This chapter aims to present the findings based on the proposed hypotheses in this study.
Here, the results of confirmatory factor analysis (CFA), linear regression, hierarchical regression, structural equation modeling (SEM), and multi-group SEM to test hypotheses were explained and discussed.
Confirmatory Factor Analysis (CFA)
Before testing each hypothesis, the confirmatory factor analysis (CFA) was performed to assess the measurement model and to provide an indication of measurement equivalence between the two countries. CFA was performed using AMOS 22.0. The study was conformed of a total of 289 respondents. The number of observed items for emotional labor was 8, customer mistreatment was 18, and burnout had 22. The Root Mean Square of Approximation (RMSEA), the ratio of chi-square to degrees of freedom (x2/df), Comparative Fit Index (CFI), Normed Fit Index (NFI), Incremental Fit Index (IFI), Tucker-Lewis Index (TLI), and the Goodness of Fitness Index (GFI) were used to examine the model fit.
Table 4.1.
Results of Confirmatory Factor Analysis
x2 df X2/ df RMSEA CFI NFI IFI TLI GFI Emotional Labor 58.79 19 3.09 .08 .97 .95 .96 .95 .95
Burnout 865.84 208 4.16 .11 .85 .81 .85 .83 .76
Customer Mist. 762.17 135 5.65 .13 .88 .85 .88 .86 .74 One-Factor 6024.80 1081 5.57 .13 .56 .51 .56 .54 .39 Two-Factors 2356.46 405 5.82 .13 .66 .62 .66 .64 .55 Three-Factors 2775.87 1076 2.58 .07 .85 .77 .85 .64 .69
45
A good model fit should have a chi-square value (x2) of p < .05, x2/df between 2-5; CFI >
.90; NFI > .90; IFI > .90; TLI > .90; GFI > .90; and RMSEA < .08 (Hooper, Coughlan, &
Mullen, 2008). The results of this study are shown in Table 4.1. Analyzing the 8-item model for emotional labor (Figure 4.1.), RMSEA is 0.08 which indicates a good fit, CFI, IFI, NFI, TLI are above the accepted criteria. For burnout model (Figure 4.2.), the model did not show a good fit, since RMSEA (.11) was above the accepted criteria. CFI was 0.85; IFI 0.85; NFI 0.81; TLI 0.83; and GFI 0.76; all of them are below the accepted criteria. The values of customer mistreatment as shown on Table 4.1. also, did not show good fit indices (Figure 4.3.).
The one-factor model (Figure 4.4.) included all the items of emotional labor, customer mistreatment, and burnout. This model did not show a good fit, since RMSEA (.13) was above the accepted good fit criteria, CFI was 0.56; NFI 0.51; IFI was 0.56; TLI was 0.54;
GFI was 0.39. All of these values are below the accepted criteria. The two-factors models (Figure 4.5.) as described on Table 4.1. does not show good fit criteria. The three-factor model proposed in this study (Figure 4.6.) was also analyzed. RMSEA was .07, which indicates the model has a good fit. CFI was .85, and IFI was 0.85, which both are slightly under the accepted criteria. NFI was .77, TLI 0.64, and GFI 0.69. All of them lower of the acceptable range. However, the three-factor model reported to have the best fit compared to the one-factor model, and two-factor model.
46 Figure 4.1. Measurement model of emotional labor
47 Figure 4.2. Measurement model of Burnout
48
Figure 4.3. Measurement model of customer mistreatment
49 Figure 4.4. Structural one-factor model
50 Figure 4.5. Structural two-factor model
51 Figure 4.6. Structural three-factor model
52
Hierarchical Regression
This analysis was used to test the positive relationship of surface acting on burnout (hypotheses H1a, H1b, and H1c), and the negative relationship of deep acting on burnout (hypotheses H2a, H2b, and H2c). Hierarchical regression was used to examine the moderating effect of customer mistreatment between emotional labor and burnout (hypotheses H5a, and H5b). The three steps recommended by Baron and Kenny (1986) were adopted in the study. Age, organizational level, and social desirability, were entered as the control variables in the first step. The main effects of surface acting, deep acting and burnout were entered as the second step, and the interaction was entered as the third step.
