n recent times, due to the rapid emergence of customer service industries, researchers have expanded their examination of the performance of emotion work in organizational settings (England, 2005). Emotion work is viewed as a form of emotional regulation process, where employees display appropriate emotional reactions in response to several workplace demands. This involves dealing with clients, customers, subordinates or co-workers with an optimistic approach and adopting a flexible mental or emotional state towards the target population. The term "emotion work" was first coined by the sociologist Arlie Hochschild (1983) as the "management of feeling to create a publicity observable facial and bodily display." Following her work, several attempts have been made by eminent scholars to conceptualize different aspects of emotion work and its effect on human service professionals. Situational demands in organizations act as stressors which affect these professionals both physically and mentally, leading to lifestyle diseases such as asthma, cancer, cirrhosis, chronic obstructive pulmonary disease, diabetes, heart disease, obesity etc.

Although many researchers have explored the nature of emotion work with varied perspectives (Morris & Feldman, 1996;Ashforth & Humphrey, 1993), recently health psychologists have provided new insights into the management of emotions and emotional expressions to conform with organizational display rules and requirements (Diefendorff, Croyle & Gosserand 2005;Mann 2004). Research on management of emotions as a part of the work role is based on the assumption that organizations affect people's thoughts, feelings and actions through a set of display rules that serve as standards for the appropriate display of emotions (Grandey, 2000). It is increasingly recognized that emotional expressions at workplace is strongly influenced by organizational norms and rules (Rafaeli & Sutton, 1987), that are communicated to employees through both formal means, such as selection, training, evaluation, incentive systems, and informal means, such as social influence and pressures (Schein, 1985). To comply with the situational demands, employees act by suppressing their true emotions (Ashforth & Humphrey, 1995). Emotion work entails following these display rules regardless of one's felt emotions which may involve displaying positive and negative emotions, sensitivity towards client's feelings, emotional dissonance and exerting control over interactions with clients (Zapf, Vogt, Seifert, Mertini, & Isic, 1999). The study of emotion work has been considered an essential aspect for ensuring effective interpersonal interactions and group functioning (Adelmann, 1995;Arvey et al., 1998). It has also become the focus of great deal of research in psychology and other health related disciplines in recent years (e.g., Barger, 2006;Saluan, 2009). Emotion work at organizations requires faking positive emotions through surface acting (Cheung & Tang, 2009) and suppression of negative ones (Tan, Foo, Chong & Ng, 2003). Employees experiencing greater emotion work may suffer from emotional exhaustion, depersonalization (Botma, 2009) and physical health problems such as headaches, insomnia etc. (Schaubroeck & Jones, 2000). Several studies have also found that such people are more likely to use surface acting emotion work strategy (Karim & Weisz, 2010), which impairs their cognitive health leading to a sense of stagnation towards work life.

As Hochschild (1983) noted, employees often bottle up feelings of frustration and resentment, resulting in emotional outburst. Parker and Wall (1998) have argued that intense emotion work can have detrimental health effects by weakening the body's immune system. Several studies have found that emotional dissonance and surface acting result in stress and emotional exhaustion (Kruml & Geddes, 2000;Brotheridge & Lee, 2002;Holman, Chissick & Totterdell, 2002). Pugliesi (1999) reported the existence of well-defined display rules results in increased stress leading to psychological distress. In this respect, Grandey (2000) and Schaubroeck & Jones (2000) found that the inhibition of expression of negative emotion might cause a range of physical illnesses, including serious illnesses such as hypertension and cancer. Further, Brill (2000) argued that 'emotional dishonesty' leads to antisocial behaviours, including violence. Other researchers have found that, when individuals suppress their emotions they possess less personal will, lower performance, imprecise memories for those situations (Twenge & Baumeister, 2002).

Recent studies have pointed to a causal sequence between emotion work, emotional dissonance and the negative job consequence, emotional exhaustion (Lewig & Dollard, 2003;Dijk & Brown, 2006). A study by found that healthcare workers with greater emotional dissonance reported greater emotional exhaustion. They were also found to experience frequent negative emotions. Evidence suggests that emotional dissonance is directly associated with emotional exhaustion and is commonly experienced by customer service representatives serving retail sector, healthcare and call centre industries (Gupta & Mishra, 2011;Karl & Peluchette, 2006).

