When "Omaha Crime" rolled off the presses in November, I thought to myself, "Is this all there is?"
The 12-page section about reported crime from 1990 to 1997 looked pretty simple. Good stories and photos, decent maps (despite our old presses), lots of information.
I realized that the simplicity itself was the great achievement of the project. We had produced a readable, balanced, fair and informative section about crime from nearly 13 million convoluted mainframe records.
To me, there was nothing extraordinary. We could not boast about doing complex statistical analyses or unearthing some government secret. There was no place where it could be seen plainly why this project took 17 months of my life. Then, I realized that the simplicity itself was the great achievement of the project. We had produced a readable, balanced, fair and informative section about crime from nearly 13 million convoluted mainframe records.
Many journalists can identify with the story of how "Omaha Crime" came into being. Most likely it isn't 13 million records you are after, but the number doesn't really matter. It's the process-the often confounding process of trying to obtain data from a reluctant government agency and, if you get it, the incredible amount of time it can take to clean that data and determine if it is reliable.
"Omaha Crime" is a poster child for computer-assisted reporting. During the life of the project, we used Excel, Access, Visual FoxPro, SPSS and ArcView. We also designed a poll to assess how safe residents felt in Omaha.
First steps
The journey began in January 1991 with written requests for electronic crime data from the police and sheriff. The newspaper, having limited CAR knowledge at that time, found itself unable to respond effectively to the government technobabble it got in response to those requests. So, the issue landed on the sidelines, arising only sporadically during the next five years.
In early 1996, with a more extensive understanding of CAR, we tried again. At our first negotiating session, we faced a roomful of lawyers and law enforcement officials who did not want the data in our hands.
Neither the police nor the sheriff had ever used the data they collected. They had an antiquated mainframe system that allowed for little more than counting incidents for the FBI's Uniform Crime Report. As one police crime analyst put it: "We come armed every month with a handful of pencils."
The last thing government officials wanted was a newspaper analyzing what they should have been analyzing themselves. So, at our final negotiating session in August 1996, the police department demanded a "handling fee" of $1.00 per record in addition to other data costs.
They figured we'd go away. Instead, we sued.
Two months later, they agreed to give us the data - reported incidents, bookings, arrest warrants, criminal histories and full personal information on perpetrators and victims.
Reliability check
The first installment of data arrived in June 1997. That's when crime numbers first consumed my life. More arrived throughout the year.
Here are some of the things we did to help make sure the data was as reliable as possible:
We excluded from our study places that had their own security or police forces, such as universities.
Incidents with "unfounded" status codes were removed.
Violent incidents reported as having happened at police or sheriff offices, hospitals and extended-hour medical care centers, and jails and juvenile detention centers were removed. Non-violent incidents at police or sheriff facilities and jails and juvenile centers also were removed.
Incidents reported at shopping malls were excluded from per capita calculations so they would not skew rates in surrounding neighborhoods.
We mapped where the incident happened, not where it was reported. If you are going to group your locations into larger areas like census tracts or block groups, be careful when mapping intersections. Determine on which corner the incident happened. If that is not feasible, randomly assign intersection incidents to the converging geographic areas.
If looking at incidents over a number of years, get population figures or estimates for as many years during that period as possible, then average them for your per capita rate. Using just one year may not be accurate if population is growing or declining substantially.
Because our data arrived in installments, the incident reports did not all end at the same time. We had homicides through July 9, 1997, drug offenses and robberies through July 22, arson cases through August 25, etc. So we weighted our crime rates for each offense.
Carol Napolitano can be reached by e-mail at carolnap@owh.com