While focusing on the difference between actual feeling and the feeling of expression by actors in dramaturgy (Goffman, 1959), Hochschild (1979, 1983) adjusted the concept of emotional labor into organizational culture and the original definition as “the management of feeling to create a publicly observable facial and bodily display.” After Hochschild’s study, several conceptualizations of emotional labor were proposed (Ashforth & Humphrey, 1993; Brotheridge & Lee, 2003; Glomb & Tews, 2004; Grandey, 2000; Morris & Feldman, 1996; Rafaeli & Sutton, 1987). Emotional labor was an exchange value, a wage for “the act of displaying appropriate emotion” (Ashforth & Humphrey, 1993) following the display rules for the organizational goals (Grandey, 2000).

Ashforth and Humphrey (1993) asserted the appropriateness of display rules rather than feeling rules (Hochschild, 1979, 1983) because rules for emotional labor refer to a set of certain behaviors rather than internal feelings. According to Ekman (1972), facial expressions for specific emotions that a person experiences and particular facial configurations are connected universally. For optimal control of the service quality, emotional labor has to be easily monitored by managers, co-workers, and customers (Rafaeli & Sutton, 1987). Mastracci, Newman, and Guy (2010) classify emotional labor as pretense or authentic display by cognitive skills. In emotive work, employees conform to the display rule and regulate the expression of emotion. For instance, display rule for sales personnel in fashion retail stores is to express positive emotion (happiness, passion) and to suppress negative emotion (anxiety, irritation).

Emotional labor has been conceptualized in two ways. First, job-focused emotional labor is the behavioral response to variations in the duration, frequency, intensity, variety, and emotional dissonance of service interactions (Morris & Feldman, 1996), which means “the level of emotional demands” in an occupation (Brotheridge & Grandey, 2002). However, other research has found that emotional labor does not always involve or lead to emotional dissonance (Brotheridge & Lee, 2003; Morris & Feldman, 1997; Zerbe, 2000). Emotional dissonance is a state of being rather than an objective or a process (Grandey, 2000), so it is difficult to say what connects it to the emotion management process. The relationship between emotional dissonance and emotional labor is more likely to include conditional circumstances (Christensen & Raynor, 2003; Godwyn, 2006; Gutek, 1995). Because of ambiguity and controversy in these debates, we do not consider emotional dissonance as a component of emotional labor in our research model.

Second, employee-focused emotional labor describes “the process of managing emotion and expression” by employees to meet work demands (Brotheridge & Grandey, 2002) while performing through surface acting and deep acting (Ashforth & Humphrey, 1993; Hochschild, 1979, 1983). Groth, Hennig-Thurau, and Walsh (2009) defined surface acting as “faking or amplifying emotions by displaying emotions not actually felt,” and deep acting as “attempting to modify felt emotions so that a genuine emotional display follows.” Surface acting involves employees simply pretending to fulfill their job duty, which is achieved by modulating their reaction to a situation such as faking a smile or suppressing a bad mood (Totterdell & Holman, 2003). Deep acting, in contrast, involves employees trying to align required and true feelings (Hülsheger & Schewe, 2011), which is achieved by modifying their arousal or perception of a situation. Since this study is highly relevant to employees’ perceptions and experiences along with performing emotional labor, we consider surface acting and deep acting to be components of emotional labor in our formulation.

Affectivity defined as “a general tendency to experience a particular mood or to react to objects in a particular way or with certain emotion” (Lazarus, 1991). In emotional labor, affectivity is an important employee characteristic that is associated with emotional labor and directly influences employee mood states on the job (Brotheridge & Grandey, 2002; Brotheridge & Lee, 2003). As the sum of an individual’s affective disposition, the affect-based personality trait explores the relationship with an emotional labor strategy (Cheung & Tang, 2009).

A person with negative affectivity, or neuroticism, may have a predisposition to view the world in negative terms, such as pessimism and aversion, and assign a more negative interpretation to stressful work events (Abraham, 1999). On the contrary, a person with positive affectivity, or extroversion, may have a positive disposition including enthusiasm, involvement, and commitment (Watson, Pennebaker, & Folger, 1987). Gosserand and Diefendorff (2005) found negative affectivity was positively related to surface acting, whereas positive affectivity was positively related to deep acting. In this study, we considered affectivity as one of the antecedent variables in influencing emotional labor.

