Graph showing time spent on Calls for Service, by Category. Approximately 25% on noncriminal calls, over 25% on Non-UCR Crime, 20% on traffic and 20% on Proactive
Percentage of time spent on calls for service category

Methodology

This assessment was compiled using Calls for Service data from the Austin Police Department covering January 2019 through June 2020.

This dataset encompassed 1,017,576 Calls for Service which were broken up into 7 overarching categories and 45 separate subcategories for further analysis.

Calls for Service do not perfectly capture time spent by law enforcement officers but instead provide a snapshot as to how officers interact with members of the public. Incidents relating to APD special assignments and secondary employment were not included in the overall assessment to provide a better understanding of interactions while formally “on the clock.” Overall, this analysis separated 956,256 Calls for Service into seven categories.

The seven categories of Calls for Service used in this analysis are:

Medical – Typically mental health, suicide, or death incidents. Additional data received indicates there is a separate flag for mental health incidents that stands apart from the call type. However, this data could not be joined back to the original dataset due to the absence of a linking data type. In addition, this data was only available for January through June 2020.

Responsive – Incidents that are not initiated by the officer. These may or may not be criminal in nature and include incidents like assisting other agencies, disturbances, and burglar alarms.

Non-UCR Crime – Incidents that are criminal in nature but do not fit in FBI’s Uniform Crime Report Part I categories (criminal homicide, rape, robbery, aggravated assault, theft, auto theft, and burglary). These range from city ordinance violations to kidnapping and everything in between.

Proactive – Incidents that are initiated by the officer or take place during discretionary time. These incidents include conducting follow-up investigations, investigating suspicious persons, and routine patrol activities.

Property Crime – Defined by the FBI as auto theft, burglary, and theft. This does not include theft by fraud, forgery, or embezzlement.

Traffic – These incidents typically involve responses to traffic accidents, enforcing traffic laws (other than DUI), and directing traffic.

Violent Crime – Defined by the FBI as criminal homicide, rape, robbery, and aggravated assault. Only assaults specifically identified as aggravated were included as violent crimes in this assessment.

Percentage of calls related to crime

*data visualizations created by Emma Barker

Analysis

Call distribution by category

A plurality (40.2%) of all Calls for Service during the analyzed period were responsive calls. Most of the incidents in this category fit into four subcategories: disturbances (8.5% of all incidents), assisting other agencies (6.9%), complaints (6.3%), and burglar alarms (6%).

Less than 1% of all Calls for Service from January 2019 through June 2020 were for an incident deemed a Uniform Crime Report Part I violent crime. Robbery and assault/battery incidents made up over 85% of all violent crime Calls for Service over that span.

Time spent on calls by category

There is a wide range in terms of how long officers spend on each call depending on the call type. Homicide incidents, for example, on average take over 6 hours between time officers arrive on scene and the time the scene is closed. City ordinance incidents, by contrast, average just under 8 minutes to resolve on average. Violent crime Calls for Service typically take over 3 hours to clear, significantly longer than any other crime type. Despite this, violent crime incidents took up only 2.8% of the analyzed time spent by officers during the analysis period. This finding falls roughly in line with a previous analysis of Calls for Service in New Orleans (LA), Sacramento (CA), and Montgomery County (MD).

3 subcategory types take up over 20% of officer Calls for Service time: disturbances (7.7%), traffic accidents (7.1%) and miscellaneous complaints (6.1%).

Call types by disposition

The Calls for Service data contains 15 disposition types describing what officers found or did at the scene of an incident. These range from writing a report (suggesting a crime occurred and was recorded) to an incident being cancelled before officers arrive. A report was written in just under 20% of all Calls for Service incidents though reports are written much more frequently in UCR Part I violent and property crime incidents.

Burglar alarms are a Calls for Service incident type that many municipalities are attempting to reduce to save officer time for more important tasks. Burglar alarms alone make up 6% of all Calls for Service. These incidents typically do not take long to clear – fewer than 15 minutes on average – but they still take up 2.2% of all time spent responding to Calls for Service.

 

A report is written in just 0.5% of all burglar alarm incidents. Over 95% of all burglar alarm incidents could be considered “false alarms” where the incident disposition is false alarm, cancelled incident, or no report.

Recommendations

AH Datalytics recommends the following steps for improved efficiency and transparency in Calls for Service.

 

1) Categorize Calls for Service with fewer signal/problem codes.

Regular and actionable analysis of Calls for Service data is hampered by the shear breadth of call signals. Of the initial 1,017,576 calls for service in the dataset, there were unique 837 signal/problem codes. The top 10% of all signals represent 92% of all Calls for Service and there were 237 signals with over 100 incidents making up 99% of all incidents.

More standardized signal/problem code designations would make it easier to understand trends and build replicable analyses.

 

2) Address Burglar Alarms.

