NCJ Number
83725
Date Published
1981
Length
124 pages
Annotation
This study evaluates the accuracy of measurement and substantive sources of variation for 33 survey items of attitudes toward crime and criminal justice lifted from several well-known general purpose surveys.
Abstract
Specific objectives of the research were to (1) document substantive sources of variation in common survey measures which are often confounded with variation attributed to criminal justice program effects, (2) document the impact of systematic sources of measurement error, (3) estimate the amount of random measurement error, and (4) model these random components. The study decomposed covariation among the 33 survey measures of 9 underlying attitudes toward crime and criminal justice, separating unreliability in the individual items from true stability and change in unobservable attitudes. Each of the nine attitudes were remarkably stable over the 17-week period, even though individual responses to the less reliable indicators were highly volatile. Differences in attitudes between respondents who differ in city, gender, or ethnicity were typically much larger than differences between those who had recently been victimized or affected by the criminal justice system and those who had no such experience. Overall, the usual survey measures do not seem to be good candidates for the evaluation of criminal justice programs, because individual items are too volatile while the underlying dimensions are too stable. Recommendations for improving the accuracy of such surveys are offered. A total of 36 references are listed.