In this chapter, the results of all hypotheses are provided with the usage of SPSS software version 23.0 based on 245 valid responses. First section presents descriptive statistics or the general profile of all respondents together with the reliability of each scale. Second section shows the correlation among all variables including Job Demands (Workload, Cognitive Demand, Emotional Demand), Emotional Exhaustion and Psychological Contract Breach. Lastly, hierarchical regression analysis tests all hypotheses and the moderating effect of Psychological Contract Breach.
Descriptive Statistics
In total, there were 245 people working in Vietnam service industry joined this research. The profile of all respondents was presented in Table 4.1. The majority of participants were female (62%), only 37,1% were male and 0.4% belonged to other genders. Most of them are under 35 years old. The detailed age range is as following: under 25 years old (19.6%), 25 to 30 years old (43.7%), 31 to 35 years old (26.9%), and only less than 10% of respondents were over 35 years old. This fact also highlighted the marital status of the data, with nearly 60% of people was still single, and therefore also around 62% participants did not have kids. There were one fourth of the participants had one kid, 11.4% had two kids and only 1.2% had three kids or more. Regarding educational background, over 85% of all respondents had a bachelor’s degree.
Even though all data were collected from working people, they did not work in similar company.
Around 60% of all participants worked in private company, while public and Foreign direct investment (FDI) organizations recorded only 17.6% each of the sample. In accordance with the young age of the participants, more than half of them only had 1 to 3 years job tenure, or even under one year (22%). However, the type of services that their company provided was even more diverse.
Among twelve main kinds, business service and financial services are most popular, with 19.2% and 17.6% respectively.
Table 4.1.
Respondents Profile (Total N= 245)
36
Construction and related engineering 19 7.8 4. Number of
kids
Distribution services 23 9.4
0 152 62 Educational services 19 7.8
1 62 25.3 Transport services 16 6.5
2 28 11.4 Environmental services 11 4.5
From 3 children 3 1.2 Financial services 43 17.6
Recreational, cultural and sporting 7 2.9
5. Education Health related and social services 3 1.2
High school 9 3.7 Tourism and travel related services 5 2
Bachelor 209 85.3 Other services 41 16.6
Master and above 27 11
37
Results of Pearson Correlation Analysis
Pearson Correlation Analysis was conducted to show only means, standard deviation and reliability level, but also the correlation among all variables. In this analysis, three control variables namely age, gender and tenure were also included. The result was shown in detail from Table 4.2.
Regarding reliability, all variables showed a high Cronbach’s Alpha value over 0.7, thus indicating good internal consistency. Specifically, Job Demands recorded 0.89 Cronbach’s alpha value (with Workload – 0.81; Cognitive Demand – 0.73 and Emotional Demand – 0.86). While Emotional Exhaustion also got high reliability with 0.92 Cronbach’s alpha and Psychological Contract Breach with 0.79.
All control variables did not have correlation with either independent or dependent variables, and only age and Psychological Contract Breach (moderator) had a positive correlation.
Meanwhile, Job Demands (and also three subdimensions) were positively and significantly related to Emotional Exhaustion. Psychological Contract Breach also had a significant positive relationship with Emotional Exhaustion (r = .59, p < .01). Furthermore, Psychological Contract Breach was significantly correlated to both Job Demands (r = .86, p < .01) and three dimensions including Workload (r = .32, p < .01), Cognitive Demand (r = .32, p < .01) and Emotional Demand (r = .53, p < .01). However, the result of hypotheses would be tested by using hierarchical regression analysis due to the need to include control variables in the analysis.
38
Table 4.2.
Means, Standard Deviations, Reliability Coefficients, and Correlations among The Variables
Mean S.D 1 2 3 4 5 6 7 8 9
Note. Numbers in parentheses represent Cronbach’s alpha value. *p < .05. **p < .01.
