidence for the lack of any significant correlation:
If Gun Ownership were a significant factor in homicide rates then scholars investigating the relationship of homicide to other things, such as poverty, would need to control for it in their research. These scholars probably have no political interest in the gun control debate and will only control for variables that have a real influence.
This paper from the British Journal of Criminology reviews 47 cross national studies of the relationship between poverty and homicides. Conveniently for our purposes it also lists the variables that the researchers controlled for in each study.
Not one of them controlled for gun ownership!
If scholars without a vested interest don’t regard gun ownership as a variable worth controlling for in their own field of research, I think it is safe to assume it is not important.
@Simon916 responds with:
“My answer to this is that I don’t know and it would be difficult to reach any
meaningful conclusion without looking back at each of those reviews. Frankly, I don’t have the time or the inclination.
It may be that some or all of these reviews looked at a range of countries within a geographical area which happened to have similar gun laws, therefore rendering this issue pointless as a control. It may be that the bulk of these studies, which are from decades ago, pre-dated many of the gun laws and restrictions which have since been put in place. Given that the conclusion of the article is that nearly all of the studies failed to control for one of the most important factors – poverty – it may be that they also
missed another important one.”
This paper is a review of ALL the leading research in the area. A look at the paper shows that many of the papers cover over 100 countries, many were published in the 1990’s or 2000’s and they cover a mind boggling array of control variables, to list a sample from just one paper:
Typical settlement size
Change in moral codes
Change in trad. authority
Change in subsist. occup.
Largest settlement size
If the entire scholastic community in all its research into homicides and inequality do not consider it worth controlling for gun ownership then it simply cannot be significant.
If @Simon916 (or anyone else) can point me to research papers investigating homicide rates (not primarily related to gun ownership), that do control for gun ownership then of course that would weaken this point.
The rest of the points raised I think can be condensed into the fact that we agree, (from a utilitarian perspective if not a rights based one) that the issue boils down to deciding whether the additional murders caused by gun ownership are greater or smaller than the murders avoided by gun ownership.
On the additional murders side of the equation we have:
The subset of criminals who commit murder who would not have substituted an alternative weapon for a firearm
A subset of previously lawful individuals (Such as school shooters) who would have decided not to committed murder by a substitute method without access to a firearm
The subset of criminals who kill with a firearm, who would not have carried a firearm if their potential victims were unarmed and would not have killed them by alternative means
On the murders avoided side of the equation we have:
Deaths avoided because an armed civilian prevented a crime that would have escalated to become a murder
Deaths avoided because potential criminals who would have committed murders in the course of their crimes decided a life of crime was too risky and did something else.
We have different views on the relative scale and importance of each element in each group. In most cases they are very difficult to measure and in many cases impossible.
The zero correlation between gun ownership and homicide rates both internationally and within US states in recent years would seem to indicate that the balance between these factors is, in reality, quite close.
A strong positive correlation would support the anti-gun lobby and a strong negative correlation would support the pro-gun lobby. Neither is apparent in the data.
However, there are two remaining factors that go against gun control.
- It is not enough that something has a neutral effect to ban it, even under utilitarian principles. A clear harm must be shown by those seeking the ban.
- The fundamental reason that the US constitution sought to allow citizens to own firearms was not to deter domestic criminals but to provide a deterrent to tyranny and state genocide.
The numbers killed in state genocides of an unarmed civilian population in the 20th century, dwarf, by orders of magnitude, the firearms murders in civilian cases.
In the USA in 2011 there were 8,583 firearms murders, compared to:
# 1915 – 1917 Ottoman Turkey, 1.5 million Armenians murdered;
# 1929 – 1953 Soviet Union, 20 million people that opposed Stalin were murdered;
# 1933 – 1945 Nazi occupied Europe, 13 million Jews, Gypsies and others;
# 1927 – 1949 China, 10 million pro communists;
# 1948 – 1952 China, 20 million anti communists;
# 1960 – 1981 Guatemala, 100,000 Mayan Indians Murdered;
# 1971 – 1979 Uganda, 300,000 Christians and Political Rivals of Idi Amin murdered;
# 1975 – 1979 Cambodia, 2 million educated persons murdered.
# 1994 Rwanda 800,000 Tutsi’s murdered.
# 1992 – 1995 Bonsia 200,000 Muslims murdered
If an armed citizenry would have prevented even the smallest of these genocides it would have saved more lives than a hundred years of US firearms murders.