Measurement and nonmeasurement processes with negative affectivity and employee attitudes.

Abstract

The partial-correlation approach to investigating negative affectivity (NA) is discussed, and 2 alternative latent-variable techniques are presented. The results of these 3 analytic techniques were compared using a data set consisting of NA, job satisfaction, affective commitment, role stressors, leader-member exchange, and job complexity. The findings using the partial-correlation technique supported a biasing effect of NA on relationships between the substantive variables. Alternatively, although 2 latent-variable approaches provided evidence consistent with the measurement contamination and substantive influences of NA, relationships between the predictors and outcomes were not biased by NA. Both the measurement and substantive relationships effects of NA found in this study and implications for future research on self-reports are discussed.