Why negative affectivity should not be controlled in job stress research: don't throw out the baby with the bath water

Abstract

In 1987 Watson, Pennebaker, and Folger wrote an in uential paper in which they noted the potential importance of negative a€ectivity (NA) in job stress research, going so far as to suggest the provocative hypothesis that NA biased self-reports of most job stressors (and other measures of work conditions) and job strains. A number of concerned researchers, noting correlations between NA and other variables, have recommended that we statistically control for NA bias in general stress (McCrae, 1990) and job stress (e.g., Brief et al., 1988; Payne, 1988) studies with some form of partialling. While we agree that a€ective dispositions are important, we disagree that it should become routine to treat them as bias factors to be statistically controlled. Although a number of researchers have noted possible non-bias or substantive roles for NA (e.g., Moyle, 1995; Schaubroeck, Ganster and Fox, 1992; Spector and O'Connell, 1994; Williams, Gavin and Williams, 1996), many researchers have started to routinely control for a NA bias in their studies. However, if indeed NA has a substantive role, one should not partial it as this can lead to removing the e€ects of the very variables one wishes to study. In this paper we will present arguments that partialling or related statistical techniques merely provide unrealistic hopes and illogical inferences in studying the potential biasing e€ect. Moreover, we will discuss the role played by NA in the job stress process and demonstrate that it should not just be considered a bias in need of statistical control. Rather NA can play a variety of substantive roles in the job stress process.