Model-Based Small Area Estimation (SAE) for the National Crime Victimization Survey (NCVS)

• In Event: Enhancing the Measurement of Crime in the U.S.

Abstract

As part of the NCVS subnational program, BJS developed model-based methods using small area estimation (SAE) to produce subnational crime estimates. In this presentation, we will cover the benefits and drawbacks of model-based estimation compared to directly derived survey estimates. For example, model-based estimates can be produced for all 50 states whereas direct estimates cannot. Direct estimation, however, allows for flexibility in the types of analyses which can be conducted while model-based estimation involves a fixed set of outcomes. Additionally, we will present how analysts can use the SAE methods developed specifically for the NCVS to better understand the model-based estimates already produced or to produce estimates for their own analyses. This presentation will not delve into the technical details of SAE, but, rather, describe the specific process for developing SAE for the NCVS and detail how users can replicate the process.

Authors

• Dan Liao, RTI International

• Marcus Berzofsky, RTI International

• Lynn Langton, RTI International

• Chris Krebs, RTI International

• Grace Kena, U.S. Department of Justice, Bureau of Justice Statistics

• Heather Brotsos, Bureau of Justice Statistics