The regression results demonstrated that surface acting is positively and significantly related to the three facets of burnout as shown on Table 4.2. Surface acting has a positive and significant relationship with emotional exhaustion as shown in Model 2 (β = .67, p < .001), depersonalization as shown in Model 5 (β = .52, p < .001), and personal accomplishment as shown in Model 8 (β = .19, p < .01) supporting hypothesis H1a, H1b, and H1c accordingly.
Based on the results, hypothesis 1 predicting the positive and significant effect of surface acting on burnout was supported.
Hypothesis 2 stated that deep acting is negatively related to burnout. Surprisingly, the regression results did not support this hypothesis. Deep acting did not show negative significant effect on emotional exhaustion, but a positive significant effect, as shown on Table 4.2. in Model 3 (β = .42, p < .001). The effect of deep acting on depersonalization showed a significant positive effect, and not negative as predicted in this study, as shown in Model 6 (β = .38, p < .001) therefore, hypothesis 2b is not supported. The effect of deep acting on persona accomplishment did not show any negative significant effect as described in Model 9 (β = .10, p > .05). Therefore, hypothesis 2 predicting the negative and significant
53
effect of deep acting on burnout was not supported, rejecting hypotheses H2a, H2b, and H2c accordingly.
Hypothesis 5 stated that customer mistreatment has a positive moderating effect on burnout. Results demonstrated that customer mistreatment did not moderate the relationship between surface acting and emotional exhaustion as shown in Table 4.3., Model 3 (β = .00, p
> .05), depersonalization as shown in Model 6. (β = -.01, p > .05), and personal accomplishment, as shown in Model 9 (β = -.05, p < .05). Therefore, hypothesis H5a, H5b, and H5c were not supported since the relationship between surface acting and burnout is not positively moderated by customer mistreatment. Hypothesis 6 predicted that customer mistreatment has a negative moderating effect on the relationship between deep acting on burnout. As shown on Table 4.4., Model 3, deep acting has no moderating effect on emotional exhaustion (β = .02, p > .05), and depersonalization as shown in Model 6 (β = .06, p > 0.05), rejecting hypotheses H6a, and H6b. However, negative moderating effect of customer mistreatment was found between deep acting and personal accomplishment (β = -.11, p < .05) as shown in Model 9, supporting hypothesis H6c.
54 Table 4.2.
Results of Regression Analyses for Surface and Deep Acting and Burnout
Emotional Exhaustion Depersonalization Personal Accomplishment
Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
Step 1: Controls
Age -.02* -.01 -.02* -.01 -.00 -.01 -.01 -0.01 -.01
Organizational Level -.20 -.09 -.14 -.13 -.05 -.07 -.16* -.13 -.15**
Social Desirability 1.84*** 1.34** 1.28*** 2.43*** 2.05*** 1.92*** 1.01** .87* .88**
Step 2: Main Effect
Surface Acting .67*** .52*** .19**
Deep Acting .42*** .38*** .10
Step 3: Interaction
R2 .10 .29 .20 .09 .20 .41 .06 .09 .07
Adj. R2 .09 .28 .18 .08 .19 .16 .05 .07 .05
F 10.38*** 29.36*** 17.29*** 9.36*** 17.91*** 14.30*** 5.80*** 6.56*** 5.12***
∆R2 .10 .19 .10 .09 .11 .08 .06 .03 .01
∆F 10.38*** 77.91*** 34.38*** 9.36*** 39.78*** 26.62*** 5.80*** 8.37** 2.95**
Note. *p < .05, **p < .01, ***p < .001
55 Table 4.3.
Results for Regression Analysis for Moderating Effect of Customer Mistreatment
Emotional Exhaustion Depersonalization Personal Accomplishment
Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
Step 1: Controls
Age -.02** .00 -.00 -.01 .00 .01 -.01 -.01 -.00
Organizational Level -.12 -.00 -.00 -.11 -.01 -.01 -.13 -.09 -.09
Social Desirability 1.27*** .33 .33 1.53*** .66** .66** 1.06*** .76* .73*
Step 2: Main Effect
Surface Acting .37*** .37*** .22*** .22*** .26*** .26
CSR .47*** .47*** .53*** .54*** .02 .04
Step 3: Interaction
SAxCSR .00 -.01 -.05
R2 .11 .57 .57 .10 .49 .49 .07 .14 .14
Adj. R2 .10 .56 .57 .09 .49 .49 .06 .12 .12
F 11.63*** 74.01*** 61.53*** 10.79*** 55.85*** 46.41*** 7.43*** 9.19*** 7.78***
∆R2 .11 .46 .00 .10 .40 .00 .07 .07 .00
∆F 11.63*** 149.61*** .00 10.79*** 110.96*** .09 7.43*** 11.05*** .77
Note. SA= Surface Acting; CSR= Customer Mistreatment.
