Supplemental Work in Constructing and Disseminating Small Area Estimates from the National Crime Victimization Survey (NCVS)

The National Crime Victimization Survey (NCVS) is the Nation’s primary source of information on criminal victimization. For several years, BJS has pursued the goal of increasing the available geographic detail from the NCVS to better serve the needs of stakeholders. As part of Project 2008-BJ-CS-K067, BJS sponsored research by Westat on the possible application of small area estimation techniques to estimate critical crime rates for states and for selected large counties and metropolitan areas. In 2014, a Westat team led by Dr. Robert Fay produced 3-year state averages of the incidence of property crime by three types of crime (theft, household burglary, and motor vehicle theft) and of the incidence of violent crime also by three types (rape, sexual assault, and aggravated assault combined; robbery; and simple assault) and by relationship to the perpetrator (intimate partner, stranger, or other) for the period 1999-2013, estimates that were subsequently published by BJS (Fay and Diallo, 2015). The team produced similar estimates for large counties and metropolitan areas for 1998-2012, which BJS also published.

The small area estimation results produced in 2014 generally appeared useful and promising, but three limitations are evident. First, the small area model requires as inputs both the direct NCVS sample estimates at the state or other small area level and a model for the variance-covariance relationships for the estimates over time. The variance-covariance model that had been developed was serviceable only for the core NCVS design through 2015, because it did not anticipate the effect of the introduction of the 2016 sample and the differential sampling rates by state to achieve the supplementation in 22 states. Secondly, further analysis to validate the small area model remains desirable. Third, to extend the utility of the statistical methods to other variables and geographic levels, the software implementation for the 1999-2013 estimates can be improved in design, generality, and documentation.

Under this new award, the proposed research will address four major goals, objectives, and deliverables identified by the guidance for invited applications for this project: 1)To extend the current SAE estimates to cover the period from 1999 through 2015; 2)To understand the extent to which increased sample sizes in the largest states alter the expected state estimates; 3) To extend the original models to produce prevalence estimates and estimates for a wider set of victim and incident characteristics; and 4) To document the extended system to ensure that technical staff with adequate backgrounds in small area estimation, R, and SAS could continue to use the system for new estimates and extensions in the future.

(NCA/NCF)

Note: This project contains a research and/or development component as defined in applicable law.