TriggerFinger


Sometimes, it doesn't pay to fact-check...


So, I get an email from a reader of mine complaining that James Elias, Highland Ranch, Colo needs math lessons.

In that post I linked to an article (that link might not be permanent) that claimed Canada has a crime rate about twice that of the US. The article itself is a letter-driven column called "The AnswerMan", and the above-named individual had written a letter to that column pointing out the crime rates -- hence the subject of the email.

Obviously, that Canadian crime rates are twice the US crime rates is counter to the conventional wisdom (which is why I found it noteworthy in the first place). The email had helpful links to the StatCan site and the Bureau of Justice Statistics National Crime Victimization Survey page, both of which have an overall violent crime rate category (the US page is per 1000, and must be multiplied by 100 to match the Canadian per-100,000 rates).

Now, I don't mind being fact-checked by my audience. But it helps to get your facts right if you're going to do it. Although this challenge looks OK at first glance, there's a subtle flaw in the analysis that renders it meaningless. Can you spot it?


So, I get an email from a reader of mine complaining that James Elias, Highland Ranch, Colo needs math lessons.

In that post I linked to an article (that link might not be permanent) that claimed Canada has a crime rate about twice that of the US. The article itself is a letter-driven column called "The AnswerMan", and the above-named individual had written a letter to that column pointing out the crime rates -- hence the subject of the email.

Obviously, that Canadian crime rates are twice the US crime rates is counter to the conventional wisdom (which is why I found it noteworthy in the first place). The email had helpful links to the StatCan site and the Bureau of Justice Statistics National Crime Victimization Survey page, both of which have an overall violent crime rate category (the US page is per 1000, and must be multiplied by 100 to match the Canadian per-100,000 rates).

Now, I don't mind being fact-checked by my audience. But it helps to get your facts right if you're going to do it. Although this challenge looks OK at first glance, there's a subtle flaw in the analysis that renders it meaningless. Can you spot it?

Here's what I wrote back:

James' math is fine. You, however, are confusing apples with oranges.

Specifically, you're directing me to "crimes" data in Canada and "victimization" data in the US. The difference is that a "crime" is either a charge or a conviction, while "victimization" is simply someone reporting that they were victimized on a survey. Reporting to police and filing charges is often a substantial barrier that results in a large discrepency between crime and victimization rates -- especially for petty crimes (too inconvenient to report to police), rapes (embarassment and social stigma), and defensive gun uses (why report a crime you prevented unless someone got shot?).

The fact that the the US numbers on the page you provide don't match those noted in the article should have tipped you off.

So I went looking for the other common measure of crime in the US, which is the FBI's Uniform Crime Reports: http://www.fbi.gov/ucr/03cius.htm

Those rates are conveniently per-100,000 to match the Canadian rates, and indicate for "Violent Crime" a rate of 475 per 100,000 in 2003.

The article I linked to specified 958 (Canada) vs 523 (US) for overall violent crimes in 2003. That's incorrect; those numbers are both from 1999. I suspect that if you redo your comparison with the UCR, you'll find that the numbers match.

It should be noted that victimization surveys are important tools in gathering statistical information about crime. They have several advantages over crime rates as recorded by the police and court system. But it is precisely because of those differences that you can't compare the two directly and get a meaningful result.

If you're not sure that the Canadian numbers are crime numbers, rather than victimization numbers, the definitions should clear that up. The key words to look for are "police reported crime statistics" and "substantiated by police":

The Canadian Centre for Justice Statistics (CCJS), in co-operation with the policing community, collects police-reported crime statistics through the Uniform Crime Reporting Survey (UCR). The UCR Survey was designed to measure the incidence of crime in Canadian society and its characteristics.

UCR data reflect reported crime that has been substantiated by police. Information collected by the survey includes the number of criminal incidents, the clearance status of those incidents and persons-charged information. The UCR Survey produces a continuous historical record of crime and traffic statistics reported by every police agency in Canada since 1962.

That is clearly describing a survey of police data, rather than a victim survey. The Nation Crime Victimization Survey in the US is an actual citizen survey that asks respondents to self-identify as crime victims.


This entry was published Sat Sep 24 10:43:35 CDT 2005 by TriggerFinger and last updated 2005-09-24 10:43:35.0. [Tweet]

comments powered by Disqus

Subscribe to Atom Feed

I am not a lawyer, and nothing on this site should be taken as legal advice.

This site is run on custom blog software and is being actively developed. Please be forgiving of errors.


This website is an Amazon affiliate and will receive financial compensation for products purchased from Amazon through links on this site.