While each jurisdiction maintains data differently, Calls For Service (CFS) data will ideally include:
The above information can be analyzed to answer a range of questions. To name just a few: when are police busiest, which types of calls police spend the most time on, which areas of the jurisdiction generate the most 911 calls for violent crime and which for quality of life issues, how many officers deploy to certain call types and how long they spend there, which calls for service result in arrest or criminal charges, which calls are flagged as being eligible for alternative responders, and how often is an alternative responder on scene.
Importantly, even where the data are well-maintained and complete, there is much that CFS data do not include. For this reason, additional data sets (if available) can be helpful.
Calls for service data may only get you so far — you may want additional information. Here are examples of some data sets you may want to consider:
Use of force: Layering use of force data onto CFS data can provide critical insight into what occurs when an officer responds to certain call types. However, the use of force should not, in and of itself, be dispositive in determining whether a law enforcement response is needed at such calls. CFS data alone will not indicate, for example, whether force was used because an officer’s presence ended up escalating a situation, or, conversely, if force was necessary to protect the safety of a victim or the officer.
Neighborhood information, including race and socio-economic data: Because of the historic and ongoing disproportionate impact of law enforcement on communities of color, most notably Black communities, layering information about race with CFS data can provide critical context for analysis and decision-making. For example, such data can reveal if a call taker’s or dispatcher’s assessment of risk is influenced by the demographics of a neighborhood, which may, in turn, impact the decision to assign either police officers or alternative responders. Data about race, ethnicity, and socio-economic status may need to be drawn from an external source, such as the Census Bureau's American Community Survey, and then mapped onto CFS data.
911 Call Recordings: A recording of a 911 call or a 911 call narrative can provide further helpful information and context. This data, because of its content and format, is complicated to obtain and analyze and will require a data sharing agreement.
Crime Data: Local crime data, such as Uniform Crime Report (UCR) data or National Incident-Based Reporting System (NIBRS) data, can provide helpful insight into the relationship between public safety and first response. For example, they can show whether call volume is proportionate to rates of actual criminal activity.
Enforcement Data: A broad category, enforcement may include information such as arrests, citations, summonses, and tickets. It can also include police department data on investigatory stops (also known as “stop and frisk”), vehicle stops, as well as warrant executions. Together, these types of indicators can paint a picture of where police activity (and activity of other agencies, to the extent that other agencies participate in enforcement) occurs and what impact it has. Additionally, data may illustrate the efficacy of enforcement activity, such as a department’s clearance rate (the rates at which police solve crimes), and information from prosecutors and courts about declination rates (cases in which prosecutors refuse or decline to prosecute), as well as conviction rates for felonies and misdemeanors.
988: On July 16, 2022, the National Suicide Prevention Lifeline number switched from an 800 number to the three-digit number 988. This number serves as a hotline for people in mental health crisis, routing callers to local call centers where they can talk to a counselor and receive a referral for services or, if needed, emergency assistance. The expectation is that 988 will become a resource for those in mental health crisis who previously would have called 911, though many such callers will still likely continue to rely on 911. 988 does not currently have geolocation capabilities — calls from mobile phones are routed to local call centers based on their area code and not their physical locations.1
A final note: Depending on certain system features, cross-referencing data sets can be challenging. For example, if use of force data and CFS data are not recorded in the same systems and do not share a common incident number or other common identifier, it may be necessary to match up data manually using information such as call times and types. However, this non-exact matching can introduce errors (for instance, the time on a call for service and the time for the associated use of force may not be the same because documenting and investigating the use of force requires additional effort), so this approach should be taken with appropriate caution.
Using data to understand current public safety practices and outcomes presents several limitations:
Not all data is good data. Police and government agency data can suffer from quality and integrity issues. The process of analyzing data should include evaluating how reliable and accurate the data is, and whether it meaningfully captures what it purports to. The scope and depth of analysis will likely need to be adjusted based on the quality of the underlying data.
One common example of such a problem is missing data or data present only in a fraction of cases. For example, a large share of 911 calls may be missing data fields that record response time. While it is tempting to ignore cases with missing values and analyze rows with complete data, such filtering can potentially skew results.
Data sets are not self-explanatory. Data sets may include technical terms, terms of art, or acronyms that require additional explanation, whether from professional training materials, standard operating procedures, or a data dictionary. Additionally, even when information is understandable, it may not always be meaningful — calls for service, for example, may be coded into overly broad catch-all categories such as “meet complainant,” “service assignments,” or “info call.” In some cases, 911 call transcripts and notes might enhance further understanding, but if such data are not available or cannot be reviewed systematically, then any analysis must account for the inherent limits of the available data.
Privacy concerns may limit access and detail. Whether through freedom of information act requests or a data use agreement, there may be limitations to what data can be shared with the public due to concerns about privacy and/or protection of sensitive law enforcement information. This may limit how granular or complete a data set can be. CFS data, for example, is typically provided at the block level rather than the address level. Similarly, the transcript of a 911 call or the narrative detail in a CAD system would be difficult to obtain without onerous redactions or negotiating a complex agreement to protect the caller’s privacy. This missing information may be highly relevant to an accurate understanding of events. For example, if a call for service results in an arrest, then a transcript or call notes could help determine whether the arrest was necessary and justified, or instead the result of an avoidable escalation.
Data systems often do not interface well with each other. Law enforcement information is often distributed among multiple databases, each of which tells part of a story, and none of which tells the whole. For example, no single system is likely to record all interactions that result from a 911 call for a behavioral health crisis, making it difficult to establish a clear and accurate determination of how many individuals in mental health crisis receive a police, other governmental, medical, or social service response. Law enforcement officers may report information about a subject’s mental health condition in any of a number of different locations: an arrest report, use of force report, incident report, stop documentation form, or crisis intervention or mental health data collection form, to name a few. At the same time, it may be difficult to determine if these records overlap — if, for example, subjects of use of force who were experiencing mental health challenges were or were not also captured in a database of arrest reports, incident reports, or crisis intervention documentation procedures. It can also be difficult to connect police responses to systems that track available social services, in cases when law enforcement has steered individuals experiencing a crisis toward such services.
Obtaining data presents additional challenges for those outside the government. Jurisdictions’ practices vary widely with respect to how they keep data and the type of data they make publicly available: some have open data portals that provide granular CFS data, force statistics and crime data; others may even provide analytical tools or regular, comprehensive reports, while others post only minimal basic crime data. (Unlike crime data, CFS data is not collected and released at the federal level, so the information can only come from jurisdictions directly.)