Literature indicates that individual factors may also play an important role in helping employees perform interpersonal tasks more flexibly (Gross & John, 1997) and with less strain (Tews & Glomb, 2003), thus, influencing the degree of emotion work. However, emotion work has also been linked to some negative effects on employees, such as emotional exhaustion and psychological wellbeing (Brotheridge & Grandey, 2002;Zapf, Seifert, Schmutte, Mertini & Holz, 2001). Based on the notion that individual dispositional factors have some stability over time and are useful in predicting individual behavior across situations (Staw & Ross, 1985), researchers attempted to demonstrate how these individual characteristics relate to emotion work. Specifically, prior studies found that emotionality in employees engaged in service jobs including customer interactions such as those belonging to tourism and hotel industries tend to have a positive impact on emotion work (Tronvoll, 2011;Hopp, Rohrmann, Zapf & Hodapp, 2010). They have shown that individuals high on emotionality are prone to experience a diverse array of negative mood states (e.g., fear, anger, distress). On the other hand, studies have confirmed that employees with positive affectivity are less likely to be affected by emotion work (Austin, Dore & Donovan, 2008;Bono & Vey, 2007). Evidence shows that emotional intelligence also plays an important role in determining one's satisfaction with social relationships (Lopes, Salovey & Straus, 2003). Research reported that emotional intelligence correlates negatively with surface acting and positively with deep acting (Ghalandari, Ghorbani & Jogh, 2012). Furthermore, it is evident that when confronted with emotion work, nurses experienced lower levels of burnout and somatic complaints and this effect was found to be mediated by the choice of emotion work strategies (Mikolajczak, Menil & Luminet, 2007).

Studies highlighted that customer service employees with high emotional intelligence experienced greater psychological wellbeing (Gardner & Stough, 2003) and low levels of psychological distress (Karim & Weisz, 2011). Evidence suggests that negative emotions was inversely related with psychological wellbeing (Karim & Weisz, 2010;Botma, 2009). Specifically, prior studies found that emotionally intelligent employees engaged in customer service sector tend to experience more positive emotions (Austin, Dore & Donovan, 2008;Cheung &Tang, 2009 andPrati &Karriker, 2010). Also, a study by Adil & Kamal (2012) found that negative affectivity mediated the relationship between display rule physical and mental health. Researchers found that emotional dissonance was positively related to emotional exhaustion (Arora & Falade, 2011;Gupta & Mishra, 2011). A study by Adil & Kamal (2012) demonstrated hiding negative emotions, negative affectivity, and emotional intelligence as positive predictors and positive affectivity as a negative predictor of emotional exhaustion. Several studies also suggested that surface acting and emotional exhaustion were positively related (Ghalandari, Mortazavi, Abbasi & Jogh, 2012). Higher levels of emotional exhaustion were also found to be strongly associated with faking and suppression of negative emotions . demands of hiding negative emotions and emotional exhaustion.

In customer service work, many psychological factors dominantly mediate an employee's performance of emotion work which ends up having a long term impact on the individual's health. This is typically a product of the array of challenges that service providers typically encounter which impact their physical and psychological health. Their attempts to meet the situational demands by exhibiting appropriate display rules impair their cognitive health in the form of several illnesses such as depression, hypertension, coronary heart disease and alcoholism. Such demands results in heightened reactions at psychological and physiological levels. The inability to deal with stressors arises from lack of competence to deal with negative emotions and manage one's own feelings in difficult circumstances.

Therefore, an understanding of emotion work and its mediating role on employees' health can provide a framework for developing more effective interventions to mitigate the negative effects of emotion work.

To determine the mediating role of emotion work between (a) emotionality and emotional exhaustion, and (b) emotional intelligence and psychological wellbeing.

IV.

The participants of the study include 400 employees across the country from different geographical locations employed in customer service industries. Table 1 presents their demographic information.

