At present, the academic research on employee job satisfaction was more and more mature. However, the academic research on employee job satisfaction in group level is very rare. At the beginning of this century, Mason and Griffin [8] first put forward the concept of “Group Task Satisfaction” (GTS). They pointed out that GTS was a group-level concept, i.e. group members’ consistent attitude to work task and work environment instead of the aggregation or the average of employee individual job satisfaction. Mason and Griffin [8] argued that measuring GTS should ask group members to assess their perceptions of job satisfaction in group level, namely using transformation reference object consensus method to measure. Empirical study about antecedents of group task satisfaction was little. So, the present study mainly explores the influence of two group-level variables (i.e. group cohesion and group job characteristics) on group task satisfaction.

Concept of cohesion was first put forward by Lewin [9] who argued that cohesion was the “glue” of keeping group members together and maintaining the relationship between them. According to the team-member exchange theory, each member in the group will produce different emotional reactions when contacting with other members in the group, and one may spread to other members and produce similar sentiment. In a high-quality team-member exchange relationship group, positive emotion diffuses among group members, harmonious group climate forms, emotional cohesion increases among members, and the group members are more likely to gain the satisfaction from group relationship. On this basis, the author puts forward the hypothesis:

H1: Group cohesion has a significant positive influence on group task satisfaction.

Group job characteristics, as the name implies, are inherent attributes of a group work or task. At present, a typical job characteristics theory proposed by Hackman and Oldham [10] showed that dimensions of job characteristics (i.e., task importance, autonomy, diversity, completeness, feedback) would affect employees’ outcomes through “the key psychological state”. Therefore, group job satisfaction, as a key psychological state in a group level, might be influenced by group job characteristics. Mason and Griffin [8] had also pointed out that group task satisfaction and job characteristics were correlated. For example, a challenging and significance tasks are more attractive than a regular task without any feedback. Job involvement, tasks diversity, task significance and other job characteristics would affect group members’ job satisfaction. On this basis, the author concludes that:

H2: Group job characteristics have a significant positive influence on group job satisfaction.

Mason and Griffin [8] argues that group job satisfaction has a direct impact on employee’s job satisfaction. The consistent work attitude among group members would prompt them to obey conduct code in the group and keep consistent with work attitude in the group, thus in a high overall satisfaction group, the members are more likely to experience the satisfaction with work. On this basis, the author puts forward hypothesis:

H3: group job satisfaction has significant positive influence on job satisfaction.

From the above, the author argues that group job characteristics and group cohesion do not directly affect employee job satisfaction, but indirectly affect it through group task satisfaction. On this basis, the author puts forward hypotheses:

H4a: Group task satisfaction mediates the positive influence group job characteristics on job satisfaction;

H4b: Group task satisfaction mediates the positive influence group cohesion on job satisfaction.

The more satisfied employees are more willing to stay in the enterprise instead of quitting [11] , i.e. employee job satisfaction negatively correlates with turnover intention. In addition, employees with high job satisfaction are more likely to love the job, to be highly engaged in job and to improve job performance. Therefore satisfied employees would improve their job performance. Combining with hypothesis 3, the author proposes hypotheses:

H5a: Employee job satisfaction mediates the negative influence of group task satisfaction on turnover intention;

H5b: Employee job satisfaction mediates the positive influence of group task satisfaction on job performance.

Little research regarded group cohesion as a moderator variable, and cross-level moderating effect was less. Group cohesion could be divided into two dimensions. One is emotional cohesion, also called social cohesion, which refers to close relationship among group members, and interdependence tendency formed in daily life in which they get along with each other. Another is task cohesion which puts emphasis on coordination and distribution of responsibilities and complementary skills of group members when they are completing group task together [12] . In a group of strong emotional cohesion, satisfied employees would emotionally attach to their group, reducing their turnover intention; in a group of strong task cohesion, group members would closely cooperate, improving job performance of their group. Therefore, the author concludes that:

H6a: Group cohesion moderates the positive effect of employee job satisfaction on turnover intention;

H6b: Group cohesion moderates the negative effect of employee job satisfaction on job performance.

From what has been discussed above, the author puts forward hypothesis model shown in Figure 1.

The author issued formal questionnaires to the employees in the front-line service group and back office group of three five-star hotels and a four-star hotel in Guangzhou and Shenzhen. Before issuing, the author packed corresponding questionnaires into the envelope according to the number of each group and marked the group name on it to avoid packing by mistake. After employees completed, the author sealed the envelope and took them back. A total of 252 questionnaires from 40 groups were collected. Eliminating invalid questionnaires such as incomplete questionnaires, 204 questionnaires from 35 groups were valid, with an 81% effective rate. In the valid study samples, women accounted for 56.1%; employee between 18 and 28 years old accounting for 74.6%;between 29 and 45 years old accounting for 23.9%; 46 years old or above accounting for 1.5%; below high school degree, 43.4%, college degree or above, 56.6%; monthly income below 3000 RMB, 68.2%, between 3000 and 8000, 30.3%, above 8000, 1.5%; respondents who work in current hotel are less than 1 year, 55.6%, between 1 year and 5 years, 39.5%, above 5 years, 4.9%.

The measurements of key variables in present study adopt Likert 5-points scale, 1 for complete disagreement

and 5 for complete agreement. Group task satisfaction (GTS) was measured with group task satisfaction scale developed by Mason and Griffin [13] . The measurement of group cohesion (GC) referred to group identity scale by Henry et al. [14] , Henry et al. argued “group cohesion from the group identity”. Group job characteristics (GJC) was measured with job diversity, autonomy and feedback three dimension, and items design referred to the study of Teas [15] . Job satisfaction (JS) was measured with job satisfaction 5-scale developed by Smith et al. [16] . The measurement of turnover intention (TI) referred to the scale developed by Wayne et al. [17] . The measurement of job performance (JP) referred to “task-contextual performance”, two-dimension model proposed by Borman and Motowidlo [18] .

Firstly, this research has adopted the SPSS17.0 to conduct exploratory factor analysis (EFA) of the three multidimensional variables, which are group task satisfaction, group job characteristic and group cohesion. The sub- scales of group task satisfaction and group job characteristics hold very high loads among their common factors; meanwhile, they have achieved convergent validity and discrimination validity. Among the sub-scales of group cohesion, the item coefficient of GC11 was less than 0.45, it had to be deleted; while GC3, GC5 and GC7 fell over the dimension of planned construct and they couldn’t be explained easily, therefore, they also had to be deleted.

After revising the scale, this research will adopt LISREL8.80 to conducted the confirmatory factor analysis (CFA ) of group task satisfaction, group cohesion and group job characteristics, which furthermore will use indexes of ×2 testing, RMSEA, NNFI, CFI, SRMR to judge the fitting degree of models. As shown in Table 1, the measurement models of group task satisfaction, group cohesion and group job characteristics basically fit the data, demonstrating that the scales have good construct validity. After reversion, all variables “Cronbach” α reached between 0.609 and 0.832, which showed that the reliability of the scales reached acceptable scope.

Among the hierarchical linear models, level 1 (individual level) includes the employee’s job satisfaction, turnover intention, job performance and control variables. Level 2 (group level）includes group job characteristics, group cohesion and group task satisfaction. RWG value, coefficient of ICC (1) and ICC (2) of employees’ score of group job characteristics, group cohesion and group task satisfaction basically reached acceptable aggregation conditions recommended by Klein et al. [19] . Therefore, we aggregated the individual scores of the three variables to level 2 variable values.