# ANALYSING EMOTIONAL LABOR IN THE SERVICE INDUSTRIES: CONSUMER AND BUSINESS PERSPECTIVES

EDITED BY : Jungkun Park, Weon Sang Yoo and Ki-Joon Back PUBLISHED IN : Frontiers in Psychology

### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-259-6 DOI 10.3389/978-2-88963-259-6

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

Frontiers in Psychology 1 December 2019 | Emotional Labor in Service Industries

# ANALYSING EMOTIONAL LABOR IN THE SERVICE INDUSTRIES: CONSUMER AND BUSINESS PERSPECTIVES

Topic Editors: Jungkun Park, Hanyang University, South Korea Weon Sang Yoo, Korea University, South Korea Ki-Joon Back, University of Houston, United States

Citation: Park, J., Yoo, W. S., Back, K.-J., eds. (2019). Analysing Emotional Labor in the Service Industries: Consumer and Business Perspectives. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-259-6

# Table of Contents

*04 Editorial: Analysing Emotional Labor in the Service Industries: Consumer and Business Perspectives*

Weon Sang Yoo, Ki-Joon Back and Jungkun Park


Cecile K. Cho and Theresa S. Cho


Sang-Lin Han, Hyeon-Sook Shim and Won Jun Choi

*52 Research on How Emotional Expressions of Emotional Labor Workers and Perception of Customer Feedbacks Affect Turnover Intentions: Emphasis on Moderating Effects of Emotional Intelligence*

Young Hee Lee, Suk Hyung Bryan Lee and Jong Yong Chung


Myoung-Soung Lee, Sang-Lin Han, Suji Hong and Hyowon Hyun


Shenghua Huang, Hongbiao Yin and Lifang Tang

# Editorial: Analysing Emotional Labor in the Service Industries: Consumer and Business Perspectives

Weon Sang Yoo<sup>1</sup> , Ki-Joon Back <sup>2</sup> and Jungkun Park <sup>3</sup> \*

*<sup>1</sup> Korea University Business School, Seoul, South Korea, <sup>2</sup> Conrad N. Hilton College of Hotel and Restaurant Management, University of Houston, Houston, TX, United States, <sup>3</sup> School of Business, Hanyang University, Seoul, South Korea*

Keywords: editorial, emotion, emotional labor, emotional labor strategies, service industry (SI)

**Editorial on the Research Topic**

### **Analysing Emotional Labor in the Service Industries: Consumer and Business Perspectives**

This Research Topic covers the dynamics of emotional labor and its related outcomes and antecedents in various settings. Recent consumer and service management research increasingly focuses on the role of emotions and emotional labor in service delivery and, at the same time, employees' emotional status. Frontline employees in various settings are expected to display certain emotions and suppress others in their daily interactions not only with customers but also with their managers or supervisors in order to comply with their job requirements and organizational expectations. As a result, consumers are also coping with their level of emotional acceptance when purchasing goods or receiving services, such as comfort or discomfort. Thus, this topic can be approached from both within the business and from consumer perspectives, or beyond.

There are 10 manuscripts in this Research Topic. First, despite the growing body of research on the emotional labor of service providers, little has been known about the social consequences of emotional labor by exploring double roles of emotional laborer as a consumer. Drawing on emotional dissonance theory, Park et al., investigate the relationship between the felt emotional dissonance and prosocial behavior (e.g., donation to a charity) with four experiments suggesting higher emotional dissonance serially influences perceived lack of control, emotional exhaustion, lowered sympathy for others' feeling, and subsequently lower willingness to help others.

Three manuscripts focus on the emotional relationship between consumers and service providers. Jeong et al., investigate a conceptual model articulating the nature of customer expectations and satisfaction over services with several emotional factors. Five propositions about consumer emotional service expectations as a primary antecedent toward confirmation, perceived quality, and satisfaction are provided. As moderators, two dimensions of consumer detection of emotional labor (i.e., detecting deep acting and surface acting) are imposed on each of the relationships. Evidence demonstrates the roles of emotional service expectation in service confirmation and satisfaction. The moderating effects of consumer detection of employees' emotional strategies are limited to the relationship between emotional service expectation and confirmation in the hotel/restaurant industry. Zhang et al., examine how employee surface acting relates to their sabotage to customers through the mediating role of emotional exhaustion. Also, they explore the moderating roles of coworker exchange and leader-member exchange using the conservation of resources theory and social exchange theory as conceptual frameworks, and find that coworker exchange buffers the positive effect of surface acting on emotional exhaustion. In addition, the study stresses the effect of surface acting on employee harmful behaviors, the potential underlying mechanism, and boundary conditions to mitigate the negative consequences. Lee et al., explore the relationship between

### Edited and reviewed by:

*Richard Eleftherios Boyatzis, Case Western Reserve University, United States*

> \*Correspondence: *Jungkun Park viroid2007@gmail.com*

### Specialty section:

*This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology*

Received: *30 August 2019* Accepted: *24 September 2019* Published: *09 October 2019*

### Citation:

*Yoo WS, Back K-J and Park J (2019) Editorial: Analysing Emotional Labor in the Service Industries: Consumer and Business Perspectives. Front. Psychol. 10:2290. doi: 10.3389/fpsyg.2019.02290* employees' expression of emotional labor and perception of customer feedbacks touching how the perception of customer feedback affects emotional exhaustion in order to understand how emotional exhaustion affects job satisfaction and turnover intentions. The result suggests that employees with better understanding of their own emotions who experience a more significant negative effect when there is a discrepancy between what they feel and how they should act.

Within the organizational level, there are six articles. Based on social identity theory, Oh et al., examines the relationship between corporate social responsibility (CSR) perception and emotional labor strategies, and the effects of the interaction between CSR perception and moral identification on emotional labor strategies indicating main effect of CSR perception on emotional labor strategies. The results find CSR perception of employees is positively related to deep acting revealing employees' views on their organization's social responsiveness and morality affect their emotional labor strategies. Also, affective organizational commitment and moral identifications mediate the relationship between CSR perception and surface acting but not between CSR perception and deep acting. Lee et al., focus bond and fit toward emotional labor within organization. The manuscript examines the influences of relationship bonds on emotional labor through person-organization fit (P-O fit) and the moderating effects of collectivism between P-O fit and emotional labor in the financial industry identifying financial, social, and structural bonds enhanced P-O fit that improves deep acting. This study not only suggests the empirical evidence identifying the process of relationship bonds influencing emotional labor but also expands the scope of study by examining moderating roles of collectivism in cultural psychology aspect.

Two manuscripts focus on knowledge-based service providers. Huang et al., investigate school teachers' emotional labor process and the consequential outcomes for their wellbeing examining the roles of two antecedents, namely, teachers' perceptions of display rules and self-monitoring tendencies. The results show that self-monitoring generally have stronger, though maladaptive, effects than display rule perceptions on individuals' use of emotional labor strategies and well-being. Both selfmonitoring and display rule perceptions are positively related to two emotional labor strategies concluding self-monitoring may be less beneficial than previously thought. Knowledge-based service workers' display rule perceptions and deep acting may not necessarily be harmful to their well-being but reflect their role identification and commitment. Han et al., investigate the effects of level of emotional labor and its impact of emotional labor on organizational trust and organizational commitment with college administrative staffs with 3 factors (i.e., job stress, intimacy, and professionalism) as determinants of emotional labor.

Cho and Cho examine the internal state of employees' emotion how people anticipate negative emotion when faced with an uncertain outcome and try to manage their expectation. While extant research streams remain equivocal on whether managing expectation always succeeds, this research examines situations in which setting a low expectation can have an adverse emotional impact and ways to alleviate this negative emotional consequence using goal setting and false-feedback paradigm. Choi et al., explore rather negative consequences of emotional problem in workplace taking a holistic approach associating the concept of bullying with firm-level performance as well as stakeholders' responses in the market analyzing whether and how market investors react to the news of corporate harassment by top officials of publicly listed firms in Korea. The results find significantly negative stock price reactions to news of corporate bullying and harassment.

We wish to express our great appreciation to the authors of this special issue for their enthusiasm and scholarly achievements. Although the individual manuscripts reflect considerable variations in focus and perspectives, we genuinely believe that there is a tremendous cohesive message for a better understanding of the filed of emotional labor and consider applying various theories to further developing effective employee retention strategies and enhancing the well-being of all stakeholders in the service delivery process.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# ACKNOWLEDGMENTS

Guest editors would like to thank all the authors and reviewers who agreed to participate in this special issue with their original contributions.

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yoo, Back and Park. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Surface Acting, Emotional Exhaustion, and Employee Sabotage to Customers: Moderating Roles of Quality of Social Exchanges

Hui Zhang<sup>1</sup> \*, Zhiqing E. Zhou<sup>2</sup> , Yan Zhan<sup>3</sup> , Chengbin Liu<sup>1</sup> and Li Zhang<sup>4</sup>

<sup>1</sup> School of Sociology, Huazhong University of Science and Technology, Wuhan, China, <sup>2</sup> Department of Psychology, Baruch College (CUNY), New York City, NY, United States, <sup>3</sup> Department of Nursing, Taihe Hospital, Hubei University of Medicine, Shiyan, China, <sup>4</sup> Cardiovascular Medicine Department Unit 3, Taihe Hospital, Hubei University of Medicine, Shiyan, China

Using the conservation of resources theory and social exchange theory as our conceptual frameworks, the current study examined how employee surface acting relates to their sabotage to customers through the mediating role of emotional exhaustion and explored the moderating roles of coworker exchange (CWX) and leader-member exchange (LMX). We collected two-wave time-lagged data from 540 clinical nurses and found that emotional exhaustion mediated the positive relationship between surface acting and employee sabotage to customers. In addition, we found that CWX buffered the positive effect of surface acting on emotional exhaustion, while LMX buffered the positive effect of emotional exhaustion on employee sabotage to customers, such that the effects were weaker when CWX and LMX were higher, respectively. These findings shed light on the effect of surface acting on employee harmful behaviors, the potential underlying mechanism, and boundary conditions to mitigate the negative consequences of surface acting.

### Edited by:

Jungkun Park, Hanyang University, South Korea

### Reviewed by:

Jiseon Ahn, University of Houston, United States Bryan Lee, Chungwoon University, South Korea

> \*Correspondence: Hui Zhang zhanghui\_0927@hust.edu.cn

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 24 August 2018 Accepted: 23 October 2018 Published: 14 November 2018

### Citation:

Zhang H, Zhou ZE, Zhan Y, Liu C and Zhang L (2018) Surface Acting, Emotional Exhaustion, and Employee Sabotage to Customers: Moderating Roles of Quality of Social Exchanges. Front. Psychol. 9:2197. doi: 10.3389/fpsyg.2018.02197 Keywords: surface acting, emotional exhaustion, sabotage to customers, coworker exchange, leader-member exchange

# INTRODUCTION

Since the service industry accounts for above 60% of world GDP and the economy (The World Factbook, 2017), increasing research has focused on frontline service employees' behaviors, attitudes, and feelings (Grandey, 2008; Mayer et al., 2009; Dong et al., 2015). During service delivery, however, employees have to conform to organizational expectations and goals to suppress negative emotions and display positive emotions, which is characterized as emotional labor (Ashforth and Humphrey, 1993; Brotheridge and Grandey, 2002). Emotional labor contains two different displaying rules: surface acting and deep acting. Surface acting emphasizes changing outward emotional display rather than altering the inner true feelings (Abraham, 1998; Grandey, 2000), whereas deep acting highlights regulating the inner feelings to meet requirements of the work (Grandey, 2000; Hülsheger and Schewe, 2011).

The current study examines the effect of surface acting on employee sabotage to customers and investigates potential underlying mechanisms and boundary conditions. We decided to only focus on surface acting for the following reasons. First, while many studies on emotional labor

have investigated the outcomes of surface acting and deep acting in one study (e.g., Wagner et al., 2014; Deng et al., 2016), a large number of studies have only focused on surface acting (e.g., Prati et al., 2009; Shanock et al., 2013; Wang and Groth, 2014; Wagner et al., 2014; Krannitz et al., 2015). In addition, existing research on emotional labor suggests that deep acting has been linked with both positive and negative outcomes (Hülsheger and Schewe, 2011), while the literature consistently indicate that surface acting is negatively related to employee health, attitudes, performance, and well-being (Hülsheger and Schewe, 2011; Mesmer-Magnus et al., 2012), including increased emotional exhaustion (Grandey et al., 2012; Wagner et al., 2014; Yagil and Medlerliraz, 2017), higher negative affect (Judge et al., 2009), more strains (Hülsheger et al., 2010), lower job satisfaction (Giardini and Frese, 2006), lower work engagement (Bechtoldt et al., 2011), more work-to-family conflict (Wagner et al., 2014), higher turnover intentions (Becker et al., 2017), lower organizational commitment (Walsh et al., 2016), and lower task performance (Schmeichel et al., 2006; Goldberg and Grandey, 2007). Our study aims to build on this particular literature and expand the negative consequences of surface acting to employees sabotage to customers.

Second, recent literature indicates that surface acting may have an effect on employee harmful behaviors, such as interpersonal harmful behavior toward coworkers (Deng et al., 2016) and counterproductive behaviors at work (Bechtoldt et al., 2007). However, little is known about the potential effect of surface acting on employee harmful behaviors toward customers during the service interaction. As service employees have two roles at work, with one role being the member of the organization and one role providing service to customers (Côté et al., 2013), surface acting may not only affect their negative behaviors toward the organization and people working in the organization, but also their behaviors toward customers. Given that the accumulation of employee sabotage to customers may seriously decrease customer satisfaction (Schneider et al., 2005; Wang et al., 2011) and result in financial and reputational loss in service organizations (Harris and Ogbonna, 2002; Anderson et al., 2004), understanding whether and how surface acting predicts employee sabotage to customers will have potential implications to further understand how to reduce employee sabotage to customers.

Taken together, given the detrimental effects of surface acting and the focus on employee sabotage to customers as the outcome, in the current study we will examine the effect of employee surface acting on their sabotage to customers, and addresses why and how this might happen. We aim to make three contributions in the process. First, we contribute to the emotional labor literature by identifying employee sabotage to customers as an important negative consequence of surface acting. This illustrates that surface acting may have an impact on employee negative behaviors beyond coworkers and extend to customers during the interaction with customers, highlighting the importance of surface acting in the service industry.

Second, we draw on the conservation of resources theory (COR; Hobfoll, 1989) to theorize emotional exhaustion as the potential mechanism to understand why surface acting might lead to employee sabotage to customers. Surface acting can deplete employee resources and lead to emotional exhaustion (Judge et al., 2009; Grandey et al., 2012), and resource depletion can make employees less able to inhibit the impulse to engage in harmful behaviors under stress (Stucke and Baumeister, 2010), such as sabotage to customers. By identifying emotional exhaustion as the mediator, we will have a better understanding of the process from surface acting to sabotage to customers for theoretical implications, and potential ways to mitigate this process for practical implications.

Third, we examined social exchanges as moderators to buffer the negative effects of surface acting by supplying resources employees need. Previous studies have suggested that social exchanges in the workplace are important resources (Cohen and Wills, 1985; Ng and Sorensen, 2008; McCarthy et al., 2016; Schneider et al., 2017). Thus, we integrate the COR theory (Hobfoll, 1989) and Social Exchange Theory (Blau, 1964), and predict that coworker exchange (CWX) and leader-member exchange (LMX) may potentially buffer the effect of surface acting on employee sabotage to customers. It also emphasizes the importance of providing social recourses at work (McCarthy et al., 2016) to address interpersonal issues.