Result of Confirmatory Factor Analysis
After all data were collected, Confirmatory Factor Analysis - one type of structural equation modeling (SEM) was applied to test the fitness of the model or hypothesis to the data. The software that was used in this research is Mplus. First, Chi-square test indicating the degree of similarity among the observed and expected matrices was performed. Then the Comparative Fit Index (CFI), the Tucker Lewis Index (TLI), the Root Mean Square Error of Approximation (RMSEA) and the Standardized Root Mean Square Residual (SRMR) test were all conducted to measure the model fit more comprehensively. The criteria for assessing CFA indices were presented as below.
According to Tabachnick and Fidell (2007), Chi-square index ranging from 2 to 5 is considered acceptable. Meanwhile, CFI and TLI needs to be at least 0.9 to indicate a good model fit (Bentler, 1990; Tucker & Lewis, 1973). According to Browne and Cudeck (1989), the threshold for RMSEA and SRMR are lower than 0.08, indicating a good model fit. In running the CFA test, the original result indicated poor model fit. Therefore, based on modification index suggestion, a number of paths were added to improve the model. All in all, the final result of CFA for this research shown in table 4.3 presented the majority of indeces met the requirement (Chi-square is 2.88 and SRMR is 0.07). For CFI, TLI and RMSEA, the actual indices were close to cut-off point,
39
therefore the model is still considered relatively acceptable. All in all, the validity and fitness of the model in this research was confirmed.
Table 4.3.
Results of Confirmatory Factor Analysis (N = 245)
Index Acceptable Level References CFA Test Results
χ²/df 2 - 5 Tabachnick and Fidell, 2007 2.88
CFI > 0.90 Bentler, 1990 0.87
TLI > 0.90 Tucker and Lewis, 1973 0.85
RMSEA < 0.08 Browne and Cudeck, 1989 0.08
SRMR < 0.08 Browne and Cudeck, 1989 0.07
Note. df= Degree of Freedom; CFI = Comparative Fit Index; TLI= Tucker Lewis Index;
RMSEA= Root Mean Square Error of Approximation; SRMR= Standardized Root Mean Square Residual.
Result of Hierarchical Regression Analysis
In this research, there are three control variables namely age, gender and job tenure. Therefore, hierarchical regression analysis was conducted to take into account the effect of these variables when testing all hypotheses.
Hypothesis 1 proposed that Job Demands had a positive relation with Emotional Exhaustion.
Performing hierarchical regression needs two main steps to test this hypothesis. Firstly, control variables including gender, age and tenure were added in the first model. Secondly, Job Demands – independent variable was entered to examine the relationship with dependent variable – Emotional Exhaustion. From Table 4.4. the result indicated that Job Demands were positively and significantly related to Emotional Exhaustion (β = .58, p < .001). Therefore, hypothesis 1 was supported.
40
Table 4.4.
Result of Hierarchical Regression Analysis of Job Demands to Emotional Exhaustion (N = 245)
Variable β
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001.
To test the relationship of sub dimensions of Job Demands, including Workload, Cognitive Demand and Emotional Demand with Emotional Exhaustion, the same process of hierarchical regression analysis was conducted. First, control variables were inserted into the first model.
Afterwards, the subdimension was entered to test the hypothesis 1a, 1b and 1c. According to Table 4.5, Workload had a significant positive influence on Emotional Exhaustion (β = .43, p < .001).
Likewise, Cognitive Demand (β = .35, p < .001) and Emotional Demand (β = 0.35, p < .001) were also positively related to Emotional Exhaustion, as shown in Table 4.6. and 4.7, respectively.
Therefore, Hypotheses 1a, 1b and 1c were all supported.
Table 4.5.
Result of Hierarchical Regression Analysis of Workload to Emotional Exhaustion (N = 245)
Variable β
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001.
41
Table 4.6.
Result of Hierarchical Regression Analysis of Cognitive Demand to Emotional Exhaustion (N = 245)
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001.
Table 4.7.
Result of Hierarchical Regression Analysis of Emotional Demand to Emotional Exhaustion (N = 245)
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001.
