Difference Scores: Rationale, Formulation, and Interpretation

Difference scores have long been of interest in management research because of the need to assess agreement or congruence between constructs. For example, some form of difference, dissimilarity, or discrepancy score is typically at the core of investigations contrasting subordinate and supervisor work values (e.g., Posner, Kouzes & Schmidt, 1985), perceived and desired job characteristics or attributes (e.g., Dawis & Lofquist, 1984), job demands and worker abilities (French, Caplan & Harrison, 1982), and self and referent other rewards (e.g., Goodman, 1977). In such instances, difference scores are typically used as correlates or predictors of individual or organizational outcomes, such as changes in personal health and adjustment, and job-related attitudes and performance. Despite decades of use, the appropriateness of difference scores for estimating discrepancies between measurement units continues to be a source of much debate. Some have advised that the statistical and psychometric properties of difference scores are so problematic that their use should be discontinued (e.g., Cronbach & Furby, 1970; Johns, 1981). Others have concluded that such criticisms are unfounded, declaring difference scores to be both relatively reliable and unbiased (e.g. Rovine, in press; Zimmerman, Brotohusodo & Williams, 1981). More recently, Smith and Tisak (1993) and Edwards (1994) have each commented-independent of the other’s work-on the adequacy and appropriateness of using difference scores in management research, and arrived at opposite conclusions.