For this study a cross -sectional descriptive design was used. The participants belonged to a variety of service occupations, such as, banking and insurance (6.8%), sales (9%), marketing (10.8%), software engineer (4.5%), legal advisor (3.5%), financial advisor (6.3%), customer service (7%), psychologist (8%), HR (30.3%), consultants (10.5%) and technical (3.5%). The sample comprised 64.8% males and 35.3% females. Their age ranged from 21 to 65 years with a mean age of 35.02 years (SD =10.15).The sampling frame was narrowed to the criteria of subjects who have face-toface or voice-to-voice interactions with clients/ customers and they are of either entry-level employees or middle-level managers.

Each participant read and signed an informed consent document. The questionnaires were individually administered to 400 employees across the country from different geographical locations (Delhi, Haryana, Mumbai, Bangalore and Hyderabad) working in service industries on full time basis. The service industries comprise professions such as banking and insurance, sales, marketing, software engineer, legal advisor, financial advisor, customer service, psychologists, HR, consultants and technical. The participants were instructed to give responses which best describes their attitudes towards the statements given in the questionnaire. After collecting the data, scoring was completed and the responses were analyzed.

In order to identify the mediating effect of emotion work among customer service employees, following measures were administered.

Frankfurt Emotion work scale: Frankfurt Emotional Work Scale-E (FEWS-E: 2006; version 1.4-4.4) was developed by Zapf et.al. (1999) to measure the psychological processes necessary to regulate organizationally desired emotions as part of one's job. This comprises five subscales, namely, positive emotions, negative emotions, sensitivity requirements, emotional dissonance and interaction control. All subscales were scored independently of one another, with higher the score indicating greater emotion work. The scale was found to be internally consistent with cronbach alpha ranging from . 51 All the above scales were selected on the basis of the literature review and psychometric properties for all the psychological measures adopted. These scales were also standardised on Indian population and have proven reliability with cronbach alpha ranging from 0.40 to 0.86.

For the current study statistical analysis was done using SPSS student version 15. The analysis was done on five levels. Level I comprised of data cleaning. This was done by doing outlier analysis using box whisker diagrams (box plot analysis). Skewness and kurtosis were also carried out to see the normality of the data. Level II consists of descriptive statistics and reliability analysis. Means, standard deviations and Cronbach alpha were computed. Level III comprised relational analysis. Pearson correlation was used to assess the strength of the relationship between the predictors and the dependent variable. Level IV consists of prediction analysis. Hierarchical regression was used to evaluate the relationship between a set of independent variables and the dependent variable, controlling for or taking into account the impact of a different set of independent variables on the dependent variable. Level V comprises mediation analysis. Causal step approach (Baron & Kenny, 1986) and SPSS macro (PROCESS) (Hayes, 2013) were used to identify variables mediating the relationship between the independent and dependent variables.

V.

The objectives of this study were to examine the mediating role of emotion work between (a) emotionality and emotional exhaustion, and (b) emotional intelligence and psychological wellbeing. The survey data obtained on a sample of 400 employees were analyzed descriptively.

