There is a narrative in Canada that more guns equals more deaths.
It turns out that may not actually be the case.
I received many questions on my previous columns about the gun control debate and mass shootings in schools.
Several asked about statistics on the issue.
When I looked at some of the research that has been done on this topic, I found many of the studies I examined had significant flaws in which conclusions were made from the data that were beyond the scope of the studies.
These flaws can be grouped into a few broad categories:
1. Comparing shootings to gun ownership and then drawing conclusions about all homicides, without looking at the causes of death in all homicides. The major argument behind the less guns equals fewer deaths theory is that fewer deaths will occur. But if the study does not look at all homicides, or doesn’t explain why it is not looking at all homicides, it should not be making conclusions about all homicides.
2. Comparing gun ownership to all homicides, using a statistical proxy to represent gun ownership. A statistical proxy is basically a “counts-as.” The proxy used in these studies was gun suicides as well as gun homicides. These studies compared the incident of gun suicides or homicides against all homicides and then made conclusions about gun ownership, which was that when gun homicides go up, homicides go up, thus guns are bad. These studies assumed the relationship between gun ownership and the gun homicide rate was so consistent it could be used essentially interchangeably in doing the math, but this was not substantiated by the data.
3. Statistical conclusions made without using statistics. Some studies interviewed victims of gun crime and used their experience to make statistical arguments. Instead of focusing on the experience of the victim, which is a relevant field of study and appropriate for the scope of the study, one paper, for example, made the claim more guns equals more homicides because 10 people saw family members killed by people using guns. This is making a quantitative (measurable) conclusion by using qualitative (descriptive and immeasurable) data.
4. Misrepresenting statistics. Some studies either did not disclose their P-value (a statistical measurement used to validate a hypothesis against observed data) and R-squared value (how well the data fit the regression model) and made conclusions that conflicted with these values. This basically means there was no way to verify what the studies said, or the authors’ conclusions actually conflicted with their results, meaning they didn’t know how to read their own conclusions properly.
5. Making a causal argument using a correlational calculation. Just because there is a correlation between two sets of data does not automatically mean one caused the other. Statistics cannot prove causation. No credible academic study would make a causal conclusion from a correlational calculation.
I examined some data sets for all homicides globally separated by country, and data sets for firearms separated by country and total registered and unregistered firearms.
All homicides did not include those resulting from war or armed conflict, while the use of firearms only included use by civilians.
Military and police firearms were removed from the data set.
Some countries did not have verifiable data for both, with 123 countries remaining.
I analyzed different groups from these 123 countries, removing the United States as it is the only country with more firearms than people, which might skew the data.
I looked at all homicides vs. total firearms, all homicides vs. registered firearms and all homicides vs. unregistered firearms.
Both per-capita and non-per capita data were run in every combination with 95% confidence, meaning only P-values of less than 0.05 were accepted.
None of these examples produced a relationship that was statistically significant.
A responsible researcher would conclude there is no correlational relationship between homicides and either total firearms, registered firearms, or unregistered firearms within these data sets.
An irresponsible one would conclude that guns don’t kill people, people kill people, but these data sets did not say that.
— Alex Vezina is the CEO of Prepared Canada Corp. and teaches Disaster and Emergency Management at York University. He can be reached at firstname.lastname@example.org.