Phoenix Kidnapping Data Analysis

(photo of Phoenix's Camelback Mountain by laszlo-photo)

I hiked a few times up into the mountains when I was in Phoenix a few years ago. Really amazing views during sunrise and sunset.

Kidnappings News in Phoenix

There have been a number of news stories in recent months detailing the rising kidnapping problem in Phoenix. As is with stories like this, I am curious about the data behind the news stories and whether the news story headlines really are indicative of reality or hype. So I will pick on a recent CNN story which is how I became aware of the Kidnapping problem in Phoenix. The CNN story headline says it all: Torture a hallmark of Phoenix's drug kidnappings (just a quick warning, the detailed storyline in this news article is pretty graphic and disturbing)

As is with most news stories, there is more anecdotal info and emotional story lines but there is some data in the story so let me pick through a few of the data related points from the CNN story.

"Unprecedented kidnapping numbers - in 2007 and 2008"

The story mentions that there were an unprecedented number of kidnappings in 2007 and 2008 which initiated the formation of the HIKE squad (Home Invasion and Kidnapping Enforcement). The Phoenix Police Department was nice enough to share with me detailed historical data on kidnappings which is not readily available on their website. Here is a graph showing the number of kidnappings per month since 2004:


An exponential weighted moving average (EWMA) is also shown as it is my favorite weighted average method to show the general trend in a noisy data set. You can see here there is alot of month to month noise in the number of kidnappings. But overall, the number of kidnappings does seem to have increased in the past few years.

And if you do a 2-sample Poisson test (count data like this is generally Poisson distributed) then you do confirm that there is a statistically significant difference in the data from 2004 through 2006 compared to the 2007 through now data.

"On track to shatter those numbers this year"

The second main point regarding the kidnapping numbers is that Phoenix is on track to make a new record in 2009 and "shatter" the previous kidnapping totals in 2007 and 2008. The story says "as of March 31, the city had 101 reported kidnappings. If the trend continues, Phoenix will record an increase in kidnapping for a fourth straight year". The news story was published on May 19th and the data mentioned in the story is through March 31st. Interesting, in April the kidnapping data dropped considerably from the first three months of 2009. When I asked the Phoenix Police Department for the data, they sent me the data through the previous month (May). Here is plot showing just the last three years of data:

From this perspective, 2009 (the last 5 data points) don't really look any worse than the previous couple of years. Although at this rate, they would make a new record.

Statistically speaking, doing a 2-sample Poisson test, 2009 data compared to 2007 and 2008 is not at this point showing a statistically significant difference. What this means practically is so far, there is not evidence that 2009 is any worse than 2007 and 2008.

Looking Forward

I was curious about if I was optimistic, how big of an improvement in next months data would be needed to be "statistically significant". Playing around with some trial and error, if next months data showed less than 25 kidnappings, it would suggest statistically that something had changed for the better. There are alot of different possibilities for the coming months and I may return to this if it looks like Phoenix is getting a handle on the kidnapping problem.

One important point about the kidnapping problem. If you read the various news stories including this one from CNN, it is pretty apparent that these kidnapping crimes are primarily involving drug and human traffickers abducting each other for ransom. So primarily these are crimes involving victims who are already involved in illegal activities. Not to diminish any crime but it is important to point out that for the most part these are not involving "innocent" victims.

...