Estimating Statistical Bias for County-Level Crime Rates

• Activity Number: 417
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Government Statistics
• Author(s): Emily Griffith*+ and Samuel Berhanu
• Companies: FBI and FBI
• Keywords: Bias ; Missing Data ; Parametric Bootstrap ; Small-Area Estimation

Abstract

The Uniform Crime Reporting (UCR) Program is a voluntarily submitted crime data collection from local, county, state, and tribal law enforcement agencies in the United States. The FBI administers the program and releases crime statistics from this dataset primarily for use in law enforcement administration, operation, and management. County-level crime rates are calculated using the resident population covered by the reporting agencies. Society's growing dependence on data and analysis has resulted in administrative datasets, including the UCR, being used in statistical analyses. Motivated by concern over missing values, we explore the statistical bias present in annual county-level rates created by missing data between 1984 and 2009. Small-area estimation and parametric bootstrapping approaches are being considered as imputation techniques. Key technical issues addressed are the spatial correlations between agencies within counties, the temporal correlations within an agency throughout the years of data, and the complex structure of missing data due to the size and scope of the data collection.