Which of the Three KABC-II Global Scores is the Least Biased?

Abstract

In order to create fairer measures for the assessment of ethnic minority group children, the test authors of the Kaufman Assessment Battery for Children-Second Edition (KABC-II) created three different global indexes: the comprehensive Fluid-Crystallized Index (FCI); the Mental Processing Index (MPI), which excludes subtests that assess crystallized knowledge; and the Nonverbal Index (NVI), which excludes any subtest that requires verbal expression. The test authors encourage clinicians to use the MPI for children with cultural differences and the NVI for children with language differences. The implication is that the FCI is more biased than the MPI and NVI for assessing children from ethnic minorities. However, there has been no empirical evidence to support that hypothesis. Using structural equation modeling, we investigated the predictive validity of the three KABC-II global scores in predicting reading, writing, and math in a nationally representative sample of Caucasian, African-American, and Hispanic school-aged children across three grade groups (1-4, N=724; 5-8, N=743; and 9-12, N=534). Contrary to the test authors' predictions, the FCI emerged as the fairest predictor of achievement across all age and grade groups. Alternatively, the MPI and NVI showed a persistent intercept overprediction of African-American and Hispanic students' achievements across grades 5-8. The results provide evidence that more comprehensive global scaled scores might, in fact, be fairer predictor variables than less culturally and linguistically influenced indexes when predicting the achievement of children and adolescents from ethnic minorities.

Do Global Ability Composites Differentially Predict Academic Achievement Across Gender?

“…Thus, residual variances that were equal between genders remained constrained in subsequent model tests (i.e., slope and intercept bias tests). However, because there was a statistically significant degradation in model fit ( p < .05) for three residual variance tests, these parameters were freed across genders for subsequent model tests as this was not an assumption that must be met (Scheiber & Kaufman, 2015).…”

“…In addition, concerns associated with differential prediction bias may be counterbalanced if the disadvantaged group is overpredicted, because the test works in favor of the disadvantaged group and therefore is a fair assessment (Scheiber & Kaufman, 2015; Warne et al, 2014). However, to appropriately determine whether the student is affiliated with the disadvantaged group and affected by differential prediction bias, one must consider all possible ways in which an individual’s demographic variables interact with cognitive abilities and achievement outcomes.…”

“…If the residuals were not equal between gender groups, as evidenced by model degradation via a nested model comparison, then this constraint was released and freed in subsequent models. However, if the homogeneity of the residual assumption was supported, then that constraint was retained in subsequent model tests (Scheiber & Kaufman, 2015).…”

“…However, support for slope bias would suggest an interaction (moderation) present between gender and achievement, and support for intercept bias would suggest that a common regression line overpredicts for one gender group and underpredicts for the other gender group. In cases of slope or intercept noninvariance, post hoc analyses were conducted to probe the direction of gender bias with simple regression equations (Scheiber & Kaufman, 2015).…”

“…Weiss and Prifitera (1995) also reported an underprediction of females’ writing scores when using a common regression equation, suggesting their writing scores may be slightly underestimated. More recently, several studies reported negligible findings of differential prediction ethnicity bias (e.g., Naglieri et al, 2005, 2007; Scheiber & Kaufman, 2015; Watkins et al, 2007). However, less frequently considered is whether gender moderates the influence of global cognitive ability on academic achievement, despite the expansion of research on gender differences in cognitive ability and achievement variables.…”