Psychological Wellbeing Scale (Ryff, 1995): This 18 item scale measures the nature of wellness of the sample. This comprises six distinct components, i.e., self acceptance, personal growth, purpose in life, positive relations with others, environmental mastery and autonomy. The higher the score indicates the greater positive psychological functioning. The response format used is a six-point likert scale ranging from 'completely disagree' to 'completely agree'. The cronbach alpha reliability coefficients were low to modest, ranging from .33 to .56. The convergent validity of the scale with its 120 items measure ranged from .70 to .89. Oldenburg Burnout Inventory (OLBI) (Demerouti, Bakker, Vardakou & Kantas, 2002): This scale was designed to assess the cognitive and physical components of exhaustion. It measures two core dimensions of burnout, namely, exhaustion and disengagement. The exhaustion subscale has been used in this study. The items refer to general feelings of emptiness, overtaxing from work, a strong need for rest, and a state of physical exhaustion. The response format used is a four point likert scale ranging from strongly agree (1) to strongly disagree (4). The higher the score indicates the greater exhaustion. The scale was found to have acceptable internal consistency reliability estimate ranging from .74 to .87 and test retest reliability estimate of .50. Correlations of the scale with Maslach Burnout Inventory-General Survey (MBI-GS) provided evidence for high convergent and discriminant validity. Table 2 shows the mean, standard deviation, cronbach alpha, skewness and kurtosis value of all the study variables. The mean and SD value of positive emotions, negative emotions, sensitivity requirements, emotional dissonance and interaction control are 33.0 (6.3), 18.8 (4.9), 13.5 (2.5), 14.7(3.3) and 12.5(2.7) respectively. Among the other study variables, the highest mean value is of psychological wellbeing (M = 81.9, SD 11.1) followed by emotional intelligence (M = 52.8, SD 7.2), sociability (M=28.5, SD 7.8), emotional exhaustion (M =17, SD 3.4), activity (M =14.7, SD 2.8) and emotionality (M =13.1, SD 3.0). Normality of the collected data is established through skewness and kurtosis. It was found that the value of skewness for positive emotions was ( -0.53), negative emotions (0.75), sensitivity requirements (-0.48), emotional dissonance (-0.11) and interaction control (0.03), emotionality (-0.26), activity (-0.35), sociability (0.29), emotional intelligence (0.13), emotional exhaustion (-0.14) and psychological wellbeing (-0.47). The value of kurtosis for positive emotions was (-0.02), negative emotions (1.07), sensitivity requirements (0.39), emotional dissonance (0.20) and interaction control (-0.10), emotionality (-0.26), activity (0.46), sociability (-0.20), emotional intelligence (-0.64), emotional exhaustion (0.28) and psychological wellbeing (-0.10). These values indicate that the study variables are reasonably close to normal as its skewness and kurtosis have values between -1.0 and +1.0. 3 reports the zero order correlations among the study variables. The data indicated a significant positive correlation between positive emotions and sociability (r = 0.23, p<.01) followed by emotional intelligence (r = 0.19, p<.01), psychological wellbeing (r = 0.18, p<.01) and significant negative correlation with emotional exhaustion (r = -0.16, p<.01). A significant negative association was also observed between negative emotions and emotional intelligence (r = -0.30, p<.01), followed by psychological wellbeing (r = -0.29, p<.01) and positive correlation with emotionality (r = 0.25, p<.01) followed by emotional exhaustion (r = 0.17, p<.01). The data also indicated a significant positive correlation between sensitivity requirements and emotional intelligence (r = 0.21, p<.01), followed by sociability (r = 0.16, p<.01), psychological wellbeing (r = 0.16, p<.01) and significant negative relationship with emotional exhaustion (r = -0.10, p<.05). A significant positive correlation was also observed between emotional dissonance and emotionality (r = 0.23, p<.01) followed by emotional exhaustion (r = 0.20, p<.01) and negative association with emotional intelligence (r= -0.15, p<.01) followed by psychological wellbeing (r = -0.13, p<.01). A significant positive relationship was also observed between interaction control and sociability (r = 0.13, p<.01) followed by emotional intelligence (r = 0.13, p<.01), psychological wellbeing (r = 0.11, p<.05) and negative association with emotional exhaustion (r = -0.12, p<.05).

Table 4 : Hierarchical Regression Analysis Predicting Emotional Dissonance from Emotionality (N=400)

Before analysis of a possible mediating relationship can commence several conditions have to be met. First, the total effect of the independent variable on the dependent must be significant. Secondly, the independent variable must significantly affect the mediator, and finally the mediator must have a significant direct effect on the dependent variable (Baron & Kenny 1986). To determine if the relationships between the independent, dependent and mediator variables were significant, two hierarchical regression analyses were conducted. The regression analyses were used to determine the significance of the relationships between the variables as well as to generate the relevant statistics for use in further significance testing of mediation. The results of these analyses are presented in Tables 4 and 5. The first regression analysis (Table 4, step 2) demonstrated that emotionality (? = .23; p < .01) explained significant variance in emotional dissonance beyond that explained by the control variables.

