ABSTRACT
We hypothesized that sexual harassment is part of a broader behavioral family including aggression and discrimination. We examined whether the relationships between these types of mistreatment can be represented well by a general factor that relates to other workplace variables. Evidence from military datasets showed that sexual harassment, sex discrimination, and workplace aggression can be conceptualized as a more general factor that functioned well in an integrated model of sexual harassment and was experienced differently by men and women. Thus, there is utility in examining these types of mistreatment both together and independently, both for research and prevention purposes.
SUPPLEMENTAL DATA
Supplemental data for this article can be accessed at www.tandfonline.com/hmlp.
Notes
1. One of the original 16 was subsequently replaced by DMDC for policy reasons and was thus not included in the 2002 survey on which the present analysis is based.
2. Because the specific factors are orthogonal to the general factor and represent the variance left over after accounting for the general factor, we use the terms specific factors and residual factors synonymously.
3. An alternative to the bifactor model is a second-order factor model where the three specific factors (i.e., sexual harassment, sex discrimination, and workplace hostility) are modeled as indicators of the general construct. Although these models are closely related, the bifactor solution has several advantages over a higher-order model for comparing both general and specific factors (Chen, West, & Sousa, Citation2006): the bifactor model can investigate the role of specific factors that are independent of the general construct; the relationships between the indicators and the specific factors are explicitly tested in the bifactor model; and the bifactor model can be used to assess the incremental validity of the specific factors over and above the general factor. Moreover, the second-order factor model is a nested submodel of the bifactor model that has unnecessary proportionality constraints on the factor loadings.
4. Additional paths were also estimated to control for relationships that were unrelated to the general construct, based on previous research: (a) paths from both job level and organizational tolerance for harassment to supervisor, coworker, and work satisfaction were estimated (Fitzgerald et al., Citation1997; Williams, Fitzgerald, & Drasgow, Citation1999); (b) paths were estimated from psychological distress to the three satisfaction variables and organizational commitment (Mathieu & Zajac, Citation1990); and (c) paths were estimated from the job attitudes (i.e., supervisor satisfaction, coworker satisfaction, work satisfaction, and organizational commitment) to turnover (cf. Griffeth et al., Citation2000).
5. A statistical test of measurement equivalence also found that there was significant nonequivalence between male and female samples; thus, men and women were subsequently analyzed separately. Contact the authors for details.
6. In a separate study, we examined the same model but in longitudinal data. Results were complementary and so are presented in online supplemental material only.
7. Other limitations in our sample have to do with the potential biases attributable to common method variance and to negative affectivity. We undertook supplementary analyses to assess these potential sources of bias, which found that neither potential source contributed appreciably to our results. Contact authors for full details of these analyses.