NCJ Number
85335
Date Published
1981
Length
18 pages
Annotation
This essay reviews theoretical models of measurement that described an individual response to an item in conceptual terms that cast response errors in a manner useful for identifying their effects, and the models also suggest how the effects of response errors may be estimated.
Abstract
Most self-report studies overlook the nature and extent of response errors. While this may result from the difficult measurement problems presented by self-reports of crime, avenues do exist for empirical inquiry on the effects of errors in self-report surveys. This study identifies a measurement model of reported offenses with which such inquiry may proceed. Estimates of the elements in the model and of the validity and reliability of responses are derived using multiple indicators of reported and official offenses. Weak associations between reported and true levels and types of offenses are observed in the findings of a survey conducted on the offense histories of a sample of survey respondents. These associations indicate that response errors have systematic distorting effects on self-reports of offenses. Estimates of reliability and validity are generally low but vary dramatically among different measures of the incidence, recency, type, and social context of offenses resulting in arrest. Reports of the recency of the last or most recent offense are found to be most reliable and valid. Weak negative relationships are observed between subjects' true scores and response errors. (Author summary modified)