The second regression analysis ( In order to test for mediation, multiple hierarchical regression analyses were conducted. In table 6 step1, control variables were entered into the equation, which accounted for eight percent of the variance in emotional exhaustion (R 2 =.08, p<.01). Age demonstrated a significant direct effect on emotional exhaustion. In step 2, emotionality was entered to calculate the size of its direct effect on emotional exhaustion and was found to demonstrate a significant direct effect on emotional exhaustion (? = .43; p <.01). Thus, accounting for total of 25% of variance in emotional exhaustion (?R 2 = 0.18, p<.01). Step 3 in Table 6 represents the analysis of the proposed mediator's effect on the dependent variable in isolation from the independent variable. Emotional dissonance was entered into step 4 of the regression analysis. It explained an additional 19% of variance in emotional exhaustion (?R 2 = 0.19, p<.01). Here, a total of 47% of variance in emotional exhaustion was explained with emotionality and emotional dissonance both demonstrating significant direct effects (? = .41; p <.01; ? = .09; p <.05 respectively). In order to determine if a significant mediating effect exists, the second and fourth steps in the regression analysis were compared. A change in the size of the direct effect of emotionality on emotional exhaustion was observed. If a full mediating role existed, the contribution of emotionality in step 4 would not remain significant when emotional dissonance was included. In step 2, the direct effect of emotionality on emotional exhaustion was significant (? = .43; p <.01). With emotional dissonance included in step 4, the size of the direct effect of emotionality on emotional exhaustion was reduced, but remained

The objective of the study is to determine the mediating role of emotional dissonance between (a) emotionality and emotional exhaustion, and (b) the mediating role of negative emotions between emotional intelligence and psychological wellbeing.

The results noted a significant direct relationship between emotionality and emotional exhaustion. That is, respondents who reported higher levels of negative emotions such as fear, anger, and distress, experienced greater emotional exhaustion. This is consistent with the study on 142 health care workers, where, emotional exhaustion was found to be predominant among employees who expressed frequent negative emotions . The regression analysis showed that the effect size of emotionality on emotional exhaustion reduced from 0.43 to 0.41 with both values remaining significant (p<.01), when the mediator (emotional dissonance) is entered into the hierarchical regression model. This result indicated a partial mediating effect of emotional dissonance between emotionality and emotional exhaustion, as supported by Lewig & Dollard (2003) and Dijk & Brown (2006). Numerous studies have also In table 7 step1, control variables were entered into the equation. Here, the value of R 2 of control variables is .03 which indicates that control variables accounted for three percent of the variance in psychological wellbeing. Age demonstrated a significant direct effect on psychological wellbeing (? = 0.17, p<.01). In step 2, emotional intelligence was entered to calculate the size of its direct effect on psychological wellbeing and was found to demonstrate a significant direct effect on psychological wellbeing (? = .67; p <.01). Thus, accounting for total of 46% of variance in psychological wellbeing (?R 2 = 0.43, p<.01). Step 3 in Table 7 represents the analysis of the proposed mediator's effect on the dependent variable in isolation from the independent variable. Negative emotions was entered into step 4 of the regression analysis. It explained an additional 44% of variance in psychological wellbeing (?R 2 = 0.44, p<.01). Here, a total of 97% of variance in psychological wellbeing was explained with emotional intelligence and negative emotions both demonstrating significant direct effects (? = .64; p <.01; ? = -.09; p <.05 respectively). In order to determine if a significant mediating effect exists, the second and fourth steps in the regression analysis were compared. A change in the size of the direct effect of emotional intelligence on psychological wellbeing was observed. If a full mediating role existed, the contribution of emotional intelligence in step 4 would not remain significant when negative emotions was included. In step 2, the direct effect of emotional intelligence on psychological wellbeing was significant (? = .67; p <.01). With negative emotions included in step 4, the size of the direct effect of emotional intelligence on psychological wellbeing was reduced, but remained significant (? = .64; p <.01) indicating a partial mediating effect.

Bootstrapping method is preferred over Baron and Kenny's (1986) casual step approach because of two main reasons. First, MacKinnon et al. (2002) suggest that the test of mediation recommended by Baron and Kenny (1986) may suffer from low statistical power, and given that statistical methods used to determine intervening relationship are generally inherently complex, it is therefore difficult to establish clear mediating relationships. A recent study showed that the Baron and Kenny causal-steps approach required approximately 21,000 subjects for adequate ability to detect an effect when the effect sizes of the a and b paths were of small strength and all of the relation of X to Y was mediated (Fritz & MacKinnon, 2007). Second, causal step approach does not consider the estimate of the indirect effect. That is, it ignores the central question: Is the indirect effect different from zero? (Preacher & Hayes, 2008). In other words, it does not provide a numerical value of the strength of the mediated effect. Finally, the test requires that there be a significant overall relation between X and Y for mediation to exist.

