A latent class model of child offending

Publisher Website
Google Scholar
Abstract
A latent class model is proposed to estimate the magnitude and direction of errors of measurement in reports of young offending It is shown that, subject to the availability of information from reports of offending behaviour by two sources and information on a known predictor of offending, it is possible to fit models which estimate the magnitude of errors of reporting and the true prevalence of offending. This model is applied to self‐report data collected for over 700 New Zealand 12‐year‐old children. This analysis suggested that under‐reporting of offending was very high and it was estimated that over 50% of children who offended failed to report offending. The analysis is extended to show that the consequences of errors in reporting behaviour are to lead to a serious underestimation of the strength of relationships between risks of offending and predictor variables. The implications of these results for the interpretation of report statistics are discussed.