In the open data world, crime statistics are an important data set. But, I have had some recent thoughts about crime statistics which I would like to share.
Where do crime stats come from?
When I worked in that field, we had 3 systems in which we could pull from – CAD, ARS and RMS. The systems worked as follows:
- CAD – are the calls for service. They were the most current. But also, the crime type and information is the most inaccurate.
- ARS – These are reports that have been approved by an officer and a supervisor(SGT). They are probably 1 to 2 weeks out from the data of occurrence but are more accurate than CAD.
- RMS – These are reports that have been approved by records employees and are the final, authoritative source of information. These are possible a month old but are the most accurate.
As you can see, as time increases (we get further away from the data of occurrence) the accuracy of the data associated with the incident increases. Do you know which source the data set you are using comes from?
This dataset contains the block location, case number description and date of calls for service received by APD that have been entered into the case management system and approved by a supervisor.
This leads me to believe that the data would be from ARS. Meaning it is a few weeks behind, but fairly accurate. Looking at the data, the most recent incident is January 8th. Today is the 15th, so it’s a week behind. About right for ARS. Having this metadata means I can be confident in what I am measuring. Good Job Albuquerque!
Which System Should we Query?
If you want to see what is happening now, you need to query CAD. But these data are only good for getting an idea about where incidents are occurring, not what types of incidents.
If you want to look at a month of crime, you should use ARS.
And for any longer historical crime analysis, RMS is the way to go.
But Wait, I have an Issue with ALL Crime Data
First, most crime data only lists the highest charge. One crime. If I shoot someone during a drug deal, this is a murder/homicide. No mention of distribution. If you are studying drug crimes, this will not show up. That is a problem.
We had a case in Albuquerque recently where an individual stole a vehicle that was running. They probably would charge him with unlawful taking of a motor vehicle (not auto theft). But there was a child inside the car. Now it’s a kidnapping. The report would pick the highest charge, kidnapping. But the real criminal intent was auto theft/unlawful taking. As much as I want to see that individual locked away for what he did, the crime statistic does not properly reflect the crime – it actually neglects the primary offense. And this can now be exacerbated by my next issue with crime data.
Lastly, here is a scenario: a person calls 911 because of a shooting, an officer is dispatched, an arrest is made, an attempted murder report is filed and it makes it to RMS – the official record. Everything is good. But the case goes to trial – or it doesn’t because of a plea deal – and the individual is charged/pleas and found guilty of aggravated assault. The court is the official record. A crime of attempted murder never occurred at the location, on the date that the report states, an aggravated assault did. What if the person was found not guilty? A murder never occurred. But the police say it did? Is that libel?
I know this may seem nitpicky, but given the unbelievable number of plea deals and reduced charges, how accurate are our police reports – probably more so than the final case disposition but that is the final truth. Crime stats are not crimes, they are charges if we don’t use the final disposition.
I think this is a new area for crime research. A department reports to the FBI UCR based on RMS, but those charges may not actually be what the courts decided. I would love to see the difference between charges in RMS and final disposition. Maps comparing crimes with final disposition should show much lower levels of crime and far fewer felonies.
Just something to think about.