Job autonomy is a widely studied work characteristic, which reflects freedom, independence, and discretion in work scheduling, decision- making, and work methods (Hackman & Oldham, 1976; Morgeson & Humphrey, 2006). In emotional labor, autonomy is a broad concept that can generate specific emotions and behaviors as an antecedent (Grandey, 2000; Gross, 1998; Morris & Feldman, 1996) or can mediate the relationship between emotional labor and individual well-being (Johnson & Spector, 2007).

Employees with a low level of job autonomy are more likely to follow the organization’s display rule strictly and to experience emotional exhaustion as a consequence of surface acting (Grandey, Fisk, & Steiner, 2005). Meanwhile, employees with high level of job autonomy recognize surface acting as less demanding (Johnson & Spector, 2007) and comfortably choose it as one of their selling methods. In this study, we consider that autonomy may alleviate emotional dissonance and emotional exhaustion for service workers in jobs with a high level of emotional labor (Johanson & Woods, 2008). According to the theoretical framework (Grandey, 2000; Gross, 1998; Morris & Feldman, 1996), we investigate autonomy as one of the variables in influencing emotional labor, rather than as a moderator.

Customer incivility is defined as “low-intensity deviant behavior, perpetrated by someone in a customer or client role, with ambiguous intent to harm an employee, in violation of social norms of mutual respect and courtesy” (Silter et al., 2010). While the previously suggested concept of workplace incivility (Andersson & Pearson, 1999) is investigated in the perspective of employee-to-employee or supervisor-to-employee interaction, customer incivility is a norm newly discussed as employee- to-customer interaction.

The literature found sales- and service-related employees experience incivility more often from customers than co-workers and confront with higher burnout levels (Grandey, Kern, & Frone, 2007). When the employees are faced with customers’ disrespectful treatment at work, they endure the mistreatment and convey courteous service quality (Ben-Zur & Yagil, 2005). Customer incivility is an important point to affect emotional labor in sales and service sector since customer are not always right (Grandey, Dickter, & Sin, 2004). With the rising interest in incivility from customers, we consider it to be an important variable in influencing emotional labor.

In the literature, the effects of psychological outcomes from emotional labor were typically related to negative norms such as stress, depression, cynicism, job dissatisfaction, turnover intention, emotional exhaustion and deviation, drug and alcohol abuse, and absenteeism (Grandey, 2000; Hochschild, 1979, 1983; Morris & Feldman, 1996).

The implementation of emotional labor impairs the “real self,” then raises “self-alienation,” and extends to emotional exhaustion (Hochschild, 1979). Consequently, the severe emotional exhaustion in emotional labor leads to burnout, which includes emotional exhaustion, depersonalization, and diminished personal accomplishment (Maslach & Jackson, 1981). Turnover intention is a willingness to leave the organization as a negative outcome of emotional labor (Tett & Meyer, 1993), which is positively affected by surface acting (Chau et al., 2009). Most empirical studies have examined emotional dissonance as a mediating effect rather than a direct relationship between emotional labor and a negative outcome. We believe that those outcomes will be positively affected by surface acting and negatively affected by deep acting.

Although the negative effects came to the forefront, several scholars have found the positive effects of emotional labor (Ashforth & Humphrey, 1993; Brotheridge & Grandey, 2002; Rafaeli & Sutton, 1987). Based on the “facial feedback hypothesis,” Adelmann (1995) proposed that a positive emotional expression, regardless of actual feeling- enhanced self-efficacy, led to job satisfaction, self-esteem, and then individual and organizational well-being. That is to say, expressing positive emotion leads to feeling the actual positive emotion.

Self-efficacy is the belief that one can successfully accomplish task requirements; it is related to the positive outcome of emotional labor (Ashforth & Humphrey, 1993). Deep acting consolidates the self-efficacy by fulfilling social expectation, and voids embarrassing interpersonal problems by contributing to the formation of latent defense mechanisms (Brotheridge & Grandey, 2002). In clothing retail stores, sales associates at the beginner level are encouraged to perform surface acting intentionally, which protects genuine emotion from incivility and dissonance, and enhances task performance and self-efficacy (Lee & Kim, 2012). We believe that self-efficacy will be facilitated by the positive outcomes of emotional labor.

Caring job include emotive characteristics such as nurturance, empathy, self-sacrifice and kindness have long been regarded as women’s job. Since women are more sensible to emotional events and typically associated with emotive labor, our study selects women sales personnel in clothing store as subjects of research.