42
Result of Moderating effect of Psychological Contract Breach
Hypothesis 2 concerns the moderating effect of Psychological Contract Breach on the correlation between Job Demands and Emotional Exhaustion, and this hypothesis was also tested using hierarchical regression analysis. Specifically, control variables (age, gender and tenure) were put in model 1, then Job Demands and Psychological Contract Breach were added into model 2 before the interaction term of these two variables followed in model 3. As Table 4.8 presents, when adding the interaction in model 3 the result showed it has significant additional explanation power to the Emotional Exhaustion (∆R² = .04, β = .21, p < .01), which also suggested a significant moderating effect of PCB between Job Demands and Emotional Exhaustion. Therefore, hypothesis 2 was supported.
Table 4.8.
Result of Hierarchical Regression Analysis for PCB as moderator and Job Demands as independent variable (N = 245)
Psychological Contract Breach .36*** .36***
Step 3
Job Demands*Psychological Contract Breach .21***
F 6** 47.64*** 46.39***
R² .07** .50*** .54***
Adj. R² .06** .49*** .53***
∆R² .07** .43*** .04***
∆F 6** 102.49*** 20.56***
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001
The effect of PCB as a moderator on the relationship between each dimension of Job Demands (Workload, Cognitive Demand and Emotional Demand) and Emotional Exhaustion was also tested by following the same analysis process as above. Specifically, the interaction effect of Workload and PCB (hypothesis 2a), Cognitive Demand and PCB (hypothesis 2b) and emotional and PCB (hypothesis 2c) were tested. As Table 4.9, 4.10. and 4.11. show, all three interaction terms had positive and significant relationship with Emotional Exhaustion, respectively [Workload and PCB
43
(β = .21, p < .001); Cognitive Demand and PCB (β = .22, p < .001) and Emotional Demand and PCB (β = .19, p < .01)]. Therefore, hypotheses 2a, 2b and 2c were all supported.
Table 4.9.
Result of Hierarchical Regression Analysis for PCB as moderator and Workload as independent variable (N = 245)
Psychological Contract Breach .47*** .45***
Step 3
Workload*Psychological Contract Breach .21***
F 6** 37.44*** 36.27***
R² .07** .44*** .48***
Adj. R² .06** .43*** .46***
∆R² .07** .37*** .04***
∆F 6** 78.76*** 17.48***
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001
Table 4.10.
Result of Hierarchical Regression Analysis for PCB as moderator and Cognitive Demand as independent variable (N = 245)
Psychological Contract Breach .50*** .47***
Step 3
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001
44
Table 4.11. Result of Hierarchical Regression Analysis for PCB as moderator and Emotional Demand as independent variable (N = 245)
Variable β
1 2 3
Step 1
Age .31*** .13* .08
Gender .08 .08 .07
Tenure -.29*** -.16** -.11
Step 2
Emotional Demand .44*** .43***
Psychological Contract Breach .33*** .33***
Step 3
Emotional Demand*Psychological Contract
Breach .19***
F 6** 50.11*** 47.74***
R² .07** .51** .55***
Adj. R² .06** .50*** .54***
∆R² .07** .44*** .04***
∆F 6** 108.22*** 18.01***
Note. Dependent variable = Emotional Exhaustion; *p < .05. **p < .01. ***p < .001
45
For more detailed presentation of moderating effect, interaction graphs were created for hypotheses 2, 2a, 2b and 2c. Specifically, figure 4.1. stated the interaction between PCB and Job Demands. Job Demands had considerably stronger positive impact on Emotional Exhaustion for employees with higher Psychological Contract Breach. On the other hand, Job Demands had a weaker positive relationship with Emotional Exhaustion when people perceived lower Psychological Contract Breach. Therefore, PCB is a moderator which can strengthen the positive association between Job Demands and Emotional Exhaustion.
Figure 4.1. The moderating effect of PCB on the relationship between Job Demands and Emotional Exhaustion
46
Figure 4.2. to 4.4. demonstrate the interaction effect of sub dimensions of Job Demands including Workload, Cognitive Demand and Emotional Demand, and they also display the same pattern with Figure 4.1. Specifically, from Figure 4.2, for people who had high Workload, if they had high Psychological Contract Breach, they would suffer higher Emotional Exhaustion compared to people who had low Psychological Contract Breach. This indicates that PCB strengthens the effect of Workload on Emotional Exhaustion.
