Jody Foo

Writings by me.
Jan 22, 2013

Today the image below made its rounds on Facebook, twitter and a few blogs. It gives an example of a funny/provocative correlation.

IE = Murder

It is funny because it gives an example of how funny correlations can be found between two unrelated variables. To bad the actual chart is wrong though since it compares the absolute number of murders with the percentage of Internet Explorer users. I was however, intrigued and decided to see if any of the numbers were true, so I looked for them.

Data

I found homicide rates in the US stated in number of homicides per 100k people on wikipedia. Browsers stats I took from W3Schools's yearly logs since they were the only statistics I found (without putting to much time into the search) that stretched back to 2003.

The W3School numbers were a bit lower than those used in the original image. The Wikipedia article also only showed relative homicide rates (which is good), so I had to visit the United States Census Bureau (here and here) to get the actual population numbers so that I could calculate the actual number of homicides. One should also note that the crime statistics found on Wikipedia are biannual. Therefore, I only used browser statistics from those years.

Number of homicides vs. Internet Explorer use.

Below are two charts, one for 2003-2011 which was the range of the data I had, and one for 2007-2011 which aproximates the date range used in the original image. As you can see, the numbers are actually quite accurate.

Absolute homicide numbers 2003-2011 Absolute homicide numbers 2007-2011

Homicide rate vs. Internet Explorer use

However, the comparison is flawed, so I made two more charts, one showing the index values of the number of homicides per 100k people vs IE use from 2003-2011 and the other showing using 2007-2011 as its date range. Below are the charts.

Homicides per 100k 2003-2011 Homicides per 100k 2007-2011

As can be seen, even though both variable show a declining trend, their rate of decline differs.