# HYPOTHESES DEVELOPMENT

# Surface Acting and Sabotage to Customers

Surface acting has been considered a unique feature of service employees as part of their job (Cho et al., 2013). It refers to the requirement of front-line service employees to suppress the expression of their true feelings and express the false feelings during the interaction with customers (Gross, 1998; Grandey, 2000). As surface acting takes place during the interaction between employees and customers, it may not only have a harmful effect on employee behaviors toward coworkers (Deng et al., 2016) but also have an impact directly on customers during the interaction with customers, such as sabotage to customers. Sabotage to customers is common in the service industry as one type of interpersonal harming behavior (Crino, 1994; Skarlicki et al., 2008). It severely violates the service rule of providing friendly and professional high-quality service (e.g., Solomon et al., 1985; Skarlicki et al., 2008). Thus, understanding whether and why surface acting promotes employees engage in such behaviors will have important practical implications.

We propose that surface acting would positively predict employee sabotage to customers for the following reasons. First, displaying surface acting consumes employees' resources to inhabit their true inner feelings, leading to resources loss or ego depletion (Judge et al., 2009; Wagner et al., 2014; Deng et al., 2016). Previous studies on ego depletion suggest that individuals will be more aggressive (Stucke and Baumeister, 2010), act unethically (Welsh et al., 2014), and behave in a more antisocial manner (Friehe and Schildberg-Hörisch, 2017) under resources depletion. Thus, when employees displaying surface acting and are depleted, they may have fewer resources to obey the organizational rules, leading to rule-breaking behaviors such as sabotage to customers.

Second, as surface acting requires employees to suppress their inner negative feelings, they are less likely to regulate their negative emotions and often are more likely to experience emotional dissonance (Deng et al., 2016). While employees might engage in organization deviance and harmful behaviors toward coworkers as a result (Bechtoldt et al., 2007; Deng et al., 2016), their actions might not stop there. Because of the frequent direct interaction between customers and front-line service employees, employees are also likely to engage in aggressive behaviors toward customers (Wang et al., 2011; Groth and Grandey, 2012). For example, when employees experience negative feelings from customers, they may allow themselves to put customers on hold for a longer period (Grandey, 2003; Wang et al., 2011; Groth and Grandey, 2012). These behaviors might be engaged to compensate for the suppressed negative feelings during the frequent interaction with customers.

In line with this, previous studies have found that the surface acting is positively associated with employee harmful interpersonal behaviors toward organization (Bechtoldt et al., 2007) and coworkers (Deng et al., 2016), which provide indirect empirical evidence of the possible impact of surface acting on employee sabotage to customers. In addition, previous research has found that customer-related experiences such as customer mistreatment (Wang et al., 2011; Groth and Grandey, 2012) and customer injustice (Skarlicki et al., 2016) can positively predict employee sabotage to customers. Given that employees tend to engage in surface acting as the initial reaction to customer-related negative experiences (Grandey et al., 2012), they are likely to engage in sabotage to customers when surface acting depletes too many resources that they are unable to inhibit their negative behaviors toward customers. Thus, we expect that surface acting has the potential positive effect on employee sabotage to customers and hypothesize that:

Hypothesis 1: Surface acting will positively predict employee sabotage to customers.

# Mediating Effect of Emotional Exhaustion

Emotional exhaustion is defined as a resource depletion state when a person no longer can make a big physical or mental effort (Gaines and Jermier, 1983). We employ the COR theory (Hobfoll, 1989) to explain the potential mechanism of the association between surface acting and employee sabotage to customers through emotional exhaustion. The COR theory suggests that individuals tend to protect and build resources important to them (e.g., energy and time; Hobfoll, 1989). Front-line employees often experience emotional exhaustion because they have to face excessive customer demands, resulting in resources depletion (Wright and Cropanzano, 1998). When the resources are depleted, they may experience a higher level of emotional exhaustion (Maslach et al., 2001). The COR theory also suggests that resources deplete over time, which is a long-term process (Hobfoll, 1989), and that it is faster for resources to deplete in coping with work demands than to supplement themselves (Freedy and Hobfoll, 1994). Surface acting requires expressubg inconsistent emotions with their inner feelings, and employees will have to devote more efforts to inhibiting impulse (Ashforth and Humphrey, 1993), consuming their resources and leading to emotional exhaustion. Previous research (e.g., Judge et al., 2009; Grandey et al., 2012; Wagner et al., 2014; Li et al., 2017; Kong and Jeon, 2018) has provided sufficient empirical evidence for this link, and it is expected the same in the current study.

The resource perspective (Hobfoll, 1989) can also provide an explanation of the relationship between emotional exhaustion and employee sabotage to customers. It has been suggested that depletion of control resources is an important reason for employees to engage in more deviant behaviors (Marcus and Schuler, 2004; Thau and Mitchell, 2010) and aggressive acts (Stucke and Baumeister, 2010), and thus emotional exhaustion, as the state of resource depletion, is likely to also predict employee sabotage to customers. When employees deplete their resources due to surface acting and experience emotional exhaustion, they are more likely to engage in inappropriate or undesirable behaviors because employees will have fewer resources to regulate these behaviors (e.g., Muraven et al., 1998). In addition, when employees are depleted with resources and experience emotional exhaustion, they also tend to ignore organizational rules or/and moral standards, result in rule-breaking and normative behaviors (Thau and Mitchell, 2010), such as sabotage to customers. Given the frequent direct interaction between customers and front-line service employees, when employees are experiencing high emotional exhaustion due to surface acting, they are less likely to inhibit impulsive behaviors and customers may become the available victims.

Although theoretically reasonable, the relationship between emotional exhaustion and employee sabotage to customers has not been empirically tested in previous studies. However, recent studies found that emotional exhaustion is positively associated with interpersonal harming behavior, such as interpersonal harming to coworkers and aggression in organization (e.g., Thau and Mitchell, 2010; Christian and Ellis, 2011; Wang et al., 2011; Deng et al., 2016), which suggest that employees with a higher level of emotional exhaustion may also engage in interpersonal harming behavior toward customers such as sabotage to customers.

Hypothesis 2: Emotional exhaustion will positively predict employee sabotage to customers.

According to the COR theory (Hobfoll, 1989) as well as the theoretical argument and empirical evidence presented above, we believe surface acting can consume employees' resources and result in emotional exhaustion, which in turn will lead to employee sabotage to customers. Thus we argue that emotional exhaustion links surface acting and employee sabotage to customers and predict that:

Hypothesis 3: Employees' emotional exhaustion mediates the relationship between surface acting and sabotage to customers.

# Moderating Effects of Social Exchanges

The COR theory (Hobfoll, 1989) also highlights the conditions to protect individuals form resources losses and to cope with resources losses. It suggests that when facing potential or actual

resources losses, individuals often tend to gain available resources to supply and protect resources (Hobfoll, 1989). Social support is one of the important ways in this process (Hobfoll, 1989, 2001). First, social support can provide resources to broaden individuals' resources pool (Hobfoll, 1989) and help people alleviate the negative effect caused by resource-depleting experiences (e.g., surface acting) through a few channels, including promoting skills to cope with demands (Dunahoo et al., 1998) and decreasing work demands (Ray and Miller, 1994) and emotional dissonance (Monica et al., 2016). Second, social support also promotes the replenishment of resources pool and formation of the gain spirals after resource loss (Hobfoll, 1989) and thus buffers the negative effects of resource loss state such as emotional exhaustion (Hakanen et al., 2008).

According to the social exchange theory, there are two types of important social support as resources supplement in the workplace: leader-member exchange (LMX, support from leaders; Graen and Uhl-Bien, 1995) and coworker exchange (CWX, support from coworkers; Sherony and Green, 2002). Employees can gain social support resources form the interaction with coworkers and leaders in the work context to cope with work demands (Ng and Sorensen, 2008; Monica et al., 2016) and resource depletion (McCarthy et al., 2016). Therefore, we propose that employees receiving more social support from coworkers and leaders through high levels of CWX and LMX, respectively, will gain and supply resources and mitigate the negative effects of surface acting and emotional exhaustion caused by surface acting.

# The Moderating Role of CWX

Although CWX and LMX are both important resources for employees, their roles might be different. Compared to the relationship between leader and employees, the relationship between coworkers is more equal and less performance monitoring (Diefendorff and Greguras, 2009), and focusing more on trust and social reciprocity (Cole et al., 2002). The social exchange theory (Blau, 1964) suggests that more authentic, intimate and personal social exchange is based on reciprocity and social resources, while more economic and transactional exchanges are based on materialistic and instrument resources. Thus, employees may gain more social and emotional resources from the social exchange with coworkers (CWX), but gain more instrument resources from the economic exchange with leaders (LMX; McCarthy et al., 2016).

As coworkers own equal power and interpersonal relationship with employees, employees are more likely to share emotional events at work with coworkers (Hadley, 2014). The more frequent interaction between employees and coworkers than leaders provides social support with behavioral and emotional resources (Chiaburu and Harrison, 2008) and employees can receive more social resources to cope work demands and strains (Thoits, 2011). In line with this notion, while surface acting as a typical work strains for front-line service employees can deplete their emotional resources and result in emotional exhaustion, CWX can provide emotional resources (Karasek et al., 1982; Wu and Hu, 2009) to buffer the positive effect of surface acting on emotional exhaustion.

Although previous studies have not investigated the mitigating effect of CWX in the association between surface acting and emotional exhaustion, research on coworker support provides indirect empirical evidence of the moderating effect. For example, coworker support moderates the effects of abusive supervision on emotional exhaustion (Wu and Hu, 2009), and the relationship between workplace anxiety and emotional exhaustion (McCarthy et al., 2016). In addition, a meta-analysis by Viswesvaran et al. (1999) found that social support from coworkers has a stronger mitigating effect on the stressor-strain relation than social support from leaders. Because surface acting serves as a stress source for service employee, the similar buffering effect of CWX on the effect of surface acting on emotional exhaustion can be expected. Combining the theoretical argument and empirical evidence, we propose that CWX will moderate the relationship between surface acting and emotional exhaustion.

Hypothesis 4: CWX will moderate the relationship between surface acting and emotional exhaustion, such that the positive relationship will be weaker when CWX is high.

# Moderating Role of LMX

We also predict that LMX as support from leaders may buffer the effect of emotional exhaustion on employee sabotage to customers. As leaders have more power and high state than employees (Diefendorff et al., 2010), the interaction between leaders and employees involves less emotional sharing but more economic exchange (Hüffmeier and Hertel, 2011). Because high LMX provides important material and instructional resources for employees to supply their resource pool, they are likely to use such instrument resources to overcome emotional exhaustion and regulate their behaviors (Ng and Sorensen, 2008; McCarthy et al., 2016) and perform better (Dulebohn et al., 2012; Jiang et al., 2014).

Further, according to the dual level social exchange theory (Schaufeli et al., 1996), employees are more likely to build a balanced reciprocity relationship with leaders and organizations. Based on the economic exchange, a higher level of LMX may lead to more employees positive behaviors benefiting the company and fewer negative behaviors. Thus employees receiving more support from leaders are more likely to feel the obligation to engage in positive behaviors (Sakurai and Jex, 2012) and perform effectively (McCarthy et al., 2016), and are less likely to engage in sabotage to customers even when experiencing emotional exhaustion due to resource losses.

Previous studies on LMX also provide indirect empirical evidence of the moderating effect. LMX buffers the relationship between emotional exhaustion and performance (McCarthy et al., 2016), and supervisor social support buffers the effect of negative emotions and both work effort on CWBs (Sakurai and Jex, 2012). Thus, we propose that:

Hypothesis 5: LMX will moderate the relationship between emotional exhaustion and sabotage to customers, such that the positive relationship will be weaker when LMX is high.

**Figure 1** summarizes the relationships proposed in the hypotheses above.

# MATERIALS AND METHODS

fpsyg-09-02197 November 12, 2018 Time: 16:27 # 5

# Participants and Procedure

Emotion labor has been mainly studied with samples from service industries such as hotels, hospitals, restaurants, airline services, call centers and transit companies. Nurses in hospitals interact with patients and their family members in their daily work, and they are expected to provide good customer service to patients (Drach-Zahavy, 2010), often requiring emotional labor (Ashforth and Humphrey, 1993). Pervious studies also suggested that nurses in hospitals experience a high level of surface acting at work (Grandey et al., 2012). Thus, we decided to use nurses as our sample for the current study. We collected data from seven large hospitals in China across two waves. We first obtained permission from the hospitals' directors and their ethical committees to conduct the study. We then met with head nurses from each department to explain the aims and requirements of the current study, and we asked them to encourage nurses to participate in the survey. After that, we sent out 800 questionnaires supervisors of departments who helped pass the questionnaires onto their followers. Each participant was told that they would receive 10 China Yuan (\$1.52) as compensation for their time. Two days later, we returned to the hospitals to collect the surveys and received 640 questionnaires, with a response rate of 80.00%. The first survey included measures of demographic variables, surface acting, and CWX.

Three months after time 1, all employees who completed Time 1 survey received a second questionnaire that assessed their emotional exhaustion, LXM, and sabotage to customers. Participants received a box of chocolate as an incentive gift for returning the survey. A total of 540 Time 2 surveys could be matched with a corresponding survey at time 1, of whom 95.20% were female (n = 540). The average age was 30.17 years (SD = 5.51) and their average tenure in their hospital was 8.01 years (SD = 6.35).

# Measures

Chinese versions of the followings measures were administered to participants in the current study. We used the translation and back-translation method (Brislin, 1980) to translate the scales from English into Chinese to make an equivalent meaning. Further, as the participants were clinic nurses in hospitals, we followed the suggestion of Schaffer and Riordan (2003) to modify some wording to ensure the applicability of the hospital context.

# Surface Acting

A five-item of emotions labor scale developed by Brotheridge and Lee (2002) was used to measure surface acting. Participants rated items on a five-point frequency scale from 1 = never to 5 = always. An example item is "Put on an act in order to display for my job" (α = 0.82).

# Coworker Exchange

We used a seven-item scale originally developed by Graen and Uhl-Bien (1995) and later modified by McCarthy et al. (2016) to measure CWX by replacing "supervisor" with "coworkers." An example item is "My coworkers understand my job problems and needs" (α = 0.97).

# Emotional Exhaustion

We used a five-item subscale of the Maslach Burnout Inventory General Survey developed by Schaufeli et al. (1996) to measure emotional exhaustion. Participants rated items on a seven-point Likert scale from 1 (never) to 7 (every day). An example item is "I feel tired after get up when I have to deal with work in the day" (α = 0.96).

# Leader-Member Exchange

Leader-member exchange was assessed with a seven-item scale developed by Graen and Uhl-Bien (1995). Participants rated the items about their supervisor from 1 (strongly disagree) to 5 (strongly agree). An example item is "My supervisor understand my job problems and needs" (α = 0.97).

## Sabotage to Customers

We used Skarlicki et al. (2008) five-item scale to measure sabotage to customers. To fit the context for the nurses, we modified the words "customers" in the scale into "patients" in the current study. Participants rated items on a five-point Likert scale from 1 = never to 5 = frequently. An example item is "Hung up on the patients" (α = 0.84).