Table 1 provides an overview of final sample characteristics. A total of 245 women sales personnel in the clothing industry voluntarily participated in the study. Six inadequate samples were excluded, and a final of 239 samples were included for data analysis. The age group ranged from 20-59 with a relatively evenly distributed sample across the age brackets of 20-29 years (33.9%), 30-39 years (28.9%), and those above 40 years (37.2%), which means women in their 30s stopped working for some reason (e.g., childbirth and childrearing). Marital status was evenly distributed (unmarried = 51.5%, married = 48.5%). Participants had an average of 13.64 years of education, most stopping after a high-school diploma (41.0%). Other information collected included years employed in the service industry and position at the job.

Critical variables for this study included three causes of emotional labor from personal, organizational, and customer perspectives: salesperson affectivity, job autonomy, and customer incivility as antecedent variables. In addition, surface acting and deep acting in emotional labor were considered independent variables. Self-efficacy, job satisfaction, and burnout were explored as dependent variables; their interrelationship was empirically assessed using structural equation modeling.

Emotional labor was measured with the original scale that Choo, Kim, and Jun (2010) based on the study of Brotheridge and Grandey (2002). The scale contains 10 items rated on a 5-point, Likert-type scale (1 = never to 5 = always). The items were composed of two independent, but related factors (Grandey, 2000). Five items measure surface acting (α = .86), which refer to false expressions of emotion, and five items measure deep acting (α = .82), which refer to genuine expressions of emotion. One example of the item is: “I pretend to have emotions that I don’t really have” (see the Appendix). A factor analysis revealed a three-factor structure, in accordance with the two dimensions described above. Since the confidence coefficient of more than .60 is enough to proceed with an exploratory study (Nunnally, 1978), the reliabilities of the subscales measuring surface acting and deep acting were high enough.

Affectivity was measured by an amended version of the Positive and Negative Affect Schedule (PANAS) developed by Watson, Clark, and Tellegen (1988). The scale contains five items, rated on a 5-point, Likert-type scale ranging from 1 = not at all to 5 = extremely. The internal consistency reliability estimate for this scale was α = .87 in the current sample. The question reads: “I feel scared about my life.” Six items measuring job autonomy were adopted from the studies by Morgeson and Humphrey (2006), and measured with a 5-point, Likert-type scale (1 = strongly disagree, 5 = strongly agree). Cronbach’s alpha for this scale was .87. One sample item is: “The job provides me with significant autonomy in making decisions.” Customer incivility was assessed with the original scale developed by Wilson and Holmvall (2013). The scale consisted of 10 items, each measuring perceptions of uncivil behavior from customer with 5-point, Likert-type scale (1 = never, 5 = more than three time per day). Crombach’s alpha for this scale was .88. One sample item is: “Customers continued to complain despite your effort to assist them.”

Self-efficacy was adopted from the study of self-efficiency by Chen, Gully, and Eden (2001). The scale contains five items, rated on a 5-point, Likert-type scale (1 = strongly disagree to 5 = strongly agree). The internal consistency reliability estimate for the scale was α = .77 in the current sample. The sample statement is: “On the whole, I am satisfied with myself.” Turnover intention was extracted from the study by Lee and Kim (2012) based on Muliawan, Green, and Robb (2009). Four items were measured with a 5-point, Likert-type scale (1 = strongly disagree, 5 = strongly agree). Cronbach’s alpha for the scale was .84. One sample item is: “I like my job better than the average worker does.” Nine items for measuring burnout were adopted from the studies of emotional exhaustion and depersonalization by Maslach and Jackson (1981), and measured with a 5-point, Likert-type scale (1 = not at all, 5 = always). Cronbach’s alpha for this scale was .90. The sample statement includes: “I feel emotionally drained from my work.”

The questionnaire was designed to gain insight into women who work in sales service in clothing industry. The self-report questionnaire was written in Korean, and completion took an average of 15 minutes. To avoid geographic bias, we used convenience sampling from sales associate of small business, specialty retail store, department stores, outlet stores, and duty free shops in the Seoul and Gyeonggi area, which represent 45% of total population of Korea. Survey progress was done in cooperation with Korean Information Research during a period of two weeks in May 2014. Out of 300 questionnaires, 245 were returned (82% response rate) and 6 samples with inadequate answers were excluded. The final sample of our study consisted of 239 women.

All research questions were tested using statistical analysis: descriptive statistics, frequency analysis, correlation analysis, reliability analysis, exploratory factor analysis, and confirmatory factor analysis were completed using the Statistical Package for the Social Science (SPSS) version 20 (Sung, 2007) and the Analysis of Moment Structures (AMOS) version 21 (Lee & Lim, 2009).