*p < .05, **p < .01, ***p < .001
56 Table 4.4.
Results for Regression Analysis for Moderating Effect on Burnout
Emotional Exhaustion Depersonalization Personal Accomplishment
Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
Step 1: Controls
Age -.02** -.01 -.01 -.01 .00 .00 -.01 -.01 -.01
Organizational Level -.12 -.02 -.02 -.11 -.02 -.02 -.13 -.10 -.10
Social Desirability 1.27*** .38 .39 1.53 .67 .68*** 1.06*** .79* .71*
Step 2: Main Effect
Deep Acting .17*** .18*** .14** .14** .13* .11
CSR .58*** .57*** .59*** .58*** .10 .13*
Step 3: Interaction
DAxCSR .02 .02 -.11*
R2 .11 .50 .50 .10 .48 .48 .07 .11 .12
Adj. R2 .10 .50 .49 .09 .47 .47 .06 .09 .10
F 11.63*** 56.44*** 46.91*** 10.78*** 52.03*** 43.29*** 7.43*** 6.75*** 6.36***
∆R2 .11 .39 .00 .10 .37 .00 .07 .03 .01
∆F 11.63*** 110.29*** .12 10.79*** 102.39*** .26 7.43*** 5.39** 4.04**
Note. DA= Deep Acting, CSR= Customer Mistreatment.
*p < .05, **p < .01, ***p < .001
57
Structure Equation Modeling (SEM)
Structure Equation Modeling (SEM) and multi-group SEM were conducted to test the hypotheses as well to compare the proposed model across El Salvador and Taiwan. Two models were constructed, one that does not include the interaction terms (main model to assess the main effects and compare them across Taiwan and El Salvador) and another one with the interaction terms just to assess the presence of interactions as well as compare them between Taiwan and El Salvador.
Note. EE= Emotional Exhaustion; DE= Depersonalization; PA= Personal Accomplishment;
SA= Surface Acting; DA= Deep Acting.
The purpose of performing SEM is to determine whether the strength of the relationship between specific behaviors is the same for different groups. The unconstrained model was compared to a model where all variables were constrained and Chi-square test of difference
58
was used to assess whether there was a significant difference between the two models. The model was controlled for age, social desirability and organizational level.
Multi-group analysis showed that there was no significant difference between El Salvador and Taiwan for the main effects of surface acting and deep acting on burnout as shown on Table 4.5. There was no significant difference between Salvadorean and Taiwanese employees on the positive relationship between surface acting and emotional exhaustion (Δ𝜒2 (1) = .44, p > 0.05), depersonalization (Δ𝜒2 (1) = 1.3, p > 0.05), and personal accomplishment (Δ𝜒2 (1) = .70, p > 0.05) as shown on Table 4.5. rejecting hypotheses 3a, 3b, and 3c accordingly. No significant difference was found between Salvadorean and Taiwanese employees in the effect of deep acting on burnout. There was no significant different between El Salvador and Taiwan in the main effect of deep acting on emotional exhaustion (Δ𝜒2 (1) = .00, p > 0.05), depersonalization (Δ𝜒2 (1) = .68, p > 0.05), and personal accomplishment (Δ𝜒2 (1) = .03, p > 0.05) as shown on Table 4.5. Therefore, hypothesis 4a, 4b, and 4c were not supported accordingly. Figure 4.1. shows the multi-group analysis diagram for El Salvador. Figure 4.2. shows the multi-group analysis diagram for Taiwan.