# Control Variables

We controlled for nurses' demographic variables (i.e., age, job tenure, and education) on the study variables (i.e., surface act, emotional exhaustion, and sabotage to customers; Dahling and Perez, 2010; Wang et al., 2011). As deep acting is associated with ego depletion (Deng et al., 2016) and resources replenishment (Hülsheger and Schewe, 2011), we controlled deep acting when examining the relationships of surface acting with the outcome variables. Deep acting was measured with a three-item of emotion labor scale developed by Brotheridge and Lee (2002). Participants rated items on a five-point frequency scale from 1 = never to 5 = always. An example item is "Work hard to feel the emotions that I need to show to others" (α = 0.81).

# RESULTS

# Confirmatory Factor Analysis

A confirmatory factor analysis was conducted via Amos 21.0 to establish discriminant validity of the study variables. As shown in **Table 1**, the hypothesized five-factor model provides a better fit

### TABLE 1 | Results of confirmatory factor analysis.

fpsyg-09-02197 November 12, 2018 Time: 16:27 # 6


N = 540. CFI , comparative fit index; TLI, Tucker–Lewis index; RMSEA, root mean square error of approximation. CWX, coworker exchange; LMX, leader-member exchange.

to the data [χ 2 (df = 367) = 1062.29; RMSEA = 0.06, TLI = 0.95, CFI = 0.95] than all alternative models, showing evidence of discriminant validity.

# Descriptive Statistics

**Table 2** shows means, standard deviations, and correlations of the study variables. Consistent with our hypotheses, the correlations among surface acting, emotional exhaustion, and sabotage behavior to customers were all in the expected direction. This suggested that it was appropriate to conduct formal mediation analyses to test our hypotheses. Because of the high correlations between age and job tenure (r = 0.95, p < 0.001), we only control job tenure in the following analysis.

# Hypothesis Testing

We used the PROCESS macro for SPSS (Hayes, 2013) to test our hypotheses. As shown in **Table 3**, after controlling job tenure, education and deep acting in model 3, surface acting positively predicted employee sabotage to customers (B = 0.41, p < 0.001). Hence, Hypothesis 1 was supported. In support of Hypothesis 2, the result showed that emotional exhaustion positively predicted employee sabotage to customers in model 4 (B = 0.19, p < 0.001). After entering emotional exhaustion, surface acting was less significantly associated with employee sabotage to customers (B = 0.31, p < 0.001), whereas emotional exhaustion (B = 0.12, p < 0.001) was positively related to employee sabotage to customers in model 5. Furthermore, we calculated the indirect effect of surface acting on employee sabotage to customers through emotional exhaustion with 5,000 bootstrapped samples. The result showed that surface acting had a significant indirect effect on employee sabotage to customers through emotional exhaustion (indirect effect = 0.10, 95% CI [0.06,0.16]). Thus, Hypotheses 3 was supported.

Hypothesis 4 and Hypothesis 5 focused on the moderating effects of CWX on the association between surface acting and emotional exhaustion, and LMX on the relationship between emotional exhaustion and employee sabotage behavior to customers, respectively. We added the interactions term between surface acting and CWX into model 2 in **Table 3** to test Hypothesis 4. The interaction effect was significant (B = −0.24, p = 0.05). **Figure 2** further revealed that when CWX was lower, the positive effect of surface acting on emotional exhaustion was stronger (B = 1.08, t = 10.07, p < 0.001) than when CWX was higher (B = 0.55, t = 2.70, p = 0.007). Thus, Hypothesis 4 was supported. We added the interactions term between LMX and emotional exhaustion into model 6 in **Table 3** to test Hypothesis 5. The interaction effect was also significant (B = −0.09, p < 0.001). As shown in **Figure 3**, the positive effect of emotional exhaustion on employee sabotage behavior to customers was stronger (B = 0.22, t = 7.92, p < 0.001) when LMX was lower than when LMX was higher (B = 0.03, t = 0.51, p = 0.61), supporting Hypothesis 5.

# DISCUSSION

Our study found that surface acting has a positive effect on employee sabotage to customers through emotional exhaustion. Further, social exchanges buffer the negative effects. Specifically, the positive relationship between surface acting and emotional exhaustion is weaker for individuals with high CWX, and the positive relationship between emotional exhaustion and employee sabotage to customers is weaker for individuals with high LMX. Our findings suggest that while surface acting is ubiquitous and may result in serious negative consequences in service industries, social support from coworkers and leaders can potentially alleviate the harmful effects.

# Theoretical Implications

Our findings offer important theoretical insights. First, we contribute to the emotional labor literature by extending the effect of surface acting to employees sabotage to customers. Previous studies have suggested surface acting is positively related to employee harmful behaviors within organizations, such as deviance toward the organization (Bechtoldt et al., 2007) and harmful behavior toward coworkers (Deng et al., 2016). Our study suggests that surface acting might also lead to employee harmful behaviors toward customers during the service delivery. Surface acting might consume employees' resources to suppress their true inner feelings and result in resources depletion (Judge et al., 2009; Wagner et al., 2014; Deng et al., 2016), which might make employees be more antisocial and aggressive (Stucke and Baumeister, 2010; Friehe and Schildberg-Hörisch, 2017) during the interaction with customers, eliciting sabotage to customers. The findings supported the notion that surface acting has a broader impact and social cost not only within organizations (Deng et al., 2016), at home (Wagner et al., 2014), but also in service encounters. In addition, we contribute to the increasing literature on potential antecedents of employee sabotage to customers (e.g., Wang et al., 2011; Groth and Grandey, 2012; Skarlicki et al., 2016) and suggest that employee surface acting also has a potential to lead to their own harmful behaviors toward customers.

Second, based on the COR theory (Hobfoll, 1989), we contribute to the literature by finding that surface acting might



N = 540. <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. CWX, coworker exchange; LMX, leader-member exchange.

### TABLE 3 | Results of regression analyses.

fpsyg-09-02197 November 12, 2018 Time: 16:27 # 7


N = 540. <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. CWX, coworker exchange; LMX, leader-member exchange.

promote employee sabotage to customers through emotional exhaustion. As previous studies (e.g., Judge et al., 2009; Grandey et al., 2012; Wagner et al., 2014; Li et al., 2017; Kong and Jeon, 2018) have found, surface acting positively predicted emotional exhaustion, suggesting that surface acting might require employees to devote resources to suppress emotion impulses (Ashforth and Humphrey, 1993). Further, when people experience emotional exhaustion as a state of depletion of control

resources, they are more likely to engage deviant behaviors (Marcus and Schuler, 2004; Thau and Mitchell, 2010) and aggressive acts (Stucke and Baumeister, 2010), which might also extend to customers and lead to sabotage to customers. Our finding on the mediating effect of emotional exhaustion in the relationship between surface acting and employee sabotage to customers is consistent with this argument.

Third, based on the COR theory, our results show that social support from coworkers and leader can buffer the harmful effect of surface acting. CWX and LMX, as the main sources of social support in the workplace (McCarthy et al., 2016), have been found useful in providing resources to buffer the negative effect of work demands (Ng and Sorensen, 2008; Monica et al., 2016), such as surface acting and resource depletion (McCarthy et al., 2016). Our findings further demonstrate the importance of CWX and LMX in the resource depletion process of surface acting. On one hand, CWX can provide emotional resources (Karasek et al., 1982; Wu and Hu, 2009) to buffer the positive effect of surface acting on emotional exhaustion; on the other hand, LMX can provide more instrument resources to overcome emotional exhaustion and regulate their behaviors (Ng and Sorensen, 2008; McCarthy et al., 2016) to reduce employee sabotage to customers. This finding also adds to the previous literature that have demonstrated that individual factors such as emotion regulation self-efficacy (Deng et al., 2016) and situatonal factor such as climate of authenticity (Grandey et al., 2012; Li et al., 2017) can moderate the relationship between surface acting and resource losses. Our findings show that interpersonal factors such as CWX and LMX can be potential resource supplements to buffer the positive effect of surface acting and emotional exhaustion.

# Practical Implications

Our study provides important practical implications for organizations where service employees display surface acting frequently. First, surface acting might seriously harm the organizations when employees engage in sabotage to customers as a response. Service organizations should pay more attention to employees' emotions and train them to better deal with the negative emotions caused by customers. Previous research has suggested that perspective taking is an important way to decrease the employees' negative affect (Parker and Axtell, 2001; Bechtoldt et al., 2007), so that employees may not make efforts to regulate emotions or suppress negative affect to reduce the frequency of displaying surface acting. Besides, Li et al. (2017) found that employee mindfulness is negative associated with surface acting. Thus, service organizations can train their employees to master the strategy of perspective taking and be mindfulness. In addition, organizations can also train employees to engage in more deep acting (Deng et al., 2016) to reduce the occurrence of surface acting. Furthermore, organizations should also try to reduce internal sources of surface acting when it is inevitable from customers. For example, abusive supervision (Chi et al., 2018) and coworker interpersonal mistreatment (Adams and Webster, 2013) have been found to positively relate to surface acting, and organizations should try to establish a more supportive climate to reduce these experiences of employee and subsequent surface acting.

Second, given that emotional exhaustion might mediate the relationship between surface acting and sabotage to customers, it is vital for service employees to gain resources to recover from emotional exhaustion, which may in turn reduce sabotage to customers. Sonnentag and Fritz (2007) suggested that effective recovery activities including relaxation, psychological detachment, exerting personal control, and engaging in mastery experiences can be potential ways to recovery from emotional exhaustion. Besides, service organizations should also supply more opportunities for employees to reduce emotional demands (Goldberg and Grandey, 2007), have a rest (Trougakos et al., 2008), or engage in more social sharing (Baranik et al., 2017). Further, some targeted interventions on ego depletion also should take into consideration (Awa et al., 2010).

Third, the current study demonstrates the buffering effects of social exchange in both stages of the relationship between surface acting and sabotage to customers through emotional exhaustion. Thus, service organizations should try to promote high-quality relationships among employees, and between employees and their leaders. For example, Miles et al. (1996) suggest that open communication is a significant strategy to develop these relations. Besides, from the perspective of leadership, previous studies have suggested that the positive leadership, such as servant leadership, plays a key role in developing LMX and support climate (Wu et al., 2013; Liden et al., 2015). Thus organizations should encourage supervisors to serve as servant leaders and pay more attention to employee development and give priority to their needs and interests.

# Limitations and Future Directions

Our study has a few limitations. First, although we collected two waves of data to reduce common method variance, measuring all the variables using the same source can still raise potential concerns about common method variance. Future studies may use a third party to observe and report about the service interactions, such as coworkers perceived CWX, leaders perceived LMX, and customer reported sabotage to customers, which might provide a more objective assessment of employee behaviors and interactions. In addition, we can't draw conclusions concerning causality. Thus, future research may attempt to conduct longitudinal or experimental designs manipulating surface acting to verify causality.

Second, the data used in our studies were collected from nurses in China, limiting the generalizability of our findings; future studies should further replicate our findings with samples from other service industries such as hotels, banks, and airlines and other regions to extend our understanding of the effect of surface acting on employee harmful behaviors toward customers.

Third, our findings indicated that emotional exhaustion mediated the relationship between surface acting and employees sabotage to customers. However, it is likely that other mechanisms also exist. For example, integrating insights from work meaning theory (Rosso et al., 2010) may advance our understanding of how surface acting might result in more sabotage to customers through decreased work meaning. Besides, we only examined social exchange as potential moderators, and future research should further examine some other contextual factors such as climate of authenticity (Li et al., 2017) and individual factors such as emotion regulation self-efficacy (Deng et al., 2016) and emotional intelligence (Prati et al., 2009) as potential buffers.

# CONCLUSION

fpsyg-09-02197 November 12, 2018 Time: 16:27 # 9

Drawing upon the COR theory (Hobfoll, 1989) and social exchange theory (Blau, 1964), our study provides support for the mediating effect of emotional exhaustion in the relationship between surface acting and employee sabotage to customers, and social exchange as boundary conditions to buffer the effect of surface acting on employee emotional exhaustion and the effect of emotional exhaustion on sabotage to customers. These findings shed light on employee harmful behaviors as potential consequences of surface acting, as well as the potential underlying mechnisms and boundary couditions.

# ETHICS STATEMENT

An ethics approval was not required as per institutional guidelines and national laws and regulations because no unethical

# REFERENCES


behaviors existed in this study. We just conducted paper– pencil test and were exempt from further ethics board approval since this research did not involve human clinical trials or animal experiments. All subjects gave written informed consent in accordance with the Declaration of Helsinki. Research respondents were ensured confidentiality and anonymity. All participation was voluntary.

# AUTHOR CONTRIBUTIONS

HZ, ZZ, YZ, and CL: making substantial contributions to design, models, and hypotheses. HZ, YZ, and LZ: acquisition of data. HZ, ZZ, and YZ: analysis and interpretation of data. HZ, ZZ, and CL: wrote and revised the article. HZ, ZZ, YZ, CL, and LZ: final approval.

# FUNDING

The present research was supported by self-determined research funds from the colleges' basic research and operation of MOE (Grant No. 2018WKZDJC008).




relationship through perceptions of distributive justice. J. Appl. Psychol. 95, 1009–1031. doi: 10.1037/a0020540


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Zhang, Zhou, Zhan, Liu and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# On Averting Negative Emotion: Remedying the Impact of Shifting Expectations

Cecile K. Cho<sup>1</sup> and Theresa S. Cho<sup>2</sup> \*

1 International Business Department, Korea University Business School, Seoul, South Korea, <sup>2</sup> Strategy and International Management Department, College of Business Administration, Seoul National University, Seoul, South Korea

This paper examines how people anticipate negative emotions when faced with an uncertain outcome and try to manage their expectations. While extant research streams remain equivocal on whether managing expectations always succeeds, this research examines situations in which setting a low expectation can have an adverse emotional impact and suggests ways to alleviate this negative consequence. Using goal setting and a false-feedback paradigm, we show that, although individuals who set low goals to manage expectations can end up feeling more disappointed than those who set high goals (study 1), this negative impact can be avoided when individuals are reminded of their initial goals at feedback, or made aware of inaccuracies in forecasting their future emotion (studies 1 and 2).

### Edited by:

Jungkun Park, Hanyang University, South Korea

### Reviewed by:

Nara Youn, Hongik University, South Korea Chan Jean Lee, College of Business, KAIST, South Korea

> \*Correspondence: Theresa S. Cho tcho@snu.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 30 June 2018 Accepted: 15 October 2018 Published: 20 November 2018

### Citation:

Cho CK and Cho TS (2018) On Averting Negative Emotion: Remedying the Impact of Shifting Expectations. Front. Psychol. 9:2121. doi: 10.3389/fpsyg.2018.02121 Keywords: emotion, expectation, motivated goals, comparison standards, disappointment, anticipatory strategies

# INTRODUCTION

Anticipating the future, the capacity to represent future events flexibly and imagine diverse possible outcomes are fundamental to human being's survival. Individuals anticipate circumstances by taking expected future consequences into consideration when setting present goals and standards (Aspinwall, 2005). The link between one's anticipation for future outcomes and its impact on current behavior is a central topic in self-regulation (Bandura, 1982; Higgins, 1987, 1996, 1997; Carver and Scheier, 1990, 1998). Research on bracing demonstrates that people have a robust tendency to lower predictions of their own performance (Shepperd et al., 1996). In anticipation of a potentially disappointing outcome, individuals have been found to lower their expectations in order to reduce the gap between expectations and outcomes (see Carroll et al., 2006). For example, Carroll et al. (2006) found that student predictions for upcoming grades or exam scores moved from optimistic (over-estimating the actual results) to pessimistic (underestimating) as the time of receiving the grade approached (Shepperd et al., 1996). The finding from bracing research suggests that lowering expectations can indeed serve to cushion and avert disappointment when negative outcomes occur. This is because such lowered expectations are then used as a reference standard against which an outcome is judged. In this research, we investigate whether lowered expectations always serve as the reference standard for evaluating the outcome. Should the lowered expectancy fail to serve as the reference standard, the assumed cushioning effect may not take place.