Officers spend a combined average of 25 hours and 20 minutes each day responding to burglar alarms. Reducing these types of incidents – which largely reflect false alarms – would result in tremendous time and resource saving for law enforcement officers in Austin.

 

3) Alternate responses to non-injury traffic incidents and traffic enforcement can yield significant time savings for Austin PD.

Traffic related calls consume over 20% of officer Calls for Service time. The biggest drivers of this not only include accidents (7.1% of all officer Calls for Service time), but also traffic direction (4.5%), traffic enforcement (3.6%), and traffic stops (3.5%). Traffic related incidents consume significant amount of resources for the Austin PD and could potentially be handled in several alternative scenarios that can significantly reduce time spent on traffic response and enforcement by the Austin Police Department.

 

4) Identify potential non-police responses to mental health crises.

A separate dataset was provided for this analysis. This dataset shows 23,800 incidents in 2020 with either a mental health component though many of these incidents have no clear relationship to mental health issues in the larger Calls for Service dataset. The separate dataset indicates that 7.8% of all 2020 incidents between January 1 and June 30 involve a mental health issue. These calls originate in a variety of manner with 64% coming from welfare checks, disturbances, and trespassing incidents. Identifying which mental health incidents could be handled in a non-police manner would reduce police time spent while helping to solve underlying public health issues as well.

 

5) Locate addresses with repeat incidents.

Mental health and domestic violence incidents often occur at repeat addresses – though locational data was not provided for this analysis. Identifying the addresses where incidents are occurring again and again can inform non-police responses to help solve the underlying issues within these locations.

 

6) Make signals clearer

Signals should provide key information clearly to responding officers, supervisors, and analysts to understand the nature of incidents and identify trends. For example, there are 29 different signal types for assault/battery incidents, and shooting incidents are not differentiated from stabbing incidents. Making the nature of calls clearer would be an important step in making the Calls for Service data easier to digest.

 

7) Fewer dispositions.

The Calls for Service data shows 16 categories of disposition with 52 unique dispositions representing incidents with multiple dispositions. More dispositions make the task of analyzing Calls for Service data more difficult, so dispositions should be simplified wherever possible.

 

8) Publish data online and update daily.

Calls for Service data should be made available online and updated daily to facilitate transparency. This data should be published in machine readable format, updated regularly, and contain as many years of data as possible (with each year published separately if necessary). Public access to Calls for Service data enables the community to engage more thoroughly with the police department and local government to develop a shared understanding of crime and policing in Austin.

Appendix A – Detailed data table

About AH Datalytics

AH Datalytics is a consulting firm focused on bringing 21st Century analytics to the criminal justice system. Our work helps organizations better understand their problems and figure out if their solutions are working. We bring a wealth of expertise in analyzing, disseminating, and presenting information to organizational leadership and helping organizations effectively convey their analytics to the public.

AH Datalytics has extensive experience with providing comprehensive and data driven analyses to support criminal justice agencies, provide complex data reporting to agency leadership and an interested public, and build sustainable analytic capacity within organizations. Prior to forming AH Datalytics, co-founders Ben Horwitz and Jeff Asher greatly enhanced the use of data analytics at the New Orleans Police Department leading to national recognition and becoming a model agency in the use of data for police reform.

AH Datalytics is currently working with the City of Ferguson, Missouri to support the analytic components of their Consent Decree implementation. Working as subject matter consultants with the DOJ, IACP, and National Police Foundation, AH Datalytics has consulted with numerous agencies including in Spokane (WA), Baltimore (MD), Arlington (TX), Newark (NJ), and Tucson (AZ). Their work has been featured in the NYTimes most recently in the piece entitled, “How do the Police Actually Spend Their Time” and previously, “The Missing Numbers in Preventing Murders.”

Prior to AH Datalytics, Mr. Horwitz created the first audit unit at NOPD and then went on to create a nationally recognized Analytics Unit resulting in the development of NOPD’s MAX program. The MAX program features a robust and comprehensive dynamic reporting platform to facilitate Consent Decree reporting, close and effective supervision, and crime analysis. Prior to working at NOPD, Ben was the Data and Operations Manager at the Data Center in New Orleans in which he created methodologies, authored analyses, and disseminated economic, demographic, and other data sets to the public. Ben holds an MS in Public Policy and Management from Carnegie Mellon University with a focus on data analysis and information systems.

Mr. Asher initiated the role of public safety consultant for the New Orleans City Council providing data driven analyses to help the Council and public understand criminal justice issues. The resulting dashboards greatly enhanced the public’s access to data on crime, consent decree issues, and more. Before launching AH Datalytics, Jeff served as a crime analyst for the City of New Orleans and Jefferson Parish Sheriff’s Office, and prior to that he worked as an analyst for the Central Intelligence Agency and Department of Defense. Jeff’s analyses have appeared nationally on data journalism website FiveThirtyEight, The New York Times, Slate, and more. Jeff holds a MA from George Washington University.