As can be seen in the table 8, indirect effect of the relationship between emotionality and emotional exhaustion through emotional dissonance is 0.01 (SE = .01), CI 0.95 = 0.001 and 0.022. The indirect effect of the relationship between emotional intelligence and psychological wellbeing through negative emotions is 0.04 (SE = .02), Sobel z = 0.04, p < .05, CI 0.95 = .01 and .09. As indicated in table 8, the 95 percent confidence intervals for both paths does not include zero, therefore emotional dissonance mediated the relationship between emotionality and emotional exhaustion. Also, negative emotions mediated the relationship between emotional intelligence and psychological wellbeing.

shown that cabin attendants, healthcare workers and call centre employees with high emotional dissonance were more likely to experience emotional exhaustion and express feelings of anger and frustration (Karl & Peluchette (2006; Heuven & Bakker (2003) ;Zapf et.al. (2001)). It has been observed that in order to meet the situational demands, employees with high emotionality tend to exert increased emotive effort when asked to express personality incongruent emotions, i.e., portraying a warmth and friendly attitude towards clients. As a result, their personalities were observed to be incongruent with the job characteristics of the service occupations considered in the sample. Such characteristics lead to dissatisfaction, emotional fatigue and hopelessness among employees.

Furthermore, the result shows that emotional intelligence is directly related with psychological wellbeing. In its support, prior researchers suggest that customer service employees with high emotional intelligence tend to experience greater psychological wellbeing (Gardner & Stough, 2003). This was further supported by Karim & Weisz (2011). The regression analysis showed that the effect size of emotional intelligence on psychological wellbeing reduced from 0.67 to 0.64 with both values remaining significant (p<.01), when the mediator (negative emotions) is entered into the hierarchical regression model. Thus, a partial mediating effect of negative emotions was found between emotional intelligence and psychological wellbeing. This is also consistent with Adil & Kamal (2012). Various scholars have emphasized that emotionally intelligent employees engaged in customer service roles, such as, nursing, social work, sales and teaching tend to display greater degree of positive than negative emotions (Austin, Dore & Donovan (2008); Cheung & Tang (2009); Mikolajczak, Menil & Luminet(2007)). Evidence also suggests public sector customer service workers engaging in emotion work report more physical complaints and poor mental health (Karim & Weisz (2010) and Schaubroeck & Jones (2000)). This may be because emotionally intelligent employees were more inclined to understand people, be empathetic to their circumstances, and regulate emotions appropriately in order to comply with the organizational display rules. Their ability to internalise others feelings enables them to avoid psychological distress and attain a sense of personal accomplishment and wellbeing. Thus, the research findings shows that emotional dissonance and negative emotions emerged as mediators of emotion work, which tend to have adverse impact on employees' physical and psychological health.

The findings of this study are subjected to several limitations. First, the research design for the present study was cross section. The limitations inherent in such designs have been well documented (Neuman, 2003), including the difficulty of supporting proposed causal relationships. Secondly, all measures were self report based measures which might have caused socially desirable responses and affected the results of this study. Thirdly, the age group of the participants employed in the study ranges from 21 to 65 years, which is very large. The study can be conducted for different age groups like adolescents, younger adults, middle adults, older adults. Future researchers might adopt quasi-experiments or longitudinal designs and refer to different sources such as supervisors and coworkers for eliminating the bias, especially for emotion work measure. Although this study's results support the mediating role of emotion work between (a) emotionality and emotional exhaustion, and (b) emotional intelligence and psychological wellbeing, further exploration incorporating other mediating and moderating variables of emotion work need to continue in attempts to clarify the links between its determinants and the consequences.

To conclude, for organizations where the performance of emotion work is an integral component of service delivery, the present findings suggest that training programs targeting the performance of emotion work need to consider different strategies for employees to cope when emotions felt are in conflict with those required to be displayed. These approaches should aim to reduce the negative effects of emotion work.