Using NIBRS data to estimate crime reported to law enforcement
Without accurate, detailed data on crime, law enforcement agencies at every level would have a difficult time identifying problems and allocating resources. Law enforcement has recognized this need for decades, but as law enforcement agencies begin reporting more detailed information, the statistical methods, standards, and analyses need to evolve as well in order to provide maximum benefit to law enforcement and the community they serve.
Since the 1930s, the Federal Bureau of Investigation (FBI) within the U.S. Department of Justice has gathered and published an annual crime report through the Uniform Crime Reporting Program (UCR), which relies on data voluntarily submitted by law enforcement agencies. The data are divided into the five categories: summary data, county-level data, incident-level data, hate-crime data, and nonrecurring data. The data collected across the five categories are also collected in the FBI's National Incident-Based Reporting System (NIBRS). These data are important because they provide crime-related information to the public, researchers, law enforcement, media, and criminal justice students.
While the UCR Program has updated and expanded its scope over the decades, the FBI and the Bureau of Justice Statistics (BJS) saw the need to modernize annual crime reporting effort. In doing so, the federal agencies would provide a more inclusive, accessible, and comprehensive data report.
Benefits of transitioning to the NIBRS for uniform crime reporting
At the beginning of the 2021 data year, the UCR Program transitioned to solely producing crime reported to law enforcement estimates based on submissions to NIBRS. The transition to solely using NIBRS offers several benefits compared to the prior crime source of crime data, which blended NIBRS with aggregated count data from the Summary Reporting System (SRS).
First, basing crime estimates on NIBRS allows all crimes in an incident to be counted. The prior methodology only counted the most serious offense in an incident, while the new approach counts each offense reported in an incident. Second, a more complete picture of crime is estimated. NIBRS includes 58 different offense types compared to the ten available through the SRS. The NIBRS-based estimates provide information on all offense types. Lastly, NIBRS provides more information on the characteristics of each crime incident. Solely relying on NIBRS for crime reported data produces a more accurate and detailed set of crime estimate.
Challenges with transitioning to NIBRS-based estimates for crime reporting
While these benefits to the crime data users are immense, the transition to NIBRS-based estimates was not without challenges. For example, the FBI and BJS needed a methodology to produce estimates for agencies, including those that are actively submitting data to NIBRS and those that are still transitioning to the system. Because the NIBRS is a large database, there was a need for a process that can handle millions of estimates. Furthermore, given the amount of information collected in NIBRS, careful consideration was needed to determine what estimates could be reliably produced and meet the federal statistical standards of the FBI and BJS.
RTI contributing to NIBRS data modernization
Our team overcame the challenges and responded to the need to develop estimates based solely on NIBRS. We worked with the FBI and BJS to develop a new methodology that presents a major step forward in modernizing the data and methods used by the federal government to produce the official estimates of crime reported to law enforcement. The new methodology – which uses cutting edge methods in statistics and data science – relies on data submitted by state and local law enforcement agencies to NIBRS. The methodology minimizes potential bias in the estimates through a combination of statistical weights and statistical imputation. Statistical weights are used to account for agencies that have not transitioned. Statistical imputation is used to account for agencies that only provide partial information during the year (i.e., less than 12 months’ worth of information).
Additionally, the methodology utilizes a more robust method to account for the uncertainty in the estimates and provide more accurate confidence intervals. Ultimately, the methodology allows the public to have access to higher-quality data and more detailed data about crime known to law enforcement.
We helped design the framework to expand the level of detail available through NIBRS. For example, the NIBRS-based system generates representative crime estimates at the national, state, and local levels and provides information about the incident, the offenses, the victims, and those arrested. Furthermore, the methodology leverages the additional information available in NIBRS to produce estimates by demographic characteristics such as age, race, and sex.
Future of data modernization methodology for the NIBRS
Because the methodology we developed is expandable, it allows for more estimates in smaller areas of geography over time. As more agencies transition to NIBRS, the methodology can be extended by two critical factors. First, representative estimates below the state-level (e.g. collections of counties) can be produced. Second, agencies can add estimates of additional indicators. For example, the 2021 estimates include a detailed breakdown of drug offenses and their characteristics. Additional breakdowns of offense type and characteristics of the offense (e.g., firearm use) can be provided.
Our work with data modernization and new methodology has resulted in a system that will produce more than 105 million estimates annually, with the estimates accessible through the FBI’s Crime Data Explorer. (Indicators for Crime Estimates Using NIBRS Data).
This is the first time crime has been reported through representative estimates of detailed characteristics. Characteristics of a crime, not just the number of offenses, can now be tracked over time. This new time series will provide a better understanding of not just how much crime is occurring from year to year, but how the characteristics of each offense are changing over time (see CDE :: NIBRS Estimation Data). To produce a set of estimates of this size, innovative data science methods are used to efficiently process all the different components of the estimation process using open source software. These methods optimized the processing time needed to produce the estimates reducing it from 2.5 months under traditional methods to 2.5 days.
Learn more about RTI’s capabilities serving law enforcement, communities, and victims as part of the RTI Center for Policing Research and Investigative Science.