Seen from a self-regulation perspective, avoiding disappointment could be conceptualized as a meta-goal that motivates the subsequent lowering of expectations. Given an uncertain future outcome, when a potential salient failure looms, a negative emotion accompanying this failure is also likely to be salient and trigger the defensive lowering of the forecasted outcome. The lay belief that one needs to "manage expectations," as doing so would help one avoid the negative emotion of failure, is common and influences how people assess what they will find satisfactory. In this paper, we consider such lowering of expectations as synonymous with choosing an easy goal, along with implications for a self-regulation strategy whereby avoiding a negative affect is part of successful goal pursuit.

# Setting Low Goals to Manage Expectations

fpsyg-09-02121 November 20, 2018 Time: 14:37 # 2

Goal setting is one of the most widely accepted paradigms of motivation and self-regulation behavior (e.g., Locke et al., 1990). Decades of research have consistently demonstrated that setting specific and challenging goals can powerfully motivate an individual. Recently, however, research on goals has begun to examine the negative consequences. Setting specific and challenging goals has been demonstrated to increase individuals' engagement in unethical behavior and excessive risks, as well as to demotivate and derail goal striving (Larrick et al., 2009; Ordonez et al., 2009; Townsend and Liu, 2012). If setting a challenging goal becomes a salient reference that can derail goal striving, other research has found that setting an easy and safe goal can also be demotivating, as people compare their performance to a better imagined outcome (Cho and Johar, 2011).

In this research, we start from the premise that setting low goals demotivates and leaves one disappointed. We draw on the recent finding that evaluation standards can shift over time (Van Dijk et al., 2003; Monga and Houston, 2006; Cho and Johar, 2011). In particular, Cho and Johar (2011) found that setting goals low could be less satisfying because people often compare to a better possible outcome. In our research, we examine the conditions under which such a shifting of goal standards is likely and test ways to intervene. Across two studies, we explore conditions under which setting easy goals (even if achieved) can lead to perceptions of failure and show that reference standards used to evaluate outcomes is not the initial expectation (study 1); and, making participants aware that comparative standards can shift might remedy the perception of failure and accompanying negative emotions. In study 1, we manipulate motivational states to induce goal levels while measuring the chronic disposition to anticipate failure and negative emotions in order to predict the anticipatory lowering of goals and expectations. In study 2, we test the assumption that individuals concerned with prevention goals are more likely to manage expectations. In this way, we show that increasing awareness of the futility of trying to manage a future affect could serve to intervene against this counterproductive tendency.

# Anticipated Emotion and Goal Setting

In order to demonstrate the self-regulatory nature of a goallowering strategy, we draw on motivation literature to create conditions under which low-vs.-high goals are set. Similarly, research on regulatory focus suggests that avoidance orientation (prevention focus) is associated with pursuing minimal goals, whereas approach orientation (promotion focus) is associated with pursuing maximal goals (Brendl and Higgins, 1996; Forster et al., 1998; Jain et al., 2006). The motive to avoid failure is associated with the anticipated negative emotion that accompanies this failure (McClelland, 1953). The tendency to set low goals is likely to be exacerbated when consumers are concerned about minimizing the negative emotion that accompanies failure (Van Dijk et al., 2003).

# STUDY 1

The purpose of this study is twofold. First, we seek to show that people lower their expectations of what will be satisfactory as an anticipatory coping strategy to avoid disappointment. Second, we seek to show that lowering expectations may not help to avert disappointment even if expectations are confirmed. Consistent with previous research, our prediction is that lowering expectations will lead to feeling more, not less, disappointed, because the initially set expectation is not used as a comparison standard when the performance outcome is revealed and evaluated (Cho and Johar, 2011). In this study, we use avoidance and approach motivations to operationalize and induce setting of low and high goal levels, respectively (Friedman and Forster, 2001; Zhou and Pham, 2004). Our logic in using approach and avoidance motive as antecedents of lowering expectation comes from the motivation literature which states that avoid motivation should trigger a greater sensitivity to potential failure; the greater concern with failure and negative affect, the greater tendency to lower goals (for example, Forster et al., 1998). It also follows from the bracing literature which found that when faced with potentially negative news, people lower their expectations in order to prepare for this bad new (Shepperd et al., 1996). Therefore, we predict that under avoidance motivation, participants will lower their expectation (set lower goals), whereas under approach motivation, participants will set a comparatively higher expectation (set higher goals).

If the lowered expectation is used to evaluate the outcome (which is confirmed using the false feedback paradigm) then the measure of disappointment should be minimal. If people end up comparing their outcome to imagined outcomes that are better ("it could have been better"), then even after initially lowering their expectation strategically (Shepperd et al., 1996), the assumed benefit of the lowered expectation is likely misguided and needs to be corrected.

We draw on previous research and vary the information present at the time of performance feedback to test the preceding hypothesis that potential performance is likely to be the spontaneously recruited comparison standard (Cho and Johar, 2011). In their study, Cho and Johar (2011) showed that when respondents were provided with their performance results, they seemed to compare their outcomes to the highest potential, which likely served as a spontaneous comparison standard. However, making the initially set goal salient at the time of feedback should result in no difference between the low-andhigh goal conditions: that is, both groups should compare their performance to the salient goal and realize they have met the goal and should therefore be equally satisfied. In our study, we manipulate approach and avoidance motives so as to invoke fear of failure and a negative affect. We seek to show that the desire to avoid a negative affect from failure is driving the lowering of expectation and thereby imparting a negative consequence on satisfaction because the comparison standard used is not the lowered expectation. Specifically, we seek to show that even when expectation is confirmed, those who had set a low expectation to avoid failure actually end up feeling disappointed and less happy compared to those who did not.

# Experimental Design

fpsyg-09-02121 November 20, 2018 Time: 14:37 # 3

We tested our hypothesis using a 2 (approach vs. avoidance motive) × 2 (information at feedback: performance-only vs. performance-and-goal) between-subjects design. One hundred and thirty-nine undergraduate and graduate business students at a large public university participated in the experiment for extra credit in an on-campus lab. Once seated, all participants were welcomed and provided information on the purpose and description of the study and time required (approximately 20 min). They were also informed that their responses will be unidentifiable and remain anonymous, and that participation was entirely voluntary. The students clicked "yes" to provide consent before proceeding and were debriefed of the purpose of the study upon completion. Ethics approval was not required as per the institution's guidelines and national regulations; the study was exempted from the ethics review process by the Ethics Committee (Dr. Augustine Kposowa, Chair, and Dr. Rollanda O'Connor, Vice Chair, Human Research Review Board, Office of Research Integrity, University of California, Riverside).

### Stimuli and Procedure

The experiment included two phases. In the first phase, participants were asked to complete two tasks framed either in an approach or an avoidance manner. The first task involved proofreading a short article in which the participants, in the approach condition, were instructed to "find the maximum number of misspelled words." In the avoid condition, participants were instructed to "avoid missing any misspelled words" (Zhou and Pham, 2004). The second task was to solve a paper-andpencil maze (Friedman and Forster, 2001) in which participants had to guide a cartoon mouse from the center of the maze to the exit. In the approach condition, a piece of cheese was depicted as sitting at the exit, whereas in the avoidance condition, a prowling owl was depicted as looking over the mouse from the opposite end of the exit. According to Friedman and Forster (2001), the cheese and the owl operationalize cognitive representation of "seeking reward" and "avoiding punishment," respectively. It was predicted that, compared to those who were primed with an approach frame, participants primed with an avoidance frame would set more conservative return goals. It is noted that the manipulation used was originally designed for regulatory foci priming in which constructs promotion and prevention foci are close correlates of approach and avoidance motivations (Friedman and Förster, 2005, 2013).

In the second phase, participants completed the main experiment using an online financial investment interface. Performance is always at the level of the goal; that is, goals are met in all conditions. Participants were told that they would make investment decisions and receive feedback on their performance based on the actual performance of the stocks they picked. They were told:

"Imagine that you are living in a foreign country and need to invest your money. You have an investment budget of \$5,400 and want to invest it in the stock market of this country. Given the market conditions in this country, at the end of a month, you can expect your portfolio to yield between 0 and 20% in return. As with any investment in a financial market, investing in stocks involves risk."

On the next page, participants read, "You make investment decisions on the first of every month – that is, you trade on the 1st of each month. It is now the first day of March. What is the rate of return you would be satisfied with for this month?" Participants then picked a target level of return from the following possible target returns: 0, 2, 4, 6, . . .20%. Next, they responded to a manipulation check question to verify their awareness of the relative level of their expectation levels: "What is the level of expectation you have set for your portfolio's performance?" (1 low expectation; 9-high expectation).

On the following page, participants constructed stock portfolios using an interactive interface. They were presented with a list of 20 fictitious stocks along with key information such as P/E ratio, price, ROE, debt-to-equity ratio, and EPS (quarter vs. year ago). The interactive interface simulated information layout of the E∗Trade website. The experimental program recorded the time participants spent on reviewing, selecting, and allocating the three stocks. Low- and high-goal setters did not differ in terms of the amount of time that they spent on the task [Mlogoal = 7 min 48 s vs. Mhigoal = 7 min 12 s; F(1,139) = 0.88, n.s.].

After a ten-minute filler task, participants were given feedback and received their stock portfolio returns on the subsequent screen. The returns matched their goals (goal+0.04%, the latter added to increase believability of the feedback). Participants were led to believe that their stock picks and allocations were used to calculate the actual returns using real data for that month for which they made their decision. For the "performance-only feedback" (default) condition, only the performance information was provided at feedback. Those in the "performance-and-goal at feedback" condition were told, "You had predicted that you would be satisfied with (actual goal) % for the past month. Your portfolio has resulted in a return of (actual goal plus 0.04%)."

Respondents then recorded their thoughts about their performance ("Please write down all thoughts that came to mind when you saw the performance level of your stock portfolio."). They next rated their disappointment with the performance of their stocks on a 9-point scale. As a final separate study, respondents were asked to complete a set of "personality questionnaires," which included a set of questions designed to measure their affect-management concern ("I chose a performance goal that would reduce my future disappointment"; "Choosing a low goal is better than high goal because it is helpful in dealing with anxiety"; "Setting a low goal is a good way to prepare for an uncertain outcome"; "I kept in mind that not meeting my goal would make me unhappy"; 9-point scale; 1 = not at all agree; 9 = definitely agree; alpha = 0.72). Affect-management concern measures were collected to test the role of an anticipatory

negative affect in goal setting. Our prediction is that concern regarding a future affect, and the desire to avoid a negative affect, will mediate the predicted effect of avoidance prime on lowering goals. Finally, we collected involvement and expertise measures, which did not differ across conditions.

# RESULTS

## Manipulation Check

To verify that those participants who managed their expectation by setting low goals were aware that they chose a low target (vs. high) we regressed the self-rated level of expectation item to the actual level of goal. As anticipated, the level of goals set and perceived level of these goals were strongly correlated (r = 0.36, p < 0.0001).

### Setting Goals Low to Avoid Negative Emotion

We predicted that individuals lowball their goals to avoid uncertainty and a negative affect from failure. As expected, participants under an avoidance frame (vs. approach frame) were found to set significantly lower goals (i.e., target rates of return) (β = −0.61, p < 0.0001). We then tested the mediating role of affect-management concern (combined measure of four question items; min = 1, max = 9) on level of goals set, using Hayes' (2013) PROCESS model 4. A bootstrapping analysis (5,000 iterations) revealed that affect-management concern had a significant, indirect effect on the level of goal set (β = 0.43, 95% confidence interval [CI]: [0.15, 0.79]). This result suggested that goal setting under uncertainty could serve as an anticipatory coping mechanism to avoid a negative affect. Isolating the measure for salience of disappointment, the indirect effect of the desire to avoid disappointment on the goal level was stronger under avoidance motivation compared to approach motivation (β = 0.51, 95% confidence interval [CI]: [0.18, 0.89]).

Analyses were conducted using only the respondents in the approach frame group, who set their goals on the high end of the 0 to 20% return scale (≥10% median goal set) and those in the avoidance frame group, who set their goals on the low end of the scale (<10%; N = 124). Restricting the data was to ensure against concerns of self-selection, or third variable problem, due to chronic tendency such as general optimism and pessimism. The main question was whether this strategy of lowering goals actually helps one to avoid a negative emotion. In other words, does lowering one's goal help in avoiding disappointment?

### Disappointment

The hypothesis was tested using a regression analysis with "disappointment" as the dependent variable (DV) and the selected goal level (between 0 and 20% return, mean-centered) and feedback (performance-only vs. performance-and-goal) as the inDVs. The analysis revealed a significant main effect of goal level (goal β = −0.29, p < 0.0001) such that disappointment increased as the level of goal decreased. There was also a main effect of providing a goal at feedback such that reminding participants of their goal at feedback led to increased satisfaction (β = 0.80, p < 0.01). The interaction between the feedback variable (goal reminded or not) and the goal level was directionally consistent but not significant (β = −0.43, p = 0.13). More to our interest is whether those who lower their goals and actually achieve their goals are helping themselves to avoid the negative affect of disappointment. As illustrated in **Figure 1**, within the performance-only condition, those who achieve their low, safe goals report greater disappointment than those who set and achieve a higher goal. Follow-up contrasts confirmed that participants were more disappointed when they had set a low-vs.-high goal (based on median of 10%) in the performanceonly feedback condition (Mlogoal = 5.42 vs. Mhigoal = 3.14), [F(1,120) = 27.17, p < 0.0001]. Providing the initial goal along with the performance at the time of feedback eliminated the effect of setting low-vs.-high goals (Mlogoal = 3.04 vs. Mhigoal = 2.18), (F = 3.45, p = 0.09).

# Discussions and Limitations

The higher disappointment reported by the low-expectation individuals suggests that setting low expectations is not conducive to averting disappointment despite having such expectations confirmed. Those who manage their expectations and have them confirmed were more disappointed compared to those who did not. The interaction of providing goals at the feedback evaluation suggests that the portfolio-performance comparison was not to the initially set goals. This result is consistent with Cho and Johar (2011) finding on shifting reference standards whereby people were observed to invoke superior, alternative reference standards when they evaluate their performance. Results from experiment 1 suggest that the desire to manage a future affect by lowering a target goal is not conducive to success feedback. This finding suggests that in fact, trying to manage future success undermines the selfregulatory feedback loop by instigating perceptions of failure. Awareness of the initial goal at feedback remedies this effect. It is interesting to note that all participants reported feeling disappointed with the outcome, although in different degrees. This finding is consistent with extant research on counterfactual thinking and the general tendency to upwardly compare to unattained, alternative outcomes (see, for example, Roese, 1997).

