| Job stress, i.e., the adverse reactions employees experience in response to job stressors (Spector, Chen, & O’Connell, 2000a), is thought to be one of the antecedents of organizational commitment. Although organizational commitment is a multidimensional construct that consists of multiple forms (Allen & Meyer, 1990; Johnson, Chang, & Yang, 2010), in this study, organizational commitment indicates “identification with, involvement in, and emotional attachment to the organization^ (Allen & Meyer, 1996, p. 253), or affective organizational commitment. The understanding of the relationship between job stress and organizational commitment has been both inconsistent and incomplete. Although a large number of studies have reported a negative impact of job stress indices on organizational commitment (e.g., Cropanzano, Rupp, & Byrne, 2003; Jamal, 1990, 2005; Tourigny, Baba, Han, & Wang, 2013), other studies have not supported this link (Majchrzak & Cotton, 1988; Parasuraman & Alutto, 1984). Elangovan (2001) found that job stress did not have a direct relationship with organizational commitment; instead, job satisfaction, i.e., the extent to which individuals have positive affective emotions and attitudes toward their jobs (Cramer, 1996), mediated this relationship. However, most prior studies on the relationship between stress and commitment have used a cross-sectional design, and their ability to identify the direction of this relationship has thus been problematic. This issue is particularly relevant to the proposed mediating role of job satisfaction. Although the causal relationships implied by mediation take time to unfold, the cross-sectional approach assumes that these relationships are instantaneous (Selig & Preacher, 2009). Accordingly, the application of a mediation model to cross-sectional data may lead to severe bias (Maxwell & Cole, 2007; Selig & Preacher, 2009), ultimately suggesting a substantial indirect effect between two constructs through a mediator even when there is no real mediation effect (Maxwell, Cole, & Mitchell, 2011). Surprisingly, however, the majority of field studies on mediation have been based on cross-sectional data (Maxwell et al., 2011). Even the few cross-lagged studies in the literature have used only two time points and thus have not been sufficiently informative with regard to the direction of influence among the study variables. The use of two waves is insufficient because this type of design cannot determine the form of change over time (i.e., steady, delayed, or consistent) (Gottman & Rushe, 1993; Singer & Willett, 2003). Furthermore, a twowave design cannot distinguish true change from measurement error, which may lead to the erroneous conclusion that true change is occurring, although a longer temporal view would suggest the opposite. Therefore, Singer and Willett (2003) insisted that two-wave panel studies are only marginally better than cross-sectional studies. Accordingly, Ployhart and Vandenberg (2010) limited longitudinal research to studies of change that contain a minimum of three repeated measurements. |