The shocking news from Las Vegas reminded us, should we need it, of the seemingly infinite capacity of humans to be cruel. Each new atrocity prompts debate on how to fix the problem, with gun control usually somewhere near the top of the agenda on the other side of the Atlantic.
The BBC website this morning promised a greater understanding of “America’s gun culture in seven charts” (http://www.bbc.co.uk/news/world-us-canada-41488081) and revealed that 11,000 people in the US were killed by a gun last year.
America is a big country and you would therefore expect it to have more deaths from guns, cars, falling off ladders and lightning strikes. What we need is data per capita, relative to the size of the population. Unsurprisingly, a quick trawl of the internet finds such data care of the wonderful Wikipedia:
The table is not perfect as some of the data is many years old, and it only contains 70 countries (notable exceptions being Russia and China), but it’s good enough to be going on with. What do we see?
America has 112 guns per 100 people. With the exception of Serbia, no other country comes close. Think about that for a moment: there are more guns than people in America, and there are an awful lot of people, so that means an awful lot of guns. But study further and the table makes fascinating reading.
America doesn’t actually feature in the top ten of countries with most gun-related deaths per capita. Honduras holds the unwanted accolade of most likely place to die from gunshots, followed by a variety of its neighbours.
Should we exclude these Latin and South American countries from the analysis as they are clearly little more than lawless savages? Hmmm, you have to be really careful when doing this, as what you usually end up doing is throwing away the data that doesn’t fit your preconceived notion.
We all know that the more guns people have the more they are likely to kill each other, yeah? So if we put these points onto a scatter diagram (apologies to those of you having anxious flashbacks to GCSE Maths classrooms) we would expect to see those countries with low gun ownership rates have low gun-related death rates, and vice versa.
Before we do this let me say this: If I learnt one thing from my time studying and teaching Statistics it is that correlation does not prove causation. In other words, the fact that two variables are linked does not imply that one has caused the other.
It is vital that you understand this if you are to make sense of what follows but, alarmingly, it is widely misunderstood even in the scientific community – for how many decades have we had the ideology rammed down our throats that because CO2 is rising and global temperatures are rising the former has caused the latter?
This makes no more sense than saying that Northern Lapwings are responsible for ice cream sales in this country because figures for both increase in the summer months. However, that is a different rant of mine and one that should be left to a later blog so as not to distract from the matter at hand.
Anyhow, back to gun ownership and death rates in 70 countries, and we are expecting a scatter of points from bottom left to top right of the graph indicating a strong positive correlation, yes? (Are you still with me?)
Oh. Certainly not the classical shape of positive correlation we were expecting. However, we have the Product Moment Correlation Coefficent, or PMCC for short (trust me, I’m a statistician) to help us out by putting a number to the extent of the correlation: 1 is perfect positive correlation, -1 is perfect negative correlation, with a whole spectrum in between. 0 represents no correlation or link between the two variables.
So, essentially, there is no link between guns and gun-related deaths. If anything, the PMCC is slightly negative, indicating that the higher gun ownership the lower gun related-deaths!
It should be noted that the data includes suicides, which would presumably have happened one way or another even without guns. Suicide seems relatively high in parts of Europe, and Switzerland (which was mentioned in the BBC article) is a case in point. It has 24 guns per hundred people but around 90% of its gun-related deaths are suicide. This pattern is repeated in Scandinavia, Germany and even our nearest neighbour France.
So what would happen if we excluded suicides, unintentional and undetermined gun deaths? Surely then we would see that the countries with higher ownership rates would have higher death rates?
Astonishingly, this makes the correlation even more negative. Countries with higher gun ownership rates have slightly lower homicide rates. I was talking with my friends yesterday (in a virtual WhatsApp kind of way) about the issue and declared that if I lived in America I would want to keep a gun in the house.
To be clear, I would prefer a world without guns, nuclear missiles, hatred and avocados. But, given that all those exist, if I lived in America and heard a window breaking downstairs in the middle of the night I would rather have a gun to protect my family. The statistics seem to suggest that this actually results, if anything, in slightly lower death rates.
What if I now combined the two measures into one statistic to see what countries have an unusually high death/homicide rate per gun (essentially I am dividing the ‘Total’ and ‘Homicide’ columns of the table by the ‘Guns’ column):
I have listed the countries in alphabetical order and have not included the countries with very high numbers, like Honduras, which would have made the scale absurd. And of course I have cherry-picked the interesting countries from our near-neighbours. Before we begin analysing in earnest, please note the following:
The numbers on the vertical axis representing the number of deaths per gun in each country over a year really are very, very small. For the USA it requires over 30,000 guns to create one homicide. As they used to say on Crimewatch, please don’t have nightmares!
Astonishingly, countries such as Portugal and The Netherlands with relatively low ownership rates have relatively high death rates per gun. Both rank higher than the US for death rate and homicide rate per gun, despite having much stricter gun controls.
Of course, none of this negates the fact that the more guns in a country the more deaths are likely. The Netherlands, for example, may have over double the homicide rate per gun than America, but it has far less gun-deaths per year because there are less people and less guns.
You might think from all of this that I am pushing some pro-gun bandwagon. Rest assured, I am a statistician trying to present an unbiased analysis of the data rather than a politician promoting a certain agenda. I don’t own a gun, don’t play Call Of Duty, and I only go paintballing on stag parties.
Gun control is an incredibly complex issue and there are many aspects that I have not touched upon, including the type of gun, licensing, medical restrictions and socio-economic factors.
The most sobering chart from the BBC website was, in my opinion, this one that shows that only a tiny proportion of US murders were committed with rifles and shotguns – the only guns that are allowed under license in the UK, with private ownership of automatic firearms and handguns banned.
Nor have we examined other types of weapon and their prevalence and use throughout the world. Is it the case that people in countries with strict gun control laws just find other ways to kill each other instead? It is a sad truth of our demented modern world that most recent atrocities have been carried out with trucks, knives and home-made bombs, which are even harder to control than guns.
But as someone who has seen an awful lot of bad statistics and graphs over the years, I had an inkling that the BBC website was not presenting the whole picture, and I was right. Remember to ask yourself the next time you see a pretty graph or a persuasive statistic – who has produced it, and to what aim?
Big data is immensely powerful and has the power to do good. I fundamentally believe in statistics being the art of decision making in an uncertain world. But there is also a responsibility of those that own and use data to do so with respect and care. Just like guns, I suppose.