A limitation of this study was that the level of goal and the performance outcome were the same. Although this was by design, it lends to the concern that performance outcome that varied may drive the effect. Therefore, as the purpose was to examine the effect of managing expectations (not outcome) on emotional reactions to a confirmed expectancy, holding the outcome constant would lend additional strength to the findings. This limitation notwithstanding, we note that it does not take away from the core premise – that is, all participants in the study were choosing the target (expectations) that they considered would be satisfactory. The use of the stock-picking task, while interactive and self-involving, could be subject to the concern that it is artificial and unfamiliar. Replication of the results using real monetary consequences would enhance this concern. Further, replicating the findings using other domains of self-relevance would enhance generalizability of this study's findings. While different in scope, we explore more familiar and self-involving domains in our next study.

# STUDY 2

Study 1 demonstrated that setting goals low might be counterproductive in avoiding disappointment even if the goals are met. This study examines the possibility that greater awareness of the difficulty of forecasting future emotion may reduce the lowballing tendency among those with a chronic tendency to set low goals. Whereas the preceding study primed participants to set different goal levels, this study examines chronic tendencies to set low or high goals, with the level of goal set as the main DV. One such chronic tendency, or trait difference that may dictate whether goals are chronically set low or high is regulatory focus (Shah et al., 1998). According to Shah et al. (1998), people can construe aspirational standards as minimal goals they must attain or as maximal goals they hope to attain. Minimal goals differentiate negative from nonnegative events, whereas maximal goals differentiate positive from non-positive events (Brendl and Higgins, 1996; Idson et al., 2000). Freitas et al. (2002) have demonstrated that a promotion focus tends to foster concerns with maximal goals while a prevention focus fosters concerns with minimal goals. Because goals within a promotion focus are seen as opportunities to try for optimal outcomes – whereas goals within a prevention focus are perceived as minimal requirements – it could be argued that this dispositional tendency in goal perception influences the actual level of aspiration one sets for oneself. Specifically, an individual with a prevention focus is likely to set low goals because he is concerned with avoiding failure and achieving the minimum is perceived to fit this goal. In contrast, a promotionfocused individual is likely to set higher goals because he is less concerned with avoiding failure and the accompanying disappointment. Can this tendency to set low goals be mitigated? One method of intervention would be to increase awareness regarding the fallacy that one can avert disappointment by managing expectations. If individuals are made aware that anticipated affect is often misleading and inaccurate, the tendency to set preemptively low goals in anticipation of a possible negative affect could be prevented. Research on context effects and bias correction has shown that, when people become aware of their reactions as being due to primes rather than the target, they attempt to "correct" their reactions by consciously "resetting" and adjusting their judgments (Martin, 1986; Schwarz and Bless, 1992; Wegener and Petty, 1997). In other words, when made aware of a potential bias, individuals would attempt to correct for this bias. This motivation to correct, we reason, should also operate when an individual is made aware of the tendency to overpredict a potential negative affect and the worst outcome.

In this study, we test to see whether making participants aware of the inaccuracy of the lay theory (tendency to overpredict negative emotion and the belief that lowering goals and expectations can help to avoid it) can trigger a corrective process, whereby the lowballing of goal is mitigated. We expect that those who are aware of the inaccuracy of their predicted affect will effectively set higher goals, particularly those who are chronically likely to lower goals as measured by their regulatory orientation.

# Experimental Design

The study used a 2 (prime: awareness vs. no awareness) × 2 (chronic regulatory focus: promotion vs. prevention) betweensubjects design with the level of goal as the main DV. One hundred and thirty participants participated as part of extra credit for a class. As with study 1, all participants were informed of the confidentiality and unidentifiability of their responses and that their participation was entirely voluntary. They were told that they would perform several unrelated tasks consisting of a reading comprehension task, a stock-picking study, and a lifestyle questionnaire. The students clicked "yes" to provide consent before proceeding and were debriefed of the purpose of the study upon completion. As with study 1, this study was approved by the university's Human Research Review Board. Ethics approval was not required as per the institution's guidelines and national regulations; the study was exempted from the ethics review process by the Ethics Committee (Dr. Augustine Kposowa, Chair, and Dr. Rollanda O'Connor, Vice Chair, Human Research Review Board, Office of Research Integrity, University of California, Riverside).

# Stimuli and Procedure

In the first study, presented as a comprehension study given by the English department, respondents in the "awareness" prime condition read a newspaper article describing research by Gilbert and Wilson (2000, c.f. Gertner, 2003) on the unreliability of forecasted feelings. Those in the "no awareness" prime condition read an article about strategies to avoid disappointment to succeed in life. Participants were asked to rate their agreements on five question items using a 7-point scale (1 = strongly disagree and 7 = strongly agree): "This article was convincing"; "It is wise to avoid disappointment in navigating life's challenges"; "Managing one's expectation is a good strategy to cope with life's uncertainties and disappointments"; "To be happy, it is important not to think so much about possible failures and disappointments"; "It is important to know that what people expect will be disappointing or painful will not be so bad."

In the next ostensibly unrelated study, the scenario used was similar to that in study 1, except for returns that ranged from negative to positive, ("Given the market condition in this country, at the end of a month, you can expect your portfolio to yield between −20 and 20% in return."). Participants worked on a computer and were led to believe that the study would entail an actual stock-picking task after which they would be provided how well their chosen stocks performed (identical to study 1); but because we are interested in the target level chosen as the main DV, the study would stop when they chose a level of return. Participants read the stock-picking scenario and chose from a set of 21 possible levels of returns (0, 1, 2, 3, . . ., 20%). In addition, in the next (unrelated) "lifestyle and value survey" that followed, participants were asked to state their target in various life domains, namely saving money, exercising, losing weight, and any other goal they may have previously set, and whether these goals were easy or difficult. These questions were expected to serve as replicates of the predicted interaction effect of regulatory foci and awareness of inaccuracies of the lay theory on managing goal levels.

In the final "unrelated study," respondents answered the 11-item Regulatory Focus Questionnaire (Higgins et al., 2001) where they rated their history of promotion and prevention success and failure on 5-point scales (1-never or seldom; 5-very often). Participants were then debriefed and told that the study was concerned with goal-setting rather than financial decisionmaking. We expected that chronic orientations on the promotion and prevention dimensions would correlate with the level of goals set such that a prevention focus would lower the goal level. Specifically, it is predicted that prevention individuals will set lower goals than promotion individuals, and that within the awareness condition, this tendency will not be observed whereby prevention individuals will set their goals at a similar level as compared to promotion individuals.

# RESULTS

### Manipulation Check

An ANOVA was run on the measures of agreement with the convincingness of the article that was read. Nine respondents who rated the article as completely unconvincing were dropped from the study (seven from no awareness and two from awareness conditions), as were six participants whose numerical responses were identical throughout the questionnaires, resulting in a sample of 115 participants. Next, ANOVA was run on measures of agreement with the effectiveness of the goal-lowering strategy ("To be happy and successful, it is important to avoid disappointments in life," "Managing one's expectation is a good strategy to cope with life's uncertainties and disappointments.") and on two measures of agreement with the tendency to exaggerate future negative affect ("To be happy, it is important not to think so much about failure and disappointment"; "To be happy, it is important to know that what we think will be disappointing or painful will not be so bad."). The two pairs of measures were highly correlated and combined for analysis (r = 0.68, p < 0.0001; r = 31, p < 0.001). As expected, those in the no-awareness condition were significantly greater in agreement (M = 5.16) on the effectiveness of the lowballing strategy compared with the awareness condition [M = 4.27; F(1,113) = 12.26, p < 0.001]. The awareness condition group was also better informed regarding the inaccuracies of the lay theory in avoiding a future negative affect compared to the no-awareness condition (M = 5.69 vs. 5.03, p < 0.005).

### Goal Setting

Level of goals chosen, the main DV of interest, was analyzed first by regressing goal level to prime type (awareness vs. no awareness) and Regulatory Focus (RF) Questionnaire scores (Higgins et al., 1997). Consistent with the RF theory which holds that an individual can be high or low in prevention and promotion foci, that is, that the motivational orientations are orthogonal, the prevention and promotion scores did not correlate (Pearson r = 0.09). Regression analyses for promotion and prevention foci as predictor, goal level chosen as DV, and awareness treatment vs. control treatment as moderator were conducted. As predicted, the chronically prevention-focused

individuals were more likely to set a low goal (standardized β = −0.21, p < 0.05), whereas promotion focus had no significant effect (standardized β = −0.07, n.s.). Given that promotion and prevention foci are orthogonal, we proceeded to focus on individual differences in prevention orientation (high vs. low) as the predictor variable in our analysis. Moderator and interaction between predictor and moderator were not significant. More to our interest is whether the chronic tendency to lower the level of goal by the high prevention individuals can be intervened. Therefore, an ANOVA was run for the 2 (Prevention: High vs. Low) × 2 (prime type: awareness vs. no awareness) study, with the prevention measure separated into high or low at the median (high if >19). There was a marginally significant effect of the prevention focus on the level of goal [F(1,111) = 3.5, p = 0.07] such that high-prevention respondents were setting directionally lower goal levels compared to the low-prevention respondents. More to our interest is whether the high-prevention individuals, who are presumably the

chronic low goal setters and prone to managing expectations, can "correct" this tendency when made aware of the tendency to over-anticipate negative emotion. In other words, we tested whether after being made aware of the fallibility of predicting future emotion, the lowering of goals can be prevented. Results supported this prediction (see **Figure 2**). Within the highprevention group, those who read the awareness treatment set their goals significantly higher (M = 13.12%) than those in the no- awareness group (M = 11.00%), [F(1,111) = 4.11, p < 0.05]. In comparison, those in the low-prevention group did not differ significantly in their level of goals (Mdisapp = 12.25%, Mcorrect = 13.26%), (F < 1). Notably, the hi-prevention group under the awareness prime set goals that were not significantly different from the promotion group, suggesting that the dispositional tendency to lowball on goals (together with the negative emotion of disappointment) may be "corrected" by increasing awareness of the inaccuracies in forecasting one's future affect.

Following the stock-picking scenario study, participants also completed the "General Lifestyle Survey," in which they were asked to rate their goals in various domains, such as saving money, going to the gym, and losing weight (Studies 2a, 2b, and 2c, **Table 1**). It is noted that this study was conducted in mid-January and February, and some of the questions were designed to correspond to seasonal concerns. The participants were asked to set their goals for three other goal-setting domains: Exercise, ("Exercising and going to the gym is an important part of staying healthy. How often do you plan to go to the gym this year?"; 1 never, 2-once a month, 3-twice a month, 4-three times a month, 5-four times a month, 6-once a week, 7-twice a week, 8-three times a week, 9-four times a week or more); saving ("People often set savings goals to curb their spending. What is the amount of savings you would strive for this year?"; less than 5%, 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40% or more); and losing weight ("Losing weight is an important part of maintaining health. What is your goal for losing and maintaining your weight? In other words, how many pounds do you plan to lose this winter season?"; open-ended response). Participants were then asked to rate the perceived difficulty of the goal they indicated using a 9-point scale (For you, this goal is. . .: 1-very easy; 3-somewhat easy; 5-neither difficult nor easy; 7-somewhat difficulty; 9-very difficult). We used the perceived difficulty of the goal as the main measure of goal level chosen.

Interestingly, the patterns of results were similarly in support of the finding that the chronically low-goal setting (high prevention-focused) individuals who read the one-page description of the affective forecasting research (awareness prime) chose goals that were more difficult (compared to low prevention-focused individuals). For the exercise goal, highprevention individuals who read the awareness treatment set their goals significantly higher in rated difficulty (M = 5.58) than under the no-awareness treatment (M = 3.07), [F(1,111) = 18.40, p < 0.0001], while there was no difference for the low-prevention individuals (5.21 vs. 5.77, n.s.) (**Appendix 1**, in **Supplementary Material**). For savings, for the high-prevention individuals, those in the awareness treatment condition set their goals directionally higher in difficulty (M = 5.46) than under the noawareness treatment (M = 4.56), [F(1,111) = 2.94, p = 0.08]; there was no difference for the low-prevention individuals (5.46 vs. 6.06, n.s.) (**Appendix 2**, in **Supplementary Material**). Similarly, for losing weight, the pattern was significant and in a similar direction. High-prevention individuals in the awareness condition were setting significantly more difficult goals than high-prevention individuals in the no-awareness treatment [5.38 vs. 3.39; F(1,111) = 5.71, p < 0.05], while no such effects were observed for the low-prevention individuals (5.58 vs. 5.68, n.s.) (**Appendix 3**, in **Supplementary Material**).

# DISCUSSION

The purpose of Study 2 was fairly straightforward. That is, given that concern with managing future disappointment and negative emotion leads to a counterproductive tendency to lower one's goal, could this tendency be prevented by making better information available about the misguided assumptions behind such a tendency? To verify the premise that a salient desire to avoid disappointment can motivate one to "manage expectations" and set a low goal, we used chronic differences in concerns of avoiding negative outcomes using the motivational construct of a prevention focus. We tested our hypothesis on individuals with different regulatory orientation as a proxy for chronic low-goal setting tendency and found that the chronic tendency may be "corrected" when individuals are made aware that lowering goals is not always helpful in maximizing happiness. While financial decision-making domain was chosen for the parsimonious operationalization and replication of previous research (Cho and Johar, 2011), we test and find consistent results using three additional domains of goal setting. The simple replication across the four domains provides additional validation to the generalizability of awareness as a way to mitigate the goal setting behavior.


We manipulated motivational states to induce greater sensitivity to avoiding negative outcome (study 1) as well as measured trait differences (study 2). We tested the efficacy of this implicit strategy of goal lowering and expectations in averting disappointment (study 1) and more importantly, to explore ways to intervene against the potentially negative impact of such a strategy. Our findings suggest that an intervention of a reminder of the initial goal (study 1) or making individuals aware of the potential inaccuracies of trying to manage their future affect (study 2) may help to alleviate the likely negative emotional impact.

Study 1 demonstrated that helping consumers retain their original goals and perspectives might counter the potential impact of upward comparison and the negative emotion that accompanies this process. This study tested more directly whether adverse emotional consequence of lowballing one's expectation could be countered. It was demonstrated that better informing the consumers might be effective in correcting the misguided tendency to lower one's expectations. The results suggest that better informing individuals of the difficulties in accurately forecasting their emotional state can effectively deactivate resorting to lowering one's goals and expectations.

# GENERAL DISCUSSION

This research investigated the equivocal claim that managing expectations and setting low goals can help to prepare for a potentially negative outcome. While various streams of research including bracing (Shepperd et al., 1996) and the expectancydisconfirmation model (Oliver, 1980; Van Dijk et al., 2003) have lent support to this lay strategy of "managing expectations," more recent research has offered boundary conditions under which such a strategy may backfire (Ordonez et al., 2009; Townsend and Liu, 2012). Our findings are compatible with previous findings that show that, when goals and expectations are motivated and set low, it may not be stable enough to serve as a reference standard when one's performance outcome is revealed and assessed. Given the instability of reference standards, and the negative emotional consequences, the two studies tested ways to intervene against the counterproductive strategy of setting low expectations. More generally, making individuals aware of the likely inaccuracies in predicting what would be satisfactory in the future, namely,

# REFERENCES


the demonstrated tendency to forecast one's affect inaccurately (Gilbert and Wilson, 2000; Wilson and Gilbert, 2003), may offer a simple remedy for the negative impact of "managing expectations." An interesting avenue for future research would be whether the negative consequence of managing expectation may operate on an organizational level (Locke et al., 1990). Existing research on goals in the workplace suggests that setting goals that are too high can motivate unethical behavior (e.g., Schweitzer et al., 2004). By the same token, it would be worthwhile to examine the possibility that setting too manageable an expectation may be counterproductive in optimally motivating employees. The findings from this study also illuminate the implications of earnings management for stakeholder groups, as well as the capital market (Hirst et al., 2008).

