Lies, Damn Lies, and Statistics

As someone with a technical education, I sometimes take it for granted that people understand how to properly use statistics to either support or refute a position. I need to be reminded sometimes that there are people in the world (most people, it seems) who have not the slightest interest in using statistics legitimately, but only want to use “numbers” to try to give weight to their previously held ideological position.

In the gun control debate, there are a lot of “statistics” used by each side to try to support their position – but if you don't know anything about statistics, whose numbers do you believe?

The gun control crowd likes to use very different environments to try to show that banning guns will reduce violent crime. How many times have you heard how the UK has so much less murder than the US, and they have far fewer guns, therefore having fewer guns leads to less violent crime? Simple, isn't it?

What if I told you that there was a very high correlation between murder rates and ice cream sales in the US? Would you believe me? Well, it's true and the effect isn't random.

Wait – what? What does ice cream have to do with murder? Does ice cream cause people to go on a sugar rush and become homicidal maniacs? Who cares! Since there is a correlation between ice cream sales and murder rates, we should simply ban all ice cream sales. After all – if it saves just one life... Think of the children!!

Do you see the flaw in this analysis? The key point is that there are many variables involved in the system being analyzed, not just ice cream sales and murder. What is actually happening is that murder rates are highest during warm weather, and lowest during cold weather. Ice cream sales follow the same pattern. In this analysis, the weather is a “confounding variable” that is responsible for both the murder rate and ice cream sales. As far as I know, there is no causal relationship between ice cream sales and murder rates, and I've never heard anyone posit that there was.

Back to guns. Comparing the murder rate of the UK vs. the US is highly complex because there are a large number of variables at work that differ between the two societies. Also, comparing the murder rates of various states and cities within the US is problematic for much the same reason. Economic, educational, demographic, political, social, and other variables differ a great deal between different locations. Do you think that Chicago and Salt Lake City can be compared simply by comparing their gun laws and their murder rate?

The legitimate way to analyze the effect of gun laws on violent crime (including murder) is to see how the rate of violent crime changes in a given location when firearms laws change. This is a way to “hold everything else constant” and evaluate just the effect of the change in gun laws. Nothing is perfect, and time is an important variable as well. However, evaluating what happens to the violent crime rate in a given location after gun law changes is by a very, very wide margin the most legitimate way to evaluate the effects of gun laws on violent crime.

What do you think this type of analysis shows? Well, I'll give you a hint – there is one scholar who has done a great deal of research in this area, and his approach is precisely that – hold as many variables constant as possible, account for as many confounding factors as possible, and then see what happens to the violent crime rate when gun laws change. This scholar is John Lott.

I've said it before, and I'll say it again – no one can pretend to have an educated opinion on this subject until they have read and understood “More Guns Less Crime” and “The Bias Against Guns” by John Lott. If you are reading this, and you don't own these two books (and you want to present yourself as being informed on this topic), your next stop is amazon.com.

As to the murder rate difference between the US and the UK, this guy I've linked to before explains much of what I've described above.