An Examination of Differential Item Functioning in a Measure of Self-Reported Offending Across Race and Ethnicity Among a Sample of Justice-Involved Youth

Self-report measures of offending are widely used in criminological research and have been employed to examine potential differences across groups in offending. An assumption of these comparisons is that the items that comprise these measures function similarly across relevant groups. We evaluated differential item functioning (DIF) of a measure of self-reported offending across White, Black, and Hispanic youth involved in the justice system (n = 1289) using multigroup item response theory analysis. We then demonstrated the utility of moderated non-linear factor analysis by testing if relevant continuous-level covariates (i.e. socioeconomic status, neighborhood disorder, and impulse control) accounted for DIF associated with race/ethnicity. The findings suggest that several items displayed DIF, and that the DIF associated with race/ethnicity persisted even after controlling for covariates. We also compared two models that predicted non-DIF-adjusted scores to DIF-adjusted scores. The results indicate that summed raw scores of offending may bias estimates across groups.