# AUTHOR CONTRIBUTIONS

CC initiated the theoretical motivation for the paper, while TC and CC collaborated on the study design and data collection and analyses. CC wrote the first draft and went through multiple iterations with TC.

# FUNDING

This study was partially supported by Korea University Business School Research Grant, Institute of Management Research at Seoul National University and the University of California Riverside's School of Business Administration.

# ACKNOWLEDGMENTS

This paper was inspired by the first author's earlier work with Gita Johar. The authors thank Professor Johar for her invaluable suggestions in the earlier stage of developing this paper.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg. 2018.02121/full#supplementary-material


solving. J. Pers. Soc. Psychol. 88, 263–275. doi: 10.1037/0022-3514. 88.2.263


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Cho and Cho. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Impact of Emotional Harassment on Firm's Value

### Yun Hyeong Choi<sup>1</sup> , Hee Jin Park<sup>2</sup> and Seong-jin Choi<sup>1</sup> \*

<sup>1</sup> School of Business, Hanyang University, Seoul, South Korea, <sup>2</sup> Accounting/Tax and Management Information Systems, Kyonggi University, Suwon, South Korea

The activities and consequences of workplace bullying and harassment have been widely explored in the literature but mainly studied within the scope of individuals or at the team level. Taking a holistic approach, we associate the concept of bullying with firm-level performance as well as stakeholders' responses in the market. In this paper, we examine whether and how market investors react to the news of corporate harassment by top officials of publicly listed firms in Korea. Using a standard event study methodology and multiple regression analysis with matched sample, we find significantly negative stock price reactions to news of corporate bullying. We also find that the impact is more salient if emotional bullying is involved and discuss both the theoretical and practical implications of these findings.

Keywords: power trip, emotional harassment, conglomerate, market reaction, Korean-listed firms, event study, event system theory, emotional contagion

### Edited by:

Weon Sang Yoo, Korea University, South Korea

### Reviewed by:

Seung-Wan Kang, Gachon University, South Korea Khan-Pyo Lee, Sogang University, South Korea

# \*Correspondence:

Seong-jin Choi seongjin@hanyang.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 12 September 2018 Accepted: 07 November 2018 Published: 04 December 2018

### Citation:

Choi YH, Park HJ and Choi S-j (2018) Impact of Emotional Harassment on Firm's Value. Front. Psychol. 9:2333. doi: 10.3389/fpsyg.2018.02333

# INTRODUCTION

Workplace bullying, defined as situations in which employees suffer from persistent and hostile interpersonal behavior at work (Leymann, 1996) has received growing attention. Extant studies in psychological and organizational research explore the negative impacts not only on the victim but also on other employees (Einarsen et al., 1994; Zapf et al., 1996; Mikkelsen and Einarsen, 2002), and firms have shown considerable interest in minimizing the harm of harassment within organizations (Langan-Fox et al., 2007). However, although previous empirical studies examined the relationship between bullying and its consequences at the individual and team level (e.g., Olweus, 1993; Einarsen, 1999; Einarsen et al., 2003; Griffin and O'Leary-Kelly, 2004; Hansen et al., 2006; Langan-Fox et al., 2007; Simons, 2008; Mathisen et al., 2008; Balducci et al., 2009; Nielsen et al., 2010; Dehue et al., 2012; D'Cruz et al., 2018), few have discussed the firm-level dynamics and consequences of bullying in detail. For instance, there have been frequent occurrences of bullying and crises in firms that cannot be solved at the micro level, such as the recent "nut rage" incident on Korean Air.<sup>1</sup> Therefore, research on the corporate-level impact of bullying and its impact on a firm's value, which have been seldom studied, should be conducted.

The term "bullying" was chosen for this research for a reason. Workplace bullying can be divided into abuse by the boss, the co-worker, or the subordinate (Einarsen and Mikkelsen, 2003).

<sup>1</sup>Korean Air Vice President Cho Hyun-ah was charged with violating aviation safety law. Ms. Cho also forced the flight crew and the junior attendant who served the nuts to apologize on their knees like slaves. She used abusive language and threw documents. Her bullying undoubtedly affected many important relationships at Korean Air. Because she is the daughter of Cho Yang-ho, the airline's chairman, it is likely that when the prosecution's investigation began, the share price, which had soared to 19% before the incident, plummeted by more than 200 billion won (an organization-level feature) (https: //www.nytimes.com/2018/04/13/world/asia/nut-rage-sister-korean-air.html)

Top managers' behaviors are closely related to the firm's reputation and performance (Hambrick and Mason, 1984), and thus we focus on the bullying behaviors of TMT (Top Management Team) to identify the significance of the consequences at the level of the firm. Social undermining and abusive supervision behaviors of the heirs who acquire management rights or executives are different from bullying at work. Therefore, to more accurately define harassment by heirs and executives, we refer to assaults on subordinates as "power trips." This expresses their behavior of harassing people and making their lives difficult just because they have more harassing most other employees. In fact, as a holistic approach, the concept of the power trip should be associated with firm-level performance as well as stakeholders' responses in the market.

Moreover, to extend the prior research on workplace bullying, and especially the concept of the power trip, from the individual and organizational levels to the firm level, we consider two moderators—emotional harassment and conglomerate—in our regression model and attempt to apply them to psychological perspectives. First, we classify power trips into two categories: emotional harassment including both "verbal and physical" abuse, and non-emotional power trips. We hypothesize that not all power trip events are salient (Nigam and Ocasio, 2010) but depend on the components that constitute corporate harassment. According to Morgeson et al. (2015), the more disruptive and critical the event, the more likely it is to transmute or affect behaviors and other events. In addition, harmful talk stemming from negative emotions spreads more quickly than positive sentiments (Samson, 2006; Luo, 2009). Thus, we explore the negative interaction effects caused by emotional harassment on firms' values. Second, and in the same vein, we also suggest that the impact of power trips does not affect firms' value homogeneously; its consequences depend on the level of visibility. Visible firms tend to be subject to greater pressure to contribute more to corporate philanthropy than firms with lower visibility (Brammer and Millington, 2006; Gao and Hafsi, 2015). Moreover, if a renowned firm is the subject of an unexpected scandal, this quickly makes its way onto social media, such as Twitter and Facebook (Aula, 2010). Public anger that is rapidly spread through social media is soon reflected in market reactions. In short, most conglomerates operate in complex industry contexts and their visibility, created by the mass media, is higher than that of small firms. Therefore, the detrimental effect of power trips on conglomerates is much more salient for them than for less visible firms, such as SMEs.

This research provides an empirical contribution to the literature by extending the extant research on workplace bullying at the individual and team level to firm-level bullying through a simultaneous event study and regression analysis. The empirical findings in this study imply that firm-level power trips can be an important variable that affects the firm's value because they undermine the firm's credibility in the market. The current study also contributes to the literature on workplace bullying by applying and testing different management theories in the unique context of Korea. Event-oriented theories have been relatively rare, particularly when compared to feature-oriented theories (Morgeson et al., 2015). In addition, literature on the visibility of firms has been identified in the Korea context, and especially the chaebol environment. In the next section, we review the previous literature and develop our hypotheses. After detailing the methodology and analysis, the results are presented followed by the discussions of the study.

# THEORETICAL FRAMEWORK

# Effect of the Firm-Level Power Trip on the Firm Value in the Market

The consequences of workplace bullying among employees and teams have been well documented. This brings about counterproductive workplace behaviors (Fox and Spector, 2005) and abusive supervision is related to organizational deviance, which puts organizations at risk (Robinson and Bennett, 1995). But, a precise definition for workplace bullying by top-tier managers was not clearly developed in the literature and this, then, is remained as a very broad, inclusive, elusive term (Bartlett and Bartlett, 2011). Bartlett and Bartlett (2011) categorized work-related bullying as following three types; workload, work process, and evaluation and advancement. However, these classifications are based on within-firm analysis. Therefore, we replace workplace bullying with the terminology of "power trip" to further clarify our firm-level interactions. Power trip is called Gapjil in Korea; this term has a negative connotation to it. It is neologism made by combining the word Gap which is used to introduce the first party in a contract as well as refers to a party's superior and powerful status and Jil, a suffix for any actions that is negatively conducted. It is seemingly a similar concept, but these need to be distinct. Workplace bullying is based on hierarchy that reveals job position or relationships but, can even be caused by a subordinate. On the contrary, power trip can take place without being inside firm and the target may not be just employees. The motives of power trip often stem from wealth and social status in an authoritarian society. Korea has deeply been imbued with these thoughts which become a culture that has been followed by individuals. Therefore, an individual who was a victim in the past acts just like a perpetrator in the past when he or she is in the powerful position. As social malady, power trip in Korea is like a diving duck. It could be swimming on the surface or diving deep in the water, but whether it is seen or not, it is always there. This higher power (usually a manager or someone's boss) tends to go to their head causing them to "power trip" and abuse their rights as a manager/boss/owner. Such as picking on people or making their lives difficult, "Just because they can." is a person who is on a power trip. Bullying leadership is based on fear and control, but mainly on negative reinforcement and punishment. Negative reinforcement is an act required to avoid unpleasant and painful consequences, which forces employees to do their employers' work as instructed. Such leadership creates an oppressive culture in the workplace. In this type of work culture, relationships among employees are based strictly on vertical relationships and form a hierarchical relationship where their superiors shape and control the individuals' character. Employees lose self-esteem and self-confidence by succumbing in a servile fashion to the perpetrator's threats and abuse. This

can lower employee morale, undermine loyalty, and eventually reduce corporate productivity.

However, it is still questionable how markets react to the news of bullying by managers. This research provides two arguments concerning this issue. The first is based on the consumer's perspective. The existing research suggests that consumers prefer goods that are ethically produced, and this may affect market reactions. In other words, the authors argue that a company's good behaviors, which are reported by the mass media, create a general context for consumers' evaluations. For example, Creyer and Ross (1997) find a significant correlation between consumers' preference for a company's service and products and the level of the company's ethical performance. Murray and Vogel (1997) find that managers are more willing to purchase goods from ethical companies. Brown and Dacin (1997) find that consumers' preferences for a new product are associated with consumers' overall assessment of the company's CSR (Corporate Social Responsibility) activities. On the contrary, consumers also experience negative emotions regarding companies' unethical behaviors. For example, Choi and Lin (2009) find that consumers showed negative emotions such as anger, fear, worry, and disgust toward the Mattel Product crisis; they recalled which news was posted on online bulletin boards, and this led to a boycott campaign. In sum, prior research suggests that corporate ethnical behaviors unveiled by the mass media have a positive impact on consumers' assessment or willingness to buy, while unethical activities can have a negative effect on consumers' evaluations of products. The second argument is based on the investor's perspective. A power trip not only affects consumers but also directly undermines investors' confidence in the focal firm in a negative way. Kohut (1971) and Kernberg (1979) argue that power trips are a kind of narcissistic leadership caused by a mental disorder. The narcissism is defined as an inflated sense of self and privilege, grandiosity, and low empathy for others (O'Reilly et al., 2014); managers with this characteristic tend to make impulsive and risky decisions (Nevicka et al., 2011; Braun et al., 2015). Narcissistic leaders have a strong sense of self, but their self-concepts are unstable, so they are easy to anger when others make mistakes. In reference to a pathology, managers with narcissism tend to be more aggressive and more flawed (Bushman and Baumeister, 1998). Researchers in the upper echelon argue that characteristics of TMT can signal something about the organization's legitimacy, which influences investors' evaluations of the firm (Higgins and Gulati, 2006). The leads us to expect that managers exerting undue power over subordinates make narcissistic and risky decisions, and investors tend to undervalue this kind of firm; this premise yields the following hypothesis.

H1. The firm-level power trip has a negative effect on a firm's value.

# The Moderating Effect of Emotional Harassment

As power trips can occur at the individual, team, and organizational level, they can be classified into many different types. An abusive person can exercise power and control over the victim through physical, psychological, sexual, or financial abuse. In this study, we focus on "emotional" harassment including verbal and non-verbal harassment. On the one hand, the behaviors that simply abuse social superiority in the relatively upper class have potentially abusive power, but no emotions are involved (e.g., push-driven distribution sales, unfair labor practice, and "tunneling"). On the other hand, according to Keashly (1997) and Lutgen-Sandvik (2003), "emotional abuse is the term to articulate the hostile verbal and non-verbal behaviors that are not explicitly tied to sexual or racial content yet are directed at gaining compliance from others." For example, yelling or screaming, nagging, intimidation, aggressive eye contact, negative rumors, ridiculing someone in front of others, silent treatment, and emotional blackmail is examples of emotional abuse. Conceptually, the term "abuse" is cruel and violent treatment among people while the term "harassment" which is intended to trouble or annoy someone is persistent attacks and criticism causing worry and distress. In this respect, the latter word is more appropriate for this research because it is not on a one-off but persisting for the corporate heirs and executives to inflict physical and traumatic damage on their victims. Previous study finds that organizations must have proper processes and procedures to address workplace abuse because it is their financial self-interest to do as well as to protect their workers. (Koonin and Green, 2007). Existing studies on emotions have emphasized that they play an important role in interactions insofar as they concern attitudes, ideas, thoughts, and behaviors (Ekman, 1993; Côté, 2005; Kopelman and Rosette, 2008; Van Kleef et al., 2010). While the expression of negative emotions (e.g., anger, shame, guilty) generally makes social interactions difficult, expressing and amplifying positive emotions could be equally instrumental in shaping relationships with others as they facilitate goal attainment at work (Wong et al., 2013). Such effects are spread through what is called "emotional contagion" (Hatfield et al., 1993). Organizational and psychological researchers have investigated emotional contagion through a variety of field studies (Totterdell et al., 1998; Totterdell, 2000; Bartel and Saavedra, 2000; Barsade, 2002), which have shown that people respond heterogeneously to positive and negative emotions, while negative events are inclined to trigger stronger emotional and behavioral responses than positive or neutral events (Cacioppo et al., 1997). Barsade (2002) also found that unpleasant emotions tend to generate greater emotional contagion than pleasant ones, which is referred to as the ripple effect. Because anger is one of the most contagious emotions (Tavris, 1984; Mattila and Enz, 2002), anger derived from emotional harassment (e.g., abusive language and even assault) by a manager and revealed through media coverage will elicit a more intense response from consumers and investors than a non-emotional power trip. Decision-making is greatly affected by emotions (Luce, 1998; Ruth, 2001) and previous studies have explored how consumers use information on emotions to arrive at efficient decisions. Despite the importance of emotional contagion in making decisions, empirical research on emotional contagion is limited at the micro level. Thus, it is necessary to study how TMT events involving emotional harassment generate negative emotions among consumers and investors that affect a firm's value. In addition, according to event system theory,

not all events are homogeneous but tend to be heterogeneous (Morgeson et al., 2015). From this perspective, events can occur at any level, regardless of the individuals, teams, organizations, and environments involved. Thus, a strong individual event like a TMT's power trip could explain relationships between the event and the firm's value. Organizations confront numerous events every day; however, not all of them are influential (Nigam and Ocasio, 2010). Thus, each event could have different results depending on whether emotional harassment is involved. To identify whether emotional harassment is involved in firmlevel power trips, event system theory focuses on the crucial event feature in terms of its novelty, disruption, and criticality (Morgeson et al., 2015). First, novelty describes the extent to which an event is distinctive in comparison to past events, and thus stands for a new phenomenon (Lee and Mitchell, 1994; Morgeson, 2005). Second, disruptive events may deviate from normal routines (Zellmer-Bruhn, 2003). Disruptive events destroy entities outside of their stereotyped way of thinking (Morgeson et al., 2015). Finally, criticality indicates the degree of importance of the event (Morgeson and DeRue, 2006) and typically leads to additional analyses and changes (Vaara, 2003). Power trips by authorities lead to greater public outrage via emotional contagion than non-emotional power trips. In the same vein, negative word of mouth conveying negative emotions spreads more quickly than positive sentiments (Samson, 2006; Luo, 2009). In short, emotional harassment is more destructive because it is novel, disruptive, and critical compared to events that are emotionally neutral.