Hypothesis 7a predicting customer mistreatment towards Salvadorean service workers having a positive moderating effect of surface acting on burnout than Taiwanese service workers was partially supported. The formal Chi-square test results did not show a significant difference in the moderating effect of customer mistreatment on the relation between surface acting on emotional exhaustion (Δ𝜒2 (1) = .12, p > 0.72), and personal accomplishment (Δ𝜒2 (1) = 1.5, p > 0.05). However, the multi-group moderation test revealed that customer mistreatment towards Salvadorean service employees had a more positive moderating effect of surface acting on depersonalization compared to Taiwanese service employees (Δ𝜒2 (1) = 4.36, p < .05) which partially supports hypothesis 7a as shown on Table 4.6. Hypothesis 7b
59
was not supported since none of the components of burnout had a significant difference in the moderating effect of customer mistreatment on the relationship between deep acting and emotional exhaustion (Δ𝜒2 (1) = .41, p < .05), depersonalization (Δ𝜒2 (1) = .05, p < .05), and personal accomplishment (Δ𝜒2 (1) = 2.26, p < .05) as shown on Table 4.6. Figure 4.3. shows the multi-group analysis diagram for El Salvador and Figure 4.4 shows the multi-group analysis diagram for Taiwan with customer mistreatment as moderator.
Table 4.6.
Note. EE= Emotional Exhaustion; DE= Depersonalization; PA= Personal Accomplishment;
SA= Surface Acting; DA= Deep Acting.
60
Figure 4.7. Multi-group analysis for El Salvador (Moderator: Nationality)
Figure 4.8. Multi-group analysis for Taiwan (Moderator: Nationality)
61
Figure 4.9. Multi-group analysis for El Salvador (Moderator: Customer Mistreatment)
Figure 4.10. Multi-group analysis for Taiwan (Moderator: Customer Mistreatment)
62 Table 4.7.
Hypothesis Testing Results Summary
Hypothesis Result
H1 Surface Acting is positively related to burnout. Supported H1a Surface acting is positively related to emotional exhaustion. Supported H1b Surface acting is positively related to depersonalization. Supported H1c Surface acting is positively related to personal
accomplishment.
Supported
H2 Deep acting is negatively related to burnout. Not Supported H2a Deep acting is negatively related to emotional exhaustion. Not Supported H2b Deep acting is negatively related to depersonalization. Not Supported H2c Deep acting is negatively related to personal
accomplishment.
Not Supported
H3 The positive relationship between surface acting and burnout is stronger for Salvadorean service workers than Taiwanese.
Not Supported
H3a The positive relationship between surface acting and emotional exhaustion is stronger for Salvadorean service workers than Taiwanese.
Not Supported
H3b The positive relationship between surface acting and depersonalization is stronger for Salvadorean service workers than Taiwanese.
Not Supported
H3c The positive relationship between surface acting and
personal accomplishment is stronger for Salvadorean service workers than Taiwanese.
Not Supported
63 Table 4.7. (continued)
Hypothesis Result
H4 The negative relationship between deep acting and burnout is weaker for Salvadorean service workers than Taiwanese.
Not Supported
H4a The negative relationship between deep acting and burnout is weaker for Salvadorean service workers than Taiwanese.
Not Supported
H4b The negative relationship between deep acting and
depersonalization is weaker for Salvadorean service workers than Taiwanese.
Not Supported
H4c The negative relationship between deep acting and lack of personal accomplishment is weaker for Salvadorean service workers than Taiwanese
Not Supported
H5 Customer mistreatment has a positive moderating effect of surface acting on burnout
Not Supported
H5a Customer mistreatment has a positive moderating effect of surface acting on emotional exhaustion.
Not Supported
H5b Customer mistreatment has a positive moderating effect of surface acting on depersonalization.
Not Supported
H5c Customer mistreatment has a positive moderating effect of surface acting on personal accomplishment.
Not Supported
H6 Customer mistreatment has a negative moderating effect of deep acting on burnout
Not Supported
H6a Customer mistreatment has a negative moderating effect of deep acting on emotional exhaustion.
Not Supported
64
H6b Customer mistreatment has a negative moderating effect of deep acting on depersonalization.
Not Supported
H6c Customer mistreatment has a negative moderating effect of deep acting on personal accomplishment.
Supported
H7a Customer mistreatment towards Salvadorean service workers has a more positive moderating effect of surface acting on burnout than Taiwanese service workers.
Not Supported
H7b Customer mistreatment towards Taiwanese service workers has a more negative moderating effect of deep acting on burnout than Salvadorean service workers.
Not Supported
65