The Pearson correlations between the research variables are shown in Table 2. Negative affectivity was positively correlated to surface acting (r = .28, p < .01) and negatively correlated to deep acting (r = -.30, p < .01). Job autonomy was negatively correlated to surface acting (r = -.31, p < .01) and positively correlated to deep acting (r = .43, p < .01). Customer incivility was positively correlated to surface acting (r = .19, p < .01) and negatively correlated to deep acting (r = -.26, p < .01). Surface acting was positively correlated to turnover intention (r = .22, p < .01) and burnout (r = .38, p < .01), but not significantly correlated to self-efficacy. Deep acting was positively correlated to self-efficacy (r = .23, p < .01), but not significantly correlated to turnover intention and burnout. The table shows that H1, H2, and H3 were all confirmed in relation to the emotional labor, but H4, H5, and H6 were only partially confirmed. These results provided preliminary support for the hypothesized relationship.

All measures were subjected to confirmatory factor analysis for issues of dimensionality, convergent, and discriminant validity. Since coefficient alpha underestimates the reliability of multidimensional measure, the two-step procedure suggested by Anderson and Gerbing (1988) was examined for the effective use of coefficient alpha.

We started the confirmatory factor analysis of the antecedents (i.e., employee affectivity, job autonomy, and customer incivility). Results showed that the three-factor model was acceptable, χ² (df = 41, N = 239) = 80.749, p < .001, root mean square error of approximation (RMSEA) = .06, root mean residual (RMR) = .04, goodness of fit index (GFI) = .94, Tucker Lewis Index (TLI) = .96, and comparative fit index (CFI) = .97. All items were significantly loaded on their respective variables. Next, the confirmatory factor analysis of the dependents (i.e., surface acting, deep acting, self-efficacy, turnover intention, and burnout) was tested. Results showed that the five-factor model was acceptable, χ² (df = 46, N = 239) = 137.503, p < .001, RMSEA = .091, RMR = .05, GFI =.91, TLI =.89 and CFI =.923. Therefore, measures in this research model captured distinctive constructs.

We examined the structural equation model depicted in Figure 1 using AMOS version 21 (Lee & Lim, 2009). The results demonstrated the following fit statistics for an eight-factor model was acceptable, χ² (df = 83, N = 239) = 224.190, p < .001, RMSEA =. 09, RMR =.09, GFI =.90, AGFI =.84, TLI =.86 and CFI =.91. Although GFI were same to .09 and TLI were slightly below .09, adjusted goodness of fit (AGFI) was above .80, which indicates a good fit. Also, RMSEA and CFI were greater than a threshold value; therefore a high degree of validity as well as reliability for the research model was achieved.

Estimates of the path coefficients for the hypothesized model are presented in Figure 2. An examination of the paths in the research model revealed that hypothesis 1 was partially supported, with negative affectivity being positively related to surface acting (H1a, r = .17, p <.05), but unrelated to deep acting (H1b). Hypothesis 2 was partially supported with job autonomy being unrelated to surface acting (H2a), but positively related to deep acting (H2b, r = .37, p <.001). Perceived customer incivility was positively related to surface acting (H3a, r = .30, p <.001), but unrelated to deep acting (H3b). Self-efficacy was not affected by surface acting (H4a), but positively affected by deep acting (H4b, r = .11, p <.001). Hypothesis 5 was fully supported because turnover intention had positive path to surface acting (H5a, r = .47, p <.001) and negative path to deep acting (H5b, r = .20, p <.05). Hypothesis 6 was partially supported, with burnout being positively affected by surface acting (H6a, r = .72, p <.001), but not affected by deep acting (H6b).

Figure 2. 
Numbers near arrows represent standardized coefficients (*p < .05. **p < .01. ***p < .001.) Dashed lines represent nonsignificant paths.

In this study, we examined the relationships between emotional labor and its antecedents and outcomes for women employees in the clothing service sector. We found that, through surface acting, emotional labor was positively affected by employee affectivity and customer incivility, which positively related to turnover intention and burnout. In addition, emotional labor was positively affected by job autonomy and positively related to self-efficacy through deep acting. This finding reconfirmed that the negative antecedents were expressed through surface acting leading to negative outcomes and the positive antecedents were expressed through deep acting leading to positive results in the existing theory (Ashforth & Humphrey, 1993; Grandey, 2000; Hochschild, 1983; Morris & Feldman. 1996; Rafaeli & Sutton, 1987).