H2. The impact of firm-level power trips on a firm's value is negatively interacted when emotion harassment is involved.

# The Moderating Effect of Conglomerates

The importance of a firm's visibility and celebrity status as a source of competitive advantage is well documented in the organizational literature (e.g., Hall, 1992). Previous studies have argued that a firm's visibility and celebrity are related to the firm's performance and legitimacy (Deephouse, 2000; Roberts and Dowling, 2002). Organizations with many subsidiaries and considerable revenue are likely to be more visible. The more the media monitor the firm's behavior, the more likely their social activities are to be visible (Matten and Moon, 2008). According to Rindova et al. (2006) and Pfarrer et al. (2010), the firm's visibility is increased when the media highlight a firm as a protagonist in the corrupted institution (Rindova et al., 2006). Examples of highly visible firms in our sample include conglomerates, and especially chaebol (in Korea). These chaebols' unconventional and controversial actions have attracted media attention touting their distinctive cultures, charismatic leaders, and singular identities. In a country where workers are often expected to show unquestioning loyalty, cases have focused on immoral behavior by the rich and powerful, along with public indignation at the family-run conglomerates known as chaebol, which dominate South Korea's economy.

Existing studies suggest that larger firms tend to be more visible and more likely to pay attention to social responsibility; they may also suffer from damage that leads to a bad reputation (Udayasankar, 2008). This argument is consistent with Saiia et al. (2003), who found that highly visible firms tend to make larger philanthropic gifts. Highly visible firms are also subject to more pressure and scrutiny from the public to take practical steps and meet social needs (Campbell and Slack, 2006). Korean chaebols have been criticized for misbehavior. Since they became the pivot of the Korean economy, Koreans have considered it a measure of social success to get a job at a chaebol conglomerate but, simultaneously, they hold antichaebol sentiments about misbehaviors and owner risk. Using the catchphrase 'chaebol reformation' to improve its transparency and accountability, they argue that the chaebol focuses on accumulating wealth and inheriting the company. Because firms with higher visibility are under great pressure to contribute more to activities involving social responsibility than firms with lower visibility (Gao and Hafsi, 2015), merely symbolic donations are unacceptable and may severely damage the firm's reputation (Boiral, 2007; Lyon and Maxwell, 2011). As corporate philanthropy influences corporate perceptions for a variety of stakeholders, including investors, customers, suppliers, actual or potential employees, and the voluntary sector (Smith, 1994; Himmelstein, 1997; Saiia et al., 2003), a firm's responsiveness to its stakeholder environment plays an important role in the firm's value (Brammer and Millington, 2006). Chang (2003) conducted a study of 419 chaebol affiliates from 1986 to 1996 and found that major shareholders, who are also managers, can be valued by other shareholders. Thus, managers from a chaebol family who hold large numbers of shares tend to abuse the corporate power given by their ownership. These behaviors undermine the firm's ethical and social responsibilities. In this context, when owners or top managers of Korean chaebol behave unethically, power trips are more negatively evaluated by the media and stakeholders. This reasoning leads us to expect that conglomerates with high visibility will experience more negative consequences when power trips occur.

H3. The impact of firm-level power trips on a firm's value is negatively interacted for conglomerates.

# SAMPLE SELECTION AND RESEARCH DESIGN

# Sample Selection

We selected companies that were reported to the media as culpable of a power trip from those listed on the Korea Stock Exchange. In this study, the power trip variable was set based on the date of the first report on the media such as newspaper and news. For example, the nut rage incident was an air rage incident that occurred on December 5, 2014, at John F. Kennedy International Airport in New York City. Korean Air vice president Heather Cho, dissatisfied with the way a flight attendant served nuts on the plane, ordered the aircraft to return to the gate before takeoff. As another example, On May 8, 2013, Namyang Co., was criticized following an accusation that the firm had pressed local franchises to buy products for a long time, as well as a tape-recording of rough words to local owners. The

### TABLE 1 | Sample selection.

fpsyg-09-02333 December 1, 2018 Time: 14:8 # 5


Panel B: Number of companies reported to the media as companies guilty of power trips by year


coverage of each case quickly spread through the media after December 5, 2014 and May 8, 2013.<sup>2</sup> Since the news of power trip is rapidly spread through media such as SNS and news, the first news day was set as an event day and the stock return from day −1 to day +1 was observed based on the event day. The sample period is from 2013 to 2018. The information on these companies was collected through media press releases on the Internet and in newspapers. We obtained financial data from the KIS-VALUE, which provides the financial statements of all listed firms, and stock data from the Fn-Guide.

The sample selection process is summarized in **Table 1**. The sample used in the event study is a total of 31 firm-year observations. Specifically, there are 33 companies reported to the media as those said to be guilty of harassment. Among them, we conducted an event study on a total of 31 firm-year observations except for two companies belonging to KOSDAQ companies. Since 2015, there has been a significant increase in media coverage of corporate harassment and it is emerging as a social problem.

There are cases where the same company is reported as guilty of harassment several times within one accounting period. When the same company is reported in this way, it is included in the event study analysis, but is used only as one sample in the OLS regression analysis. OLS regression analysis was carried out on 21 firm-year observations excluded seven samples that were not available in the Fn-Guide and three companies reported as duplicate companies in the one accounting period. Each continuous variable was winsorized at the 1st and 99th percentiles to minimize the effect of outliers.

# Event Study Model

To test our hypotheses, an event study method was used.<sup>3</sup> Event studies are designed to measure the effect of an unanticipated event on stock prices. The day of the event (t = 0) is defined as the date when the company is reported to the media as a company in which an incident of bullying, and specifically a power trip has occurred. If there is no stock transaction for the company on the specified date, the first stock trading day after the media report is defined as the event date. The standard method is based on estimating a market model for each firm and then calculating abnormal returns. This study also uses a market model among the methods proposed by Brown and Warner (1985) to calculate the abnormal return (AR).<sup>4</sup> These ARs are assumed to reflect the stock market's reaction to the arrival of new information.

The method is as follows: The rate of return on the share price of firm i on day t is expressed as

$$R\_{i,t} = \alpha\_i + \beta\_i R\_{m,t} + \text{ } \tag{1}$$

where Ri,<sup>t</sup> is the rate of return on the share price of firm i on day t, and Rm,<sup>t</sup> is the rate of return on a market portfolio of stocks (such as the KOSPI index) on day t. In this study, α and β of equation (1) were estimated using the KOSPI equally weighted index (EWI) and the daily stock return on each firm.

We estimated the market model over 160 trading days, starting 5 days prior to the event day.<sup>5</sup> Estimated α and β are used to measure the excess return (AR) of each company from day −5 to day +5. The method is as follows in equation (2).

$$AR\_{i,t} = R\_{i,t} - \hat{R}\_{i,t} = R\_{i,t} - (\hat{\alpha}\_i + \hat{\beta}\_i R\_{m,t}) \tag{2}$$

where αˆ and βˆ are the ordinary least squares (OLS) parameter estimates obtained from the regression (1). Ri,<sup>t</sup> is the rate of return on the share price of firm i on day t, and Rm,<sup>t</sup> is the rate of KOSPI market return on day t.

After calculating the abnormal returns of individual firms using the market model, the following equation (3) was used to calculate the average abnormal return (AAR) for the entire firm reported by the media to have experienced an incidence of bullying or a power trip.

$$AAR\_{i,t} = \sum\_{i=1}^{N} AR\_{i,t} \times \frac{1}{N} \tag{3}$$

<sup>2</sup> In the example above, Korean Air is a conglomerate firm. On December 5, 2014, the incident occurred in a case where the person uttered a speech with a profanity. Therefore, the LARGE30 variable is 1 and the DEMOTION variable is 1. On the other hand, Namyang Co., is a non-conglomerate firm. On May 8, 2013, the incident occurred due to forced sales to subcontractors. Therefore, the LARGE30 variable is 0 and the DEMOTION variable is 0.

<sup>3</sup>The event study method in accounting has been used in numerous empirical studies since Fama et al. (1969) analyzed the disclosure effect of stock splits for an event study. The first step in the event study is to define the events to be studied and determine the duration of the events. An event period is a period of time when specific information is considered to have an impact on the stock price. In general, the event period is set up according to the event date of the specific information, and includes the before and after periods. The reason for including the pre-event date in the event period is that there is a possibility that the specific information is known to the market before the event date. On the other hand, the reason for inclusion in the event period after the event date is to take into consideration the case where information is disclosed after the stock market is closed.

<sup>4</sup>Brown and Warner (1980, 1985) found that the constant mean-return model performs as well as more sophisticated approaches.

<sup>5</sup> In the previous studies (Bruner and Simms, 1987; Rutherford, 1990; Dickinson et al., 1991; Agrawal and Kamakura, 1995), the analysis was conducted using 160 trading days.

where N is the number of events being studied among multiple firms, AR is the abnormal return of firm i in period t, and AAR is the average abnormal return in period t.

In equation (3), the AARs are calculated. The cumulative average abnormal return (CAR) is calculated as follows when the AAR is accumulated during the event window. Using CAR, it is possible to analyze how the firm's harassment behavior affects individual stock prices.

$$\text{CAR}\_N(t\_1, t\_2) \;= \sum\_{t=t\_1}^{t\_2} \text{AAR}\_t \tag{4}$$

where t<sup>1</sup> and t<sup>2</sup> are the initial and final dates of the event window, AAR is the average abnormal return in period t, and CAR is the cumulative AAR in period t.

# Regression Analysis Model

Hypotheses 1, 2, and 3 are verified by setting equations (5), (6), and (7) with the cumulative abnormal return before and after the event day as dependent variables. The method of using the CAR after the day before the event day has the advantage of mitigating the bias that may arise when analyzing only the return on the event day (t = 0). We set CAR as a dependent variable for 3 days from −1 day before the event day to 1 day after the event day. The cumulative abnormal returns over 3 days before and after the event day are a universal measure of stock returns used in event studies.

$$\text{CAR} \begin{aligned} (-1, 1)\_{i,t} &= \ \alpha\_0 + \beta\_1 \text{SPD} U M\_{i,t} + \beta\_2 \text{SIZE}\_{i,t} + \beta\_3 \text{LEV}\_{i,t} \\ &+ \ \beta\_4 \text{OCF}\_{i,t} + \beta\_5 \text{ROA}\_{i,t} + \beta\_6 \text{GROWTH}\_{i,t} \\ &+ \varepsilon \end{aligned} \tag{5}$$

$$\text{CAR} \ (-1, 1)\_{i, \ell} = \ \alpha\_0 + \beta\_1 \text{SPDUM}\_{i, \ell} + \beta\_2 \text{SPDUM}\_{i, \ell}$$

$$\begin{aligned} & \times \text{DEMOTON}\_{i, \ell} + \beta\_3 \text{SIZE}\_{i, \ell} + \beta\_4 \text{LEV}\_{i, \ell} \\ & + \beta\_5 \text{OCF}\_{i, \ell} + \beta\_6 ROA\_{i, \ell} + \beta\_7 \text{GROWTH}\_{i, \ell} \\ & + \varepsilon \end{aligned} \tag{6}$$

$$\text{CAR} \,(-1, 1)\_{i,t} = \,\alpha\_0 + \beta\_1 \text{SPDUM}\_{i,t} + \beta\_2 \text{SPDUM}\_{i,t}$$

$$\times \,\text{LARGE30}\_{i,t} + \beta\_3 \text{LARGE30}\_{i,t} + \beta\_4 \text{SIZE}\_{i,t}$$

$$+ \beta\_5 \text{LEV}\_{i,t} + \beta\_6 \text{OCF}\_{i,t} + \beta\_7 \text{ROA}\_{i,t}$$

$$+ \beta\_8 \text{GROWTH}\_{i,t} + \varepsilon \tag{7}$$

where CAR (−1,1) is the cumulative abnormal returns over 3 days from −1 day before the event day to 1 day after the event day. SPDUM is a dummy variable equal to 1 if companies were reported to media as firms charged with harassment, and 0 otherwise. DEMOTION is a dummy variable equal to 1 if companies were reported to the media as guilty of emotional harassment, and 0 otherwise. LARGE30 is a dummy variable equal to 1 if companies are conglomerate firm groups, and 0 otherwise. SIZE is computed as the natural logarithm of the total assets of the firm. LEV is defined as liabilities over equity. OCF is the operating cash flow scaled by lagged total assets. Return on assets (ROA) is net income divided by lagged total assets and GROWTH is growth of sales in year t.

Equation (5) is a model designed to test hypothesis 1 that suggests that the firm-level power trip has a negative effect on the firm's value in the market. The firm's harassment refers to the unfair conduct of an opponent who is in a favorable position in the firm. If the firm's harassment is perceived as bad news in the capital market, the firm's stock price will fall. Hypothesis 1 is supported if the coefficient β1 on the experimental variable SPDUM shows a significantly negative value.

The firm's harassment behavior can be divided into emotional and non-emotional harassment. Equation (6) is a model designed to test hypothesis 2, which suggests the capital market's response to the firm's harassment is different for emotional harassment. In equation (5), the experimental variable is interaction term (SPDUM × DEMOTION). DEMOTION is a dummy variable equal to 1 if companies were reported to media as guilty of emotional harassment, and 0 otherwise.<sup>6</sup> Hypothesis 2 is supported if the coefficient β2 on the interaction term (SPDUM × DEMOTION) shows a significantly negative value.

Equation (7) is a model designed to test hypothesis 3, which suggests the capital market's response to the firm's harassment is different if it is a conglomerate firm. In equation (7), the experimental variable is interaction term (SPDUM × LARGE30).<sup>7</sup> Hypothesis 3 is supported if the coefficient β3 for the interaction term (SPDUM × LARGE30) shows a significantly negative value.

When the number of samples is small, reliability may be questioned in regression analysis. Therefore, the propensity score matching (PSM) method is used for analysis. In this study, a "1: 5" matching was conducted based on the media reports of the firm's harassment, and OLS regression analysis was performed on a total of 126 firm-year observations.