Figure 4.2. The moderating effect of PCB on the relationship between Workload and Emotional Exhaustion
47
Figure 4.3. showed that for people who had high Cognitive Demand, if they had high Psychological Contract Breach, they would suffer higher Emotional Exhaustion compared to people who had low Psychological Contract Breach. This indicates that PCB strengthens the effect of Cognitive Demand on Emotional Exhaustion.
Figure 4.3. The moderating effect of PCB on the relationship between Cognitive Demand and Emotional Exhaustion
48
Figure 4.4. showed that for people who had high Emotional Demand, if they had high Psychological Contract Breach, they would suffer higher Emotional Exhaustion compared to people who had low Psychological Contract Breach. This indicates that PCB strengthens the effect of Emotional Demand on Emotional Exhaustion.
Figure 4.4. The moderating effect of PCB on the relationship between Emotional Demand and Emotional Exhaustion
49
Discussion
The primary objective of this study is to investigate whether Job Demands (with three different dimensions namely Workload, Cognitive Demand and Emotional Demand) have association with Emotional Exhaustion and the role of Psychological Contract Breach in this relationship. In this section, all results were analyzed in more detail, with the summary of hypotheses testing as below.
Table 4.12.
Hypotheses Testing Results Summary
Hypotheses Result
H1 Job Demands are positively related to Emotional Exhaustion. Supported H1a Workload has a positive association with Emotional Exhaustion. Supported H1b Cognitive Demand has a positive association with Emotional
Exhaustion. Supported
H1c Emotional Demand has a positive association with Emotional Exhaustion.
Supported
H2 Psychological Contract Breach strengthens the relationship between job demand and Emotional Exhaustion.
Supported
H2a Psychological Contract Breach strengthens the relationship
between Workload and Emotional Exhaustion. Supported
H2b Psychological Contract Breach strengthens the relationship
between Cognitive Demand and Emotional Exhaustion. Supported H2c Psychological Contract Breach strengthens the relationship
between Emotional Demand and Emotional Exhaustion.
Supported
First hypothesis stated that Job Demands are positively related to Emotional Exhaustion.
Accordingly, Workload, Cognitive Demand and Emotional Demand all showed a positive and significant relationship with emotion exhaustion. That means the higher Job Demands, or specifically in terms of high Workload, Cognitive Demand and Emotional Demand would all lead to higher employee Emotional Exhaustion. This finding was in line with the logic of job demand-resources theory proposed by Bakker et al. (2014), which stated that the higher the Job Demands, the more likely employees will be psychologically affected, leading to depression and using up their energy. Other prior studies also suggested the positive association of Workload with Emotional Exhaustion. For example, Leiter and Maslach (2003) found that too much Workload is the main cause of Emotional Exhaustion. And the association of Emotional Demand and
50
Emotional Exhaustion was consistent with the study of Sutton and Wheatley (2003), suggesting that emotional aspects at work had profound impact on not only Emotional Exhaustion but also Job burnout in general. For Cognitive Demand, the relationship with Emotional Exhaustion was supported by Viotti and Converso (2016) before. The study included Cognitive Demand as one type of Job Demands that resulted in not only Emotional Exhaustion, but also higher intention to leave and job dissatisfaction.
The second hypothesis was confirmed that PCB has moderating effect on the association of Job Demands (including Workload, Cognitive Demand and Emotional Demand) with Emotional Exhaustion. In other words, for employees who suffered from high demands of job, the more they feel the employers did not comply with previous promises, the higher chance of Emotional Exhaustion they may encounter. This result was consistent with Social Exchange Theory by Blau (1964), which implied that if imbalance in benefits happened, people tend to response with negative emotions, including Emotional Exhaustion. The link between PCB and Emotional Exhaustion was also proven by Razzaghian and Ghani (2015). For service industry in Vietnam, a lot of people suffered high Workload, Cognitive Demand and even Emotional Demands in working. When combined with the feeling of unfairness, specifically the contribution and effort to the company was not equal to benefits that employers should provide employees, the balance of social exchange is disrupted, leading to depression, Emotional Exhaustion and even turnover intention.
51