The National Incident-Based Reporting System (NIBRS) is the modern system for U.S.-wide collection of crime data by the Federal Bureau of Investigation (FBI). NIBRS contains structured information about dozens of aspects of each victim, offender, and incident reported to U.S. police. The NIBRS data set, amassed across more than 20 years, contains a wealth of detail from more than 6,000 police departments across 36 states. From this foundation, the publicly available data set has accumulated nearly 100 million records and will continue to grow vastly as the United States transitions from the Summary Reporting System (SRS) to the NIBRS-only platform. The major strengths of NIBRS are breadth of measurement and the flexibility to build an analysis from the data. A secondary strength is the inclusion of fields, such as county and date of an incident, that facilitates to the ability to link NIBRS to other U.S. data sets such as those of the Census Bureau, the Bureau of Labor Statistics, the Centers for Disease Control, or other Uniform Crime Reporting (UCR) collections.

Although most applications of NIBRS have centered on statistical reporting or academic research, there are operational applications of the data system as well. A data science team at the U.S. Marshals Service (USMS) has developed several projects to illustrate these applications. The projects include a dashboard to help law enforcement agencies identify other agencies that might have relevant information to share, a tool for the production of investigative analytics, and operationally relevant research.

The idea that the USMS has active data science teams may be a surprise to some, but it should not be. The USMS is actually the oldest U.S. statistical agency. From its inception in 1789 until 1870, the USMS conducted the United States Census.¹ Today, data scientists at the USMS continue this legacy as they seek ways to use data, such as the data in NIBRS, to make policing more effective, efficient, and safe.²

Although the USMS does not typically work the kinds of cases described below, the data science team often discusses case work with local detectives or chiefs when collaborating on fugitive or related issues. Building from these discussions, the USMS created these tools in hopes they would illustrate the value of NIBRS to the many local, state, tribal, and federal partners that investigate crimes and seek innovative ways to enforce laws. The USMS based these dashboarding concepts on generalized assumptions about NIBRS data that might not apply for all agencies and applications. Rather, these examples may help agencies think of ways to develop their own strategic applications for their own purposes.

Investigative Analytics

A challenging problem emerges in policing when an agency receives a case with some pieces of information known (e.g., offense behavior or victim details), but key information needed to develop a suspect or prioritize leads is unknown (e.g., offender age or gender). The challenge is all the greater when the case itself is peculiar, presenting details that preclude normal assumptions about whom the suspect might be. Consider, for example, the following possible scenarios:

• A child’s body is found in a school, the incident is clearly a homicide, and there are signs of overkill present (i.e., broken bones and teeth). There are no witnesses or suspect descriptions.
• A rural church receives gunfire from what witnesses describe as automatic weapons. There were two lines of fire on the school. There are no subject descriptions.
• A young African American girl has gone missing from a local park. Some of her clothing is found nearby, and there is evidence of sexual assault, but the DNA does not match any known offender. There are no suspect descriptions, other than the assumption of his gender.
• A masked man is attacking elderly women in the evenings, from behind, as they enter their homes. He forcefully sexually assaults them, then flees the residences. His DNA does not match any known offender and none of the women can tell his age or race. There are no suspect descriptions, other than his gender.

Each of these scenarios is rare in any one community, thankfully, but they do happen. The rarity, as well as the peculiarity of the details in these cases, means it is difficult to make assumptions about the offender based on prior experience or commonly known patterns in that type of incident, because there likely is no body of knowledge to draw on. What are the odds the offender is known to the victim and how? Is it likely the offender is a man or a woman? What is his or her likely age and race? How many offenders are there?

An investigative analytics dashboard powered by NIBRS data could help by providing probabilistic answers to these kinds of questions. When known case features for a current investigation are entered with filters, the dashboard shows the distribution of “unknown” features as observed in prior and similar incidents reported to NIBRS. The key feature of this dashboard is its flexibility; any one of the NIBRS elements can become a filter. There are thousands of combinations that can be quickly selected to focus on the particular type of victim, offender, and offense features that may be relevant. Because NIBRS can potentially supply a large sample size (from nearly 100 million rows of data representing much of the United States), even rare types of situations will often have plenty of cases to generate useful analytics.

The idea for this NIBRS dashboard grew, in fact, from considering one of these cases listed above: the serial rapist targeting elderly women in a particular city. There was little or no established research about trends in offender characteristics for such a case. Given the scarcity of prior research on this type of case and the fact investigators had little experience with this unusual type of offense, the typical approach to finding answers to these questions would not be helpful.

The tool, shown below in Figure 1, is set to known parameters of this case, which are also NIBRS data elements: an offender age range of 20–50 (presumed based on some behavioral information in the case); a location of a residence; a crime of forcible rape; and a lone offender who is male. The victims were all female and over 60 years of age. The filters for known information, as selected, are shown in ovals within Figure 1. The probabilistic information about unknown features are shown in graphics (e.g., offender race and age). Again, the probabilistic information about this case’s unknown details is simply a description of these same attributes in all prior cases in NIBRS that had identical known information as this case—but also had these other details reported (e.g., offender race, offender age). Thus, Figure 1 shows there were 5,744 prior sexual assaults of elderly women in their homes by strangers in NIBRS that also had details on offender attributes. Among those 5,744 prior times this happened, the offenders were white or Hispanic in approximately 49 percent of cases, but black in only 17 percent of cases, and offenders were an average of 37 years old.

Figure 1 shows all incidents in NIBRS, whether closed or not, in which a male stranger burgled a home and sexually assaulted a female over the age of 60. The map shows geographic concentrations of cases, and the graphs show breakdowns and averages of case characteristics.

