How we did our 911 analysis

Calculating response times from the 911 data acquired from Central Police Services is fraught with difficulty. 

For each call, the 911 log provides the moment: 

  1. A 911 operator took the call.
  2. The call was transferred to a Buffalo Police Department dispatcher. 
  3. An officer accepted the call from the dispatcher.
  4. An officer reported arriving at the scene.
  5. The responding officer cleared the call.

As far as the Buffalo Police Department is concerned, their responsibility begins with #2.

From a 911 caller’s perspective, what matters is the time elapsed between #1 and #4, so that’s the basis of our calculations. However, it’s more complicated than that. 

Read the full story based on the analysis

Often there are multiple 911 calls for a single incident, but not all those calls get time-stamped and logged out in a timely fashion. Often, just one responding officer does that, leaving misleading data in the logs.

Sometimes an officer arrives on the scene and jumps right out of the car without telling dispatch they’ve arrived. It might be the second car at a scene, which arrives a few minutes later, that apprises dispatch. 

That’s why Buffalo police don’t use 911 data to track response times, Capt. Jeffrey Rinaldo told Investigative Post in an interview before his retirement in July. That changes if there’s a complaint about when police arrived. After that, an Internal Affairs investigator might examine the records to assess what happened, he said.

“That’s where the department can then back into these numbers to try and determine, ‘Okay, if they call the police at noon, why did it take until three o’clock for the officers to arrive?’ “Rinaldo said. “Well, answer being, that district had a shooting or other high-priority call. All of the district cars were tied up in that call.”

Investigative Post began its analysis with nearly 1.8 million 911 calls made to Buffalo police from January 2014 to July 2020. We obtained the records through a Freedom of Information request to Erie County Central Police Services.

To weed out inconsistencies in the data, we discarded all response times greater than two hours, on the theory that most of those were the result of data-reporting anomalies.

Further, we used median response times rather than average response times, on the theory that statistical outliers remained that skewed the averages.

We also focused exclusively on five high-priority dispatch codes, rather than lower-priority calls where life and property were not in immediate danger.