Construct validity of the scale was

Construct validity of the scale was explored through factor analysis using component principle technique with varimax rotation in the PRC and Taiwanese sample separately. In both samples, the Bartlett's test of sphericity reached significance (p < 0.001), and the KMO value was acceptable (PRC: 0.67, TW: 0.71). Scree plot suggested a two-factor solution for both samples. For the PRC sample, the two factors had eigenvalues of 2.45 and 1.94, accounted for 54.4 percent of the total variance. For the Taiwanese sample, the two factors had eigenvalues of 2.68 and 2.24, accounted for 59 percent of the total variance. The two factors contained exactly the items from the two subscales, with average loadings of 0.74 (PRC) and 0.75 (TW) for the CPC subscale, and 0.70 (PRC and TW) for the CSC subscale. Thus, in both the PRC and Taiwanese samples, the factor structure perfectly corresponded to theoretical constructs of CPC and CSC.
Additional information regarding the reliability and validity of the scale can be found in a recent study with three independent Chinese samples (n = 740) (Lu et aL, 2001a, b). The average internal consistency reliability coefficient is 0.73 for the CPC subscale, and 0.67 for the CSC subscale. Furthermore, CPC positively correlated with the LOC subscale in the OSI-2 (r = 0.15, p < 0.001). CSC on the other hand, negatively correlated with LOC (r = -0.26, p < 0.001). However, CPC and CSC were two distinct constructs with no correlation between them (r = 0.05, ns). This evidence also corroborates the above factor analysis results.
Finally, demographic information such as age, gender, education, marital status, years in the current job, position in the organization, size and ownership of the organization were collected.

The questionnaires were written in Chinese. The wording and expressions were somewhat different in the PRC and Taiwan versions, to reflect their individual historical development. All the measures used six-point Likert scales.
Results
Sample distribution
Table II shows some of the demographic and job characteristics for the two samples from Greater China. Age was similar in the two groups, but there were 13 percent more males in the PRC group. In the Taiwanese group there were substantially more employees accomplished postgraduate level education, and more were still single. In terms of job characteristics, seniority and weekly working hours were similar in the two groups. However, in the PRC group there were more managers and more worked in medium sized organizations with 100-500 employees.

Overall, compared with their Taiwaness counterparts, the PRC employees in this study were mostly male, married, college-educated, managers serving in medium-sized public or foreign companies. However, both groups were mainly composed of middle-aged, senior employees.

As there were disparities in education level and management status between the two groups, preliminary analyses were conducted to clarify their possible effects. In the case of education attainment, analyses revealed no significant correlation with research variables in either the PRC or the Taiwan group. In the case of management status, again there was no significant correlation with research variables in the PRC group. However, Taiwanese managers did report higher overall work stress (r = 0.23, p < 0.05). Although any potentially distorting effects of education and management status seemed negligible, we took a cautious stance to enter them first into multiple regression equations when predicting strain. Neither education nor management status emerged as a significant predictor of any of the strain variables for the PRC or Taiwanese workers. They therefore were excluded from the regression models to be presented later in the paper.
Reliability of scales
On diagonals of Table III reliability of scales measured in Cronbach's alphas is presented In both samples, "CPC" or "CSC" had the lowest alpha of.66 whereas "work stress" had the highest alpha of 0.93. Researchers don't yet agree on an acceptable band for reliability coefficients. For instance, Nunnally (1978) suggested chi higher than 0.4 as acceptable, whereas Royle (1991) advocated a range of 0.3-0.7 as good indicators of reliability. We concluded that the reliability for work stress, job satisfaction and well-being scales are adequate, whereas that for control beliefs is marginal (chi = 0.66). This marginal reliability may be attributed to fewer items (four each) in both the CPC and CSC subscales.

However, a closer examination of the work-stress subscales revealed that three of them had low alpha coefficients (chi < 0.60 in at least one group). These were "managerial role", "hassles" and "organizational climate". Further analyses thus used the entire "work-stress" scale and the remaining five subscales: "workload", "relationships", "home/work balance", "personal responsibility" and "recognition". More importantly, the patterns of reliability across scales were similar in the two groups. This similarity ascertained the reproducibility of constructs of work stress, job satisfaction, well-being, and control beliefs in the two Chinese societies.