The effects of forming multi-predictor composites on group differences and adverse impact

A common assumption exists which asserts that the formation of composites of predictors represents a method for dealing with adverse impact. It is often expected that including predictors that demonstrate smaller group differences with others that demonstrate larger group differences will help to alleviate the amount of adverse impact observed at the composite level. The purpose of this paper is to answer the question "If two or more predictors are combined to form a composite, what will be the magnitude of group differences and, consequently, of adverse impact, of using that composite for selection?" In answering this question, a set of tables, figures, and formulas are presented that highlight variables influential in affecting how composites of predictors influence observed group differences. A number of conclusions are drawn that clarify the extent to which forming composites decreases group differences and subsequently adverse impact.