Evaluation of age, sex, and race bias in the Personality Inventory for Children.

Abstract

The question of whether children's ability tests are biased has received much attention, but the same issue regarding children's personality tests has been neglected. In this study, we evaluated whether the external validity of the Personality Inventory for Children (PIC), a parent-informant measure of child psychological status, was moderated by child age, sex, or race. Using a hierarchical multiple-regression algorithm within a clinic-referred sample, we found that child race and sex generally did not moderate the relation of PIC scales to symptom checklists completed by clinicians, teachers, and parents. Some expected relations were modified by child age. Implications for the potential use of the PIC as an unbiased personality measure are discussed. The question of whether children's intelligence tests are culturally biased has generated much controversy over the last 30 years. A large body of literature has been devoted to such topics as operationalization of test bias (e.g., content, construct, and predictive bias; Reynolds, 1982; Reynolds & Kaiser, 1990), appropriate statistical methods for the detection of bias (e.g., Berk, 1982), recommendations for examiners who assess minority children (e.g., Sattler, 1988), and the impact of numerous court cases on use of intelligence tests (e.g., Bersoff, 1982). In contrast to research with children's ability tests and with adult personality measures (especially the Minnesota Multiphasic Personality Inventory [MMPI]; e.g., Dahlstrom, Lachar, & Dahlstrom, 1986), there has been little evaluation of whether children's personality tests are biased (Moran, 1990). Two reasons may account for the lack of attention to this problem. First, bias in children's cognitive tests or adults' personality tests can affect obvious and important decisions, including, for example, access to special education or promotion considerations. Thus, much consideration of the problem of bias is expected. Bias in personality tests for children can also be serious, however, and can affect matters such as treatment recommendations or access to services for emotionally disturbed children. Second, the psychometric properties of many personality tests are flawed, which limits use of appropriate statistical techniques (Moran, 1990). For example, low test reliability can complicate the interpretation of results of moderator variable analyses of whether test validity changes by child race (or age, sex, etc.). Low test reliability places a ceiling on a test's external validity, which makes detection of differential validity more difficult. Subjectively scored tests (e.g., projectives) present similar problems if their interrater reliabilities are low. We evaluated in this study possible bias effects in the Personality Inventory for Children (PIC; Wirt, Lachar, Klinedinst, & Seat, 1984), a parent-informa nt measure of child cognitive,