Connecting Communities

A second type of dashboard USMS created from NIBRS data is called the Community Connector. It combines NIBRS data with census and other data at the county level to identify areas with similar crime or census features. This can be important to communities facing an unusual or new crime problem. The premise in these instances is it might be useful to identify other police departments that have already dealt with a specific type of problem to get advice.

The above case of the serial rapist preying on elderly victims is, once again, a good example. What is the best media strategy? Are special victim services needed? What kind of investigative and prosecutorial strategies work best with this kind of case? Selecting crime and community features via dashboard filters allows agencies to see which communities have encountered similar situations or crimes before and to tailor their search for advice from similar areas (e.g., cities with similar population densities).

Figure 2 shows all incidents in NIBRS in which an elderly woman was sexually assaulted by a stranger in her home during a burglary by a man between the ages of 20 and 50. The graphic allows users to also see census data about each county to identify agencies in similar areas.

Operations Research

A third kind of value USMS has extracted from NIBRS is the production of operationally relevant scientific research. Police agencies often find value in research about rare events—crimes that are especially important but difficult to study in any one department’s data system. For example, statistical models to determine the probability of a binary outcome (e.g., did an event happen: Yes/No) typically require at least 200 instances of the event to support regression or other modeling. NIBRS can provide a large enough data set to model these kinds of outcomes. Likewise, the comprehensive and consistent measurement required by NIBRS means one usually has the ability to generate a meaningful comparison group from across the United States or over time—a critical need for discovering and understanding risk factors for outcomes that matter to police and the public.

A good example of this is firearm violence directed at police. It would be rare for a single police department to experience enough of these events to support the creation of a statistical model to analyze them. Thus, it would be nearly impossible for any agency to estimate a statistical model that illuminates risk factors for this kind of violence. But NIBRS contains more than 600 of these incidents, pooled across the 6,000 police departments and 20 years of measurement. Further, because NIBRS contains data about police encounters that did not result in an officer receiving fire, the data system is in a strong position to offer insight about offender, victim, crime, and situational factors that distinguish cases where an officer was or was not the victim of gun violence.³ The same is true of other types of violence against police.⁴ USMS research, grounded in NIBRS, was able to identify a large number of NIBRS elements to signal risk for violence directed at officers. To be sure, this kind of analysis is only as strong as the completeness and accuracy of the data the FBI receives from the law enforcement community. NIBRS does not contain measures of every aspect of a crime that might matter, and, for that reason, any model built on NIBRS can help explain only a portion of the story of risk to police. Nevertheless, the insights offered are real, and they do matter. Such a model can inform police training and risk mitigation and help academics produce richer theories of this type of violence. The same is likely true of other rare events in NIBRS, such as child abductions, sexual assaults at playgrounds, female sexual offending, or possession of assault rifles by offenders. NIBRS can potentially provide a large pool of data for studying such events. For this reason, NIBRS can be a powerful tool for operationally relevant research.

Summary

Law enforcement agencies require data to continuously improve responses to routine police work and foster innovation and adaptation to new challenges. Smart policing is often limited by small or incomplete samples and data sets that lack the detail, comprehensiveness, and uniform information needed for comparison across geography and jurisdiction. This makes it difficult or impossible to study rare but significant events such as violence directed at police or generate meaningful comparison groups necessary for the study of risk.

NIBRS helps resolve these issues, especially when NIBRS data are matched with software that allows fast and flexible analysis and easy visualization. The examples in this article illustrate how this can work. The USMS does not usually work these kinds of cases; the dashboards discussed here were built as a proof of concept to show the operational value of the data. By developing these kinds of data dashboarding applications, the USMS demonstrates how other agencies could develop their own applications for NIBRS data.

The dashboards also illustrate how using NIBRS data this way is relatively simple. These dashboards were built with commercially produced software that is inexpensive and easy to use. Law enforcement agencies could produce, test, and begin using these kinds of dashboards in a few weeks. The ability to visually analyze and filter NIBRS data magnifies the data’s value and utility.

For this reason, the FBI is preparing to transition from traditional publication of static tables of crime statistics to the publication of interactive data analytics. A main example of this is the FBI’s Crime Data Explorer, which gives users the ability to filter crime statistics and create their own customized analytics. Agencies can download data files from the Crime Data Explorer.

Linking to the Future

The FBI plans to retire SRS on January 1, 2021, after which, the FBI will collect crime statistics only through NIBRS. In the meanwhile, the FBI continues to prepare the U.S. law enforcement community for the transition. As more agencies across the United States participate in NIBRS, the NIBRS data set will become more powerful and more representative of all regions of the nation.

NIBRS-participating agencies can help make U.S. crime statistics more useful by contributing to a system that is far superior to SRS. Agencies should also be aware that if they receive funding assistance based on SRS participation, they could lose their funding if they do not transition to NIBRS. The FBI offers guidance for NIBRS transition, and agencies can learn more by calling the FBI at 304-625-9999 or emailing ucr-nibrs@fbi.gov. d

Notes:

¹ Frederick S. Calhoun, The Lawmen: United States Marshals and Their Deputies (Washington, DC: Smithsonian Press, 1989), 3.

² David M. Bierie and Paul J. Detar, “Integrating Research and Researchers in the U.S. Marshals Service” Translational Criminology (Spring 2017): 4–6.

³ David M. Bierie et al., “Firearm Violence Directed at Police,” Crime & Delinquency 62, no. 4 (2016): 501–524.

⁴ David M. Bierie, “Assault of Police,” Crime & Delinquency 63, no. 8 (2017): 899–925.

Please cite as

Jeffrey Fisher and David Bierie, “Operational Applications of NIBRS to Policing,” Police Chief online, October 31, 2018.