Crime categories

Here’s a question: what’s the best, or maybe the most useful, taxonomy for crime tracking?

A Daily Californian reporter is working on this:
The data set comes from the Berkeley Open Data Portal, which looks inspired by NYC’s.

Further, the data actually comes from here where we should note that

The CrimeView Community data reflect crimes as they have been reported
to members of BPD based on preliminary information supplied to us by
the reporting parties. Preliminary crime classifications may change
based on the direction of follow-up investigations.

All of that is by way of introduction. The Daily Cal author’s question is: how to categorize crime? I and somebody else discussed using UCR and NIBRS categories, which are each different. Briefly:

UCR divides crime between violent crime and property crime. They have a crime hierarchy that labels each incident with the highest associated crime. So if someone steals your bag and beats you up (larceny-theft with assault), it counts as assault, a violent crime, and not larceny-theft, a property crime. There’s a separately tracked category for arson for some reason.

NIBRS uses what is essentially a tags system, so one incident can be tagged multiple ways with different crimes.

Ultimately this project is restricted by the City of Berkeley’s crime reports categories, which aren’t particularly clear. Do they use a hierarchy? Does each incident only get categorized with the “highest” level of offense, a la UCR?

So the question from the author is: what is a useful classificatory system for this data? Violent crimes, property crimes and minor crimes, as it is now?

I am suggesting the possibility of adding options to view by each type of report (i.e., all arsons, all sexual assaults, etc).

Any other thoughts?

1 Like

Can’t seem to edit my original post, but here’s what the reporter asked:

I’m trying to think of an intelligent way to classify crime data –
for example, is arson a violent crime? Is a misdemeanor assault really
as severe as misdemeanor vandalism? Lots of interesting questions, no
consistent or well-established answers. Here’s what I’m working with so far:….
I’m categorizing incidents as ‘Violent crimes,’ ‘Property crimes,’ and
‘Minor crimes.’ But I’m definitely not sure everything is where it
belongs. If you had to bucket these incident types into three (or maybe
four) categories, would you do it similarly?

Reporter in question here – as other examples of how news organizations have done this in the past, here’s the Chicago Tribune’s methodology and the LA Times’.

I’d probably do both a high level taxonomy, with a tagging system that allowed for arbitrary terms. That way you could tweak it a bit more, and see emergent patterns you weren’t looking for.

However, tagging may be more work if the data isn’t structured enough to run bulk operations on it (I didn’t look).

Violent, Property and Minor. Hmmm. Are some Property crimes also Minor? Can a report belong to multiple top level terms, since they are reported, and can break multiple laws?