Treatment of Missing Data in the FBI ’ s National Incident Based Reporting System : A Case Study in the Bakken Region

In response to the growing reliance upon administrative records to generate national estimates of key indicators of interest, federal statistical agencies have been expanding and enhancing activities to assess and improve the quality of data collected through administrative records systems. Given that administrative records are often collected across different agencies or local reporting units, encountering missing data at both the individual and aggregated levels is inevitable and must be addressed when developing estimates. The Federal Bureau of Investigation’s (FBI) National Incident-Based Reporting System (NIBRS) is a system designed to collect data from administrative records to be used for research and statistical purposes. It was developed as an expansion to its Uniform Crime Reporting (UCR) Program to improve the quality of crime data collected by law enforcement by capturing detailed information on each single crime occurrence. In this paper, we present an imputation method developed to handle missing data in NIBRS by leveraging other relevant external data sources. Given the hierarchical structure of NIBRS, this particular method addresses missing data occurring at multiple levels, including: a) incident level, due to item missing within each incident; b) agency level by month, due to some agencies reporting only for a partial year; and c) agency level by year, due to some agencies not submitting data to NIBRS for an entire year. The proposed method will be applied to a study in the Bakken region of the United States that utilizes the NIBRS data to examine how crime and law enforcement changed in the region as oil production increased from 2006 to 2012. A variance estimation method was also developed to evaluate the uncertainties in our estimates introduced by this imputation technique. In a broad sense, this research also can be viewed as an example of how to handle missing data in hierarchically structured administrative records.