Newsroom
RESEARCH TRIANGLE PARK, NC — The preeminent survey on crime in the United States could be streamlined, helping improve its results without adding appreciable administrative cost, RTI International researchers found in a report for the U.S. Bureau of Justice Statistics (BJS).
The report, published Nov. 15, concerns the National Crime Victimization Survey (NCVS). Instead of relying on police statistics, this in-depth, annual survey draws on interviews with approximately 160,000 people each year, providing insights on crimes regardless of whether they were reported to authorities. This is especially important for violent crimes, which are reported much less often than property crimes.
The study is part of an ongoing project at RTI. Since 2008, RTI researchers have been helping the Bureau of Justice Statistics look for ways to change the design of the National Crime Victimization Survey to maximize the accuracy and precision of the survey estimates without appreciably increasing the survey cost.
Participants currently undergo seven interviews over a three-and-a-half-year period, after which they are replaced by another set. Based on the report, RTI recommended reducing the number of interviews to four, a change the authors said will be more efficient but just as rigorous.
“Our research found that demographic groups associated with higher crime rates drop out of the panel after four waves at much higher rates than lower crime rate groups,” said Marcus Berzofsky, a senior research statistician at RTI and the lead author of the report. “By reducing the number of panel waves and rotating sampled households out before a subset self-selects out of the study, the NCVS can reduce the potential for statistical bias due to sample attrition and fatigue.”
The researchers found that the proposed design would increase the cost of the survey slightly, but recommend the change anyway because of the potential for better data.
In the report, the researchers also evaluated the merits of study designs with one, three, or five interviews. The four-interview design proved to be the strongest in terms of how close its results matched those of the current design. The researchers also recommended a more thorough cost assessment and further study of how costs change across panel waves if the Bureau of Justice Statistics decides to pursue the redesign option.
This report is the first of a series which RTI is working on with BJS to either help improve the quality of the NCVS data or make the data more relevant to the public.
“BJS is committed to ensuring the NCVS maintains a high methodological rigor,” Berzofsky said. “We are glad RTI can contribute to the improvement process.”
