So I re-requested the information, and in a few days I received all calls that resulted in a police incident report. That, police said, was the closest they could get to the data I was looking for. Both departments provided all data for free.

Combined, the data now totaled 3,700 records, a more sensible number considering Tulsa and Broken Arrow are two of the biggest cities in Oklahoma.

The data for both departments came in Excel spreadsheets. They were remarkably simple and required no joins or advanced computer-assisted reporting techniques.

Fields included location, date, time, call priority, call description and officer’s remarks, which were usually left blank. As in previous experiences with police department data, this data required a lot of basic cleaning.

The location field, for example, often included the school’s name and address, but there was rarely uniformity among records.

Sometimes the school name was abbreviated and followed directly by an abbreviated address. Sometimes the school name and address were written out. Sometimes the two were separated by a comma, dash or semicolon, and sometimes the field failed to include a school name or address at all.

I imported the data into Microsoft Access database manager and created two new fields, one titled "School" and a second titled "Address." Through a combination of update queries, I populated the new "School" field and used that data to fill in the "Address" field.

It was easy but time-consuming. I was still interested in analyzing the 18,500 calls and not just the ones that resulted in a police report.

Similarly, I had to replace about 140 description codes with a reason for the call. For instance, I replaced "INDIP" with "Indecent exposure" and "CAB" with "Child abuse." The police departments provided that information with my original request.

Once I had two cleaned databases—one of all calls from schools to police and a second of only the calls that resulted in a police report—I began running basic queries.

Finding the story
I was still without a specific story idea, so I began counting crimes by year or description, comparing crimes among Tulsa and Broken Arrow schools and among elementary, middle and high schools in both cities.

I created a new field to differentiate crimes against people, such as assault, versus crimes involving property, such as burglary. And to help put the numbers into perspective, I compared the total number of crimes per school with the school’s population. I received that data from the Oklahoma Department of Education.

The police data could have produced any number of solid stories:

• Why did Broken Arrow, a city with about a quarter of Tulsa’s school-age population, log as many calls to police as Tulsa schools?
• Why did calls from elementary schools produce almost as many police reports as calls from intermediate schools and high schools?
• Why were there twice as many assaults at Will Rogers High School than anywhere else?
  And why did schools report some 530 cases of child abuse since 2005?

My editors recommended a two-part series. The first story would investigate the different crimes at elementary, middle and high schools.

Gavin Off is the data editor at the Tulsa World in Oklahoma and a former analyst for the IRE and NICAR Database Library.