NCJ Number
77170
Journal
Statistician Volume: 29 Issue: 4 Dated: (1980) Pages: 313-350
Date Published
1980
Length
38 pages
Annotation
This article examines the effectiveness of the National Crime Survey (NCS), identifies reasons why the survey has been less successful than desired, and proposes changes which might make the NCS more useful.
Abstract
The NCS is a stratified multistage cluster sample survey using rotating panels of household locations to generate household victimization data on a longitudinal basis. Because the survey measures location victimization, household and individual attrition over the course of the study is not considered a major factor; however, the characteristics of the households and individuals leaving and entering the population may have a significant bearing both on the validity of the data and on the causes of crime. The rotating mechanism--designed to prevent subject resistance and rejection--has resulted in the development of rotation group bias in the survey's results. Because the degree of bias present varies from interview period to interview period, the estimates used to counterweigh this phenomena are not effective. Rotations should follow individuals and households rather than household locations. A followup study of individuals and households who have left the population might be helpful in correcting for the effects of attrition. Analysis of series victimizations using NCS data is difficult since a series of incidents may be reported twice if they fall across two sampling periods, may be reported as a series and one or more isolated incidents, or may be reported only as isolated incidents. A revised questionnaire developed in 1979 deals with this question but does not consider problems related to imbedding series victimizations into patterns of multiple victimization. An NCS report examining victimization trend rates exhibits technical and expository confusion and is badly marred by statistical flaws. A recently completed project which matched individual and household files over the course of the rotation period has improved the survey's longitudinal validity. These longitudinal files could be used to study repeat victimization and respondent effects of the NCS interview. Tabular data, graphs, and 35 references are included.