Meanwhile, our finding was that surface acting was not affected by job autonomy and did not correlate to self-efficacy. Also, deep acting was not affected by employee affectivity and customer incivility did not correlate to burnout. With the exception of deep acting, which was significantly related to turnover intention, surface acting did not offset the positive outcome and deep acting did not improve the negative outcome. In other words, the positive antecedents did not always improve surface acting and the negative antecedents did not always offset deep acting; the relationship is still not clear since several studies found no significance or opposite ways of relation among emotional labor concepts (Adelmann, 1995; C. Wong, P. Wong, & Law, 2007). Consequently, differences with the prevailing theory of emotional labor were also confirmed.

The study uses an integrative approach to analyze emotional labor among the antecedents and the outcomes, and newly count influence of emotional labor from customer side, customer incivility. Moreover, the result focuses especially on women employees in clothing service sector since women have been required to be kind and intimate, and have faced strong pressure to work in caring job and hospitality.

With a growing number of people working in the service sector and experiencing emotional labor, it is increasingly important to examine the causes and outcomes of emotional labor, whether they are positive or negative outcomes. This study first investigated the integrated research model of emotional labor and examined it with empirical data analysis. As noted earlier, previous researchers have not examined both the antecedents and outcomes of emotional labor in one model, which could provide a more comprehensive understanding of the mechanism of emotional labor.

This study’s research reveals a more comprehensive understanding of interaction between the antecedents and the outcomes of emotional labor. The research model faithfully followed the framework of the initial conceptual study (Grandey, 2000), and replicated previous research results suggesting that emotional labor was related to self-efficacy (Brotheridge & Grandey, 2002), turnover intention (Chau et al., 2009; Tett & Meyer, 1993). Additionally, the study extended the antecedents to customer perspective, included employee and organization perspective, and examined the research model with a homogenous sample on empirical data analysis. Moreover, emotional dissonance that had been ambiguous in nature as a situational concept is difficult to discuss due to the impact of casual relations with emotional labor, so it is excluded from this research model (Christensen & Raynor, 2003; Godwyn, 2006; Zerbe, 2000). In the implementation of emotional labor, the respondents’ behavioral concepts, such as surface acting and deep acting, were based on the verification of final research model.

The specific target - women sales personnel in the clothing service industry - was a homogenous sample, which was more informative to researchers interested in the dynamics of contextual or organizational factors in fashion and clothing. Participants were asked to provide ratings at work in real time using experience sampling methodology (Beal & Weiss, 2003), which is easier when considering emotion-related constructs. Consequently, based on the more informative data, this study may contribute to identifying problems and suggesting practical solutions for women working in emotional labor.

A possible explanation of our findings is that employees that are more likely to have negative affectivity perform surface acting rather than deep acting as a result of experiencing turnover intention and strong burnout. Thus, corporate welfare programs and cultural support are recommended to offset the negative employee affectivity. Under the organizational culture, recognizing the job autonomy of service workers rather than restricting and reinforcing the display rule gives employees a stronger tendency to perform deep acting with sincerity rather than surface acting. Also, building a good interaction with customers through deep acting may improve employee self-efficacy. However, high job autonomy will not restrain surface acting of emotional labor, which may continue to be a unique selling technique in service sector. Employees who often face customer incivility and rudeness may perform surface acting rather than deep acting to defend themselves, and consequently may experience more severe negative outcomes such as turnover intention and burnout. The organization should clearly define its position vis a vis customers’ excessive insolence and accurately inform employees how to react in alignment with the company’s customer service policy. These findings may apply to advance retail environments in encouraging service workers who face emotional labor and to conduct in-depth surveys of business efficiency and confidence.

As with most research, this study has its limitations. First, although we measured eight factors in the integrated research model, other variables may account for the relationship between causes and outcomes with emotional labor and could be controlled in future research. Second, we focused the outcome on employee perspectives, which is a crucial element to change in emotional labor, but future research could examine the relationship between customer satisfaction and organizational well-being. Third, our study targets women workers in the clothing service sector, which may not be generalized for all types of jobs or other demographic factors such as sex and age. Even though most service workers in fashion industry are women, a comparative analysis of women and men with the variables would be crucial for future research since the relationship between the variables might be influenced by demographic factor. Fourth, although we have integrated antecedents and outcomes of emotional labor in one model, we are still not clear when emotional labor has a weaker or stronger effect on an individual’s psychological status. In other words, the moderating mechanism of the effects in emotional labor was not examined in our study. Last, the research model was based on a previous theory from initial studies of emotional labor, but future researchers could progress to a conceptual study, which could encompass the partial empirical study and the recent findings regarding emotional labor.