An investigation of the effects of psychological contract and organization-based self-esteem on organizational commitment in a sample of permanent and contingent workers.

A correlation analysis of the demographics with the variables explored in this study revealed that none of the demographics shared significant relationships with the outcomes. Therefore, no covariates were considered in the following analysis.

Factor Analysis and Scale Reliabilities

Wilcox (1998) argued that researchers should collect and give consideration to the data before determining the best tool for estimation. Oftentimes the best analysis tool depends on the situation presented by the data. Limitations of this study include a relatively small sample size and the likelihood that the organization members who elected to respond to the survey were not random. This potential for imbalance may have resulted in changes in the dependent variables (Kerlinger & Lee, 2000).

A useful method for analyzing data under these circumstances is partial least squares (PLS), which is highly flexible; places less rigorous demands on measurement scales, sample size, and residual distributions (Chin & Newsted, 1999); and does not require the restrictive assumptions of data distributions, observation independence, or variable metrics (Barclay, Higgins, & Thompson, 1995). We used PLS-Graph (Version 3.0; Professor Wynne W. Chin, University of Houston; e-mail: wchin@uh.edu) to conduct the analyses.

Descriptive statistics, scale reliabilities, and average variance extracted from each variable are reported in Table 2. Composite scale reliabilities were assessed using the traditional .7 cut-off value. All scales exceeded this criterion.

PLS can also be used to explore the convergent and discriminant validity of measurement items employed in a study (cf. Howell & Avolio, 1993). Each variable's average variance was extracted from its items. The average variance shared by a variable with its items should be greater than its correlation with other variables, thus indicating adequate convergent and discriminant validity. As evidenced in Table 2, all variables met this criterion. It is important to note, however, that the average variance extracted from continuance commitment was not considerably greater than its correlation with value commitment. One reason for this stems from the previous discussion about the multidimensionality of organizational commitment and the argument that the construct is a one-factor, affective phenomenon.

We also used PLS to generate factor loadings that can be interpreted similar to principal-components analysis (Bookstein, 1986). Those loadings with a value of .7 or greater indicate that less than half of an item's variance is because of error. All items met or exceeded the cut-off value. In each case, the loading was considerably greater on its intended scale than on any other scale. The individual item loadings on the latent variables were statistically significant, with the exception of one item from the Continuance Commitment scale CI am grateful for the opportunity this organization has provided me."). This item was removed. This did not come as a surprise, because Mayer and Schoorman (1992) dropped the same item for the same reason: The factor loading was small and nonsignificant. No other indicators were removed from the model because their inclusion was not apparently detrimental to the PLS analyses.

[FIGURE 2 OMITTED]

Participants were asked whether the organization fulfilled its obligations to them, to which they answered with a yes-or-no response. Significant differences between the yes and no responses were found across both types of obligations. Furthermore, these responses were significantly correlated with both the relational (r = .43, p < .01; t = 4.21, p < .001) and transactional (r = .45, p < .01; t = 4.14, p < .001) obligations. In other words, the responses to the yes-no question were indicative of the results of the psychological contract measurement scales. As a result, we are confident that our measurement tool reliably captured worker attitudes about our dependent variable, the psychological contract.

Hypothesis Testing

Chin (1998) suggested that standardized paths in PLS should have values of about 0.20, and ideally above 0.30, to be considered meaningful. Lower path values (<0.20) represent a 1% or lower explanation of variance and may result from multicollinearity or some unknown residual effect. Path values above 0.90 might be considered spurious. Although the statistical significance of path coefficients is important in interpreting findings, the question we are most interested in is whether the relationships are theoretically interesting.

Using the aforementioned criteria for evaluating the paths in the PLS, we will turn to a discussion of the model. A resampling technique known as bootstrapping was employed with these data. Using t tests, we found that all but one of the paths was statistically significant and confirmed that these paths were representative and useful in the analyses.

The research model, with findings, is depicted in Figure 2. The paths are labeled as betas ([beta]) and are referenced accordingly throughout the following section. All hypothesized relationships were statistically significant, with the exception of the relationship between transactional psychological contract perceptions and OBSE ([beta] = .12). This was problematic by PLS standards. The path between the items measuring relational psychological contract perceptions and OBSE was significant and substantive ([beta] = .39). OBSE was related similarly to both the continuance and value factors of organizational commitment ([beta] = .70 and [beta] = .78, respectively). Therefore, with the exception of the nonsignificant relationship, all paths in this model were substantive and significant.

Hypotheses 1a and 1b. These hypotheses suggested that the participants who were contract workers would indicate higher levels of transactional, psychological contract obligations to their employer than would the permanent employees. Conversely, Hypothesis 1b suggested that permanent employees would indicate higher levels of relational obligations than their colleagues with temporary employment status. Although the means for each group represented these assertions (Table 3), the differences were nonsignificant, thus indicating that these discoveries were no greater than by chance. Therefore the first two hypotheses were not supported by these data.

Hypotheses 2a and 2b. We asserted that there would be a larger, positive relationship between perceptions of relational obligations and value commitment than with continuance commitment, regardless of employment status. This hypothesis was supported as a result of a bootstrap analysis using PLS. The path relationship between relational perceptions and the affect-based value commitment ([beta] = .37, SE =. 14, p [less than or equal to] .05) was greater than the path between relational perceptions and continuance commitment, or the desire to remain with the organization ([beta] = .34, SE = .12, p [less than or equal to] .05).

Conversely, we hypothesized that there would be a larger, positive relationship between continuance commitment and perceptions of transactional obligations, regardless of employment status. This phenomenon was hypothesized to occur regardless of employment status. This hypothesis was also supported. The relationship between transactional perceptions and the continuance commitment ([beta] = .30, SE =. 13, p [less than or equal to] .05) was greater than the path between perceptions of transactional obligations and value commitment ([beta] = .25, SE = .14, ns).

Hypotheses 2a1 and 2b1. In addition to the relationships hypothesized in 2a and 2b, we also asserted that permanent employees' perceptions of relational obligations by their employer would reveal a stronger positive relationship with value commitment. Additionally, we asserted that contingent workers' perceptions of transactional obligations by their employer would reveal a stronger positive relationship with continuance commitment.

Relational perceptions by full-time employees were more strongly related to the affect-based value commitment ([beta] = .37, SE =. 14, p [less than or equal to] .05) than were the contingent workers ([beta] = .29, SE = .22, ns). Therefore, Hypothesis 2a1 was met. Conversely, transactional obligations of the contingent workers ([beta] = .30, SE = .33, ns) and the permanent employees ([beta] = .30, SE = .14, p [less than or equal to] .05) were nearly identically related to continuance commitment. Additionally, the relationship between these two variables for contingent workers was nonsignificant, whereas the relationship for permanent employees was significant. Therefore, Hypothesis 2b1 was not supported by these data.

Hypotheses 3a and 3b. Two mediator effects were explored in this study. We hypothesized that OBSE would mediate the relationships between perceptions of relational psychological contract fulfillment and value commitment (Hypothesis 3a) and between perceptions of transactional psychological contract fulfillment and continuance commitment (Hypothesis 3b).

Complete mediation occurs when an independent variable no longer affects a dependent variable after the mediator has been added to the model. Here, we hypothesized the more common situation of partial mediation, which occurs when the direct influence of the independent variable on the dependent variable is reduced but still significantly different from zero when the mediator is introduced to the relationship.