Assessing Sample Design Options for the National Crime Victimization Survey

The National Crime Victimization Survey (NCVS) began full-scale data collection in 1972. Over time, increasing costs of data collection and a diminishing budget have forced a gradual reduction of the sample size from 72,000 households in 1972 to 38,600 in 2005, leaving the NCVS far less able to meet its initial goals. The sponsor of the NCVS, the Bureau of Justice Statistics (BJS), asked a panel of the National Academy of Sciences to review the survey. The panel’s interim report, released in 2008, recommended to BJS a series of actions, including a systematic review of a range of sample design options for the survey. Since the beginning of the NCVS, the Census Bureau has been the primary architect and data collector. The panel report addressed the question of whether BJS should consider an alternative collector, but it detailed reasons for and against continuing with the Census Bureau. The report recommended, however, that the Census Bureau provide BJS with detailed information on costs to allow external assessment of survey design options. The report also recommended that BJS consult external sources to evaluate design options. On that basis, BJS has funded the research to be reported here. This paper summarizes our research design for the study. The panel report provided two forms of guidance for this research. For one, it listed the essential components of the design of a national household survey—the design, stratification, and selection of primary sampling units; methods for selecting households; the choice between a rotating panel or cross-sectional design; and possible subsampling of individuals within households. The panel report also recommended consideration of specific alternative designs, including the design of the British Crime Survey. Our planned research responds to both forms of guidance, attempting to build on the panel’s thoughtful review of the current status of the NCVS.