The control variables of this study are as follows. SIZE is added to control the firm's size. The larger the firm's size, the more information there is about the company in the capital market (Fama and MacBeth, 1973). LEV is defined as liabilities over equity. In the capital market, the higher the liability, the higher the firm's risk. The risk due to the debt ratio is added as a control variable because it may affect the ARs. Companies with a high cash flow ratio in operating activities are more likely to be affected by bad news because they are recognized as having superior ability to respond to crises. The ROA is added to control the impact of an individual firm's performance on the excess return on the capital market. The better the company's performance, the more likely it is that the company will achieve high excess returns in the capital market. GROWTH was included to control the growth of the enterprise.

<sup>6</sup>The firm's harassment behavior is revealed to the capital market through the media (e.g., the Internet, news, etc.). In addition, we analyze whether the emotional harassment occurred by analyzing the reported media. The DEMOTION variable was not included in equation (6) because it cannot distinguish whether the emotional harassment that is not reported in the press is indicative of a firm's harassment behavior.

<sup>7</sup>Conglomerate firms are often called chaebol firms. The definition used to identify chaebol firms is that of a large business group established by the Korea Fair Trade Commission (KFTC) of which more than 30% of the shares are owned by the group's controlling shareholders and its affiliated companies.

# EMPIRICAL RESULTS

fpsyg-09-02333 December 1, 2018 Time: 14:8 # 7

# Market Reaction to Firm's Harassment

**Figure 1A** shows the average abnormal return (AAR) and cumulative average excess return rate (CAR) from −5 to +5 days. The AAR is showing a sharp decline from −1 to +1 days, and CAR is beginning to decrease from −1 day before the event day but is showing a trend of repeating ups and downs.

Specifically, the AAR was −0.008 (t = −1.87, p-value = 0.071) at −1 day before the event day and −0.006 (t = −2.12, p-value = 0.042) at event day. CAR shows negative value from −1 day before the event day but shows a tendency to fall sharply at −0.013 (t = −1.68, p-value = 0.104) at event day. This result supports hypothesis 1 that the firm-level power trip has negative effects on a firm's value in the stock market.

**Figure 1B** is the result of analyzing AAR and CAR before and after the event day after classifying the sample as either an emotional harassment group or a non-emotional harassment group. The first column in **Figure 1B** is the analysis of the emotional harassment group (N = 12) and the second column is the analysis of the non-emotional harassment group (N = 19).

According to the results of the analysis, the drop in AAR is relatively increased at the event day in the emotional harassment group. Specifically, the AAR of the emotional harassment group was −0.017 (t = −1.97, p-value = 0.074) at −1 day before the event day, −0.010 (t = −1.70, p-value = 0.117) at event day. While the AAR of the non-emotional harassment group was −0.002 at −1 day before the event day and −0.003 at event day, but there was no statistical significance found. These results are interpreted to mean that the capital market responds differently if emotional harassment is reported.

**Figure 1C** is the result of analyzing AAR and CAR before and after the event day after classifying the sample into conglomerate firm groups and non-conglomerate firm groups according to the definition of a conglomerate firm. The first column in **Figure 1C** is the analysis of the conglomerate firm groups (N = 12) and the second column is the analysis of the non-conglomerate firm groups (N = 19).

While the non-conglomerate firm group was statistically significant in the AAR during the event window, the conglomerate firm group was not statistically significant in the AAR in the same period. Specifically, the AAR of the conglomerate firm group was −0.001 at −1 day before the event day and −0.002 at event day, but this was not statistically significant, while the AAR of the non-conglomerate firm group was −0.012 (t = −2.00, p-value = 0.061) at −1 day before the event day and −0.008 (t = −2.06, p-value = 0.054) at event day. These results suggest that the capital market responds differently to events in the type of conglomerate firms. Please refer to **Appendix 1** for specific figures for the abnormal returns for windows surrounding the event day.

**Table 2** shows the results of the t-test used to analyze the impact of the firm's harassment on the capital market reaction by cumulative abnormal returns. The sample was analyzed by dividing it into emotional and non-emotional harassment groups, and conglomerate and non-conglomerate firm groups.

The cumulative abnormal returns from the firm's reported harassment show −0.006 (t = −2.12, p-value = 0.042) in CAR (0,0) and −0.012 (t = −1.85, p-value = 0.074) in CAR (−1,1), indicating that they were significant during the relatively shortterm verification period. The results of dividing the sample into emotional and non-emotional harassment groups are as follows. In the emotional harassment group sample, the cumulative abnormal returns were −0.024 (t = −1.93, p-value = 0.080). Statistical significance was found only in this group.

The results of dividing the sample into conglomerate and non-conglomerate firm groups are as follows. In the nonconglomerate firms, the CARs are significantly negative during the entire verification period. However, there was no statistical significance found for CAR in conglomerate firms. This means that in non-conglomerate firms, CAR continued to decline to negative values after the event day.

The event study method is useful for directly analyzing the impact of specific information on the short-term capital market by excluding the influence of other factors, other than the experimental variables, on the stock price. However, the method is limited by the fact that it excludes corporate characteristics that could affect the firm's abnormal returns. In this study, we conducted additional OLS regression analysis to include the financial factors that affect the firm's abnormal return.

# Descriptive Statistics and Correlations

In this study, an OLS regression analysis was performed after constructing a "1: 5" control group by propensity score matching analysis. **Table 3** provides descriptive statistics for the variables used in the regression analysis, while **Table 4** shows the results of the analysis of difference of the major variables according to the firm's harassment.

The standard variable of the difference analysis, SPDUM, is a dummy variable having a value of 1 if companies were reported by the media as harassment firms, or 0 otherwise. The mean value of SPDUM was 0.167. In other words, 21 firm-year observations of the full sample were firms reported by the media to have engaged in harassment.

Panel A of **Table 3** presents the descriptive statistics for the entire sample, Panel B shows the descriptive statistics for the harassment firm (SPDUM = 1) group, and Panel C shows the descriptive statistics for the non-harassment firm (SPDUM = 0) group. In the entire sample, the dependent variable, CAR, was 0.002, which is not significantly different from 0, but the CAR of the SPDUM = 1 group is −0.011, which was lower than the CAR of the SPDUM = 0 group. This result supports hypothesis 1: the abnormal return of the companies reported by media to engage in harassment is low.

In the SPDUM = 1 group, the mean value of the DEMOTION was 0.429. In other words, 9 (= 21 × 0.429) companies were reported to the press as engaging in emotional harassment. LARGE30, the conglomerate firm group, was 0.571 in both SPDUM = 1 and SPDUM = 0 groups.

The mean (median) value of SIZE in the entire sample was 29.303 (29.614). The mean value of the debt-to-equity ratio (LEV) was 1.244, which reflects a sample of firms that are relatively sound financially. The mean (median) value of the operating cash

flow (OCF) was 0.073 (0.070) and the mean (median) value of ROA was 0.030 (0.022). The mean (median) value of GROWTH, or the growth potential of the company, was −0.004 (−0.007).

**Table 4** shows that there is no statistically significant difference between the treatment group and the control group in firm size (SIZE), debt ratio (LEV), operating cash flow (OCF), return on assets (ROA), or growth rate of sales (GROWTH). This is because when the propensity score matching method is considered, logit analysis is performed with the firm size, debt ratio, operating cash flow, total assets profit rate, and sales growth rate. In other words, the absence of statistical significance between the control variables in the analysis of difference means that the control group was appropriately selected. In that analysis, the mean value of CAR in the treatment group (−0.011) was significantly lower than the mean value of CAR in the control group (0.004).

**Table 5** presents Pearson correlations among variables used in the main analyses. The correlation of SPDUM and CAR is negatively significant at 5%. The negative correlation between

### TABLE 2 | Cumulative abnormal returns for windows surrounding the event day.


Panel A: Results of dividing power trips into emotional and non-emotional harassment groups

Panel B: Results of dividing power trips into conglomerate and non-conglomerate groups


Variable definitions: AAR, the average abnormal return in period t; CAR, the cumulative average abnormal return in period t <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

### TABLE 3 | Descriptive statistics.


Variable definitions: CAR (−1,1), cumulative abnormal returns over 3 days from −1 day before the event day to 1 day after the event day; SPDUM, a dummy variable equal to 1 if companies were reported to media as harassment firms, and 0 otherwise; DEMOTION, a dummy variable equal to 1 if companies were reported to media as emotional harassment, and 0 otherwise; LARGE30, a dummy variable equal to 1 if companies are conglomerate firm group, and 0 otherwise; SIZE, the natural logarithm of the total assets of the firm; LEV, liabilities over equity in year t; OCF, the operating cash flow scaled by lagged total assets; ROA, net income divided by lagged total assets; GROWTH, growth of sales in year t.

### TABLE 4 | Differential analysis results for PSM matching firms.

fpsyg-09-02333 December 1, 2018 Time: 14:8 # 10


See Table 3 for variable definitions. <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

TABLE 5 | Pearson correlations.


See Table 3 for variable definitions; coefficients shown in bold are significant at p < 0.05 (two-tailed test).

SPDUM and CAR suggests that a company reported to the media as engaging in power trips is associated with a negative stock return. There is a significant negative correlation between DEMOTION, which indicates whether the power trip involves emotional harassment, and CAR. Therefore, the first hypothesis gains support from the negative correlation between SPDUM and CAR.

Also, the negative correlation between DEMOTION and CAR supports the second hypothesis which suggests that emotional harassment rather than non-emotional harassment will show more negative market reactions. The correlation between LARGE30 and SIZE is 0.59 has high positive significance in this regard. Among control variables, we find that few correlations are very high. For example, the correlation between OCF and GROWTH is 0.60, while that between LEV and ROA is −0.46. The variance inflation factor (VIF) for independent variables is 2.538 less than 10, which means that there is not a serious multi-collinearity problem.

# REGRESSION RESULTS

# Results for Market Reaction of Firm-Level Power Trips (H1)

**Table 6** presents the results for the test of our first hypothesis from the regression analysis based on equation (5). In **Table 6**, hypothesis 1 is supported if coefficient β<sup>1</sup> shows a significantly negative value. The results of the analysis show that the regression TABLE 6 | Market reaction of firm-level power trip.


See Table 3 for variable definitions. <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

coefficient of SPDUM was −0.016 (t = −1.89, p-value = 0.061). These results suggest that participants in the capital market negatively perceive the press reports on the firm's harassment behavior.

# Results for Market Reaction by Emotional Harassment (H2)

**Table 7** presents the results for the test of our second hypothesis from the regression analysis based on equation (6). In **Table 7**, hypothesis 2 is supported if coefficient β<sup>2</sup>

### TABLE 7 | Market reaction by emotional harassment.

fpsyg-09-02333 December 1, 2018 Time: 14:8 # 11


See Table 3 for variable definitions. <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

TABLE 8 | Market reaction by conglomerate.


See Table 3 for variable definitions. <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

in the interaction term (SPDUM × DEMOTION) shows a significantly negative value. The result of the analysis shows that the regression coefficient of SPDUM × DEMOTION was −0.026 (t = −1.70, p-value = 0.093). This result implies that capital market participants perceive media reports on emotional harassment as more negative than media reports on nonemotional harassment.

# Results for Market Reaction by Conglomerate (H3)

**Table 8** presents the results for the test of our third hypothesis from the regression analysis based on equation (7). SPDUM is a regression coefficient that indicates the capital market response of media reports on a firm's harassment behavior in non-conglomerate firms (LARGE30 = 0). The regression coefficient on SPDUM is −0.025 (t = −1.98, p-value = 0.050). In the non-conglomerate firm group, if the firm's harassment behavior is reported by the media, it indicates that the company will have negative ARs. In hypothesis 3, we expect that the information use's negative response in the conglomerate firm group would be greater than the information use's negative response in the non-conglomerate firm group if the firm's harassment behavior is reported by the media. The regression coefficient of SPDUM × LARGE30, which is the experimental variable of hypothesis 3, is 0.017 (t = 0.98, p-value = 0.331), which is not significant. In the conglomerate firm group, if the firm's harassment behavior is reported by the media, there is no effect on ARs. The sum of the regression coefficients of SPDUM and SPDUM × LARGE30 was −0.008 (= −0.025 + 0.017), which was not statistically significant.

This result is not consistent with hypothesis 3 that the effect of a firm-level power trip on a firm's value in the market is more salient for the conglomerate firms' group than for the nonconglomerate firms' group. These results reflect the perception of market participants that conglomerate firms will not fail because they enjoy a competitive edge in capital, labor, and products. In addition, the results of **Table 8** are consistent with the results of the event study analysis in which the CAR variable of non-conglomerate firms is negative in all periods, while the CAR variable of conglomerate firms is not significant in **Figure 1C**.

# DISCUSSION

This study examined the effects of power trips by top managers on the firm's valuation in the market. We found that power trips revealed by the mass media have significant and negative effects on a firm's value. We also found that this negative impact is moderated by the type of power trip. This study contributes to the literature in several ways. First, it complements the existing literature by explaining how markets evaluate unethical behavior by top managers. Although many studies have highlighted the malfunctioning and counterproductive workplace behaviors caused by power trips within the firm, few have explored the firm-level consequences of power trips. The issue of power trips should be considered at both the micro- (i.e., the individual and team level) and the macrolevel (of the organization or firm). This study is an effort to fill gaps in the research by exploring the consumers' and investors' perspectives. Second, this study suggests the importance of emotion in the analyses of power trips based on emotional contagion and event system theory. The role of emotions in both the service encounters, where positive or negative emotions spread from employee to consumer, and in media coverage, where negative emotions spread from the emotional harassment event by TMTs to consumers, is still important. The event system theory assumes that events are not homogeneous but heterogeneous. The strength of an event evolves systemically in organizations across space and time. Events cannot be treated as discrete and isolated matters of concern at only one level. Thus, power trips are disruptive events that force organizations out of their conventional response

modes. Prior research highlights how employees experience emotional events at work, while affective event theorists have considered how workplace events affect employees' emotions (Weiss and Cropanzano, 1996; Weiss et al., 1999). In this study, we have explored how markets estimate the news of emotional harassment. Our finding is aligned with previous studies of leadership theories. Humphrey et al. (2016) emphasize the important role of emotional factors and represent the usefulness of merging research on leadership and emotions. Empirical studies have also shown that a leader's emotional expressions are more important than the content of the message (Newcombe and Ashkanasy, 2002) because leaders influence follower's attitudes, cognitions, affective states, and behavior (Koning and Van Kleef, 2015). In this vein, we suggest that emotional power tripping by leaders leads to detrimental consequences. Third, our research provides a unique hypothesis of power trips in which effects are contingent on the firm's visibility. However, this hypothesis is not supported, and this can be explained by the following reason. As most conglomerate firms are not businesses that directly deal with consumers, the information that conglomerate firms are reported to the press as firms guilty of harassment may not significantly affect the market participants' judgment on the future profitability of the company. On the other hand, non-conglomerate firms have many substitute goods and often deal directly with consumers. Therefore, if media reports precipitated by firms' harassment behaviors (unethical behavior) are spread by consumer boycotts, they will directly impact the profitability of the company. Thus, the negative relationship between the firm's harassment behaviors and abnormal returns is not apparent in the conglomerate firms, but the relationship for non-conglomerate firms is significantly negative. In fact, Namyang Co., which was reported to the media as a conglomerate firm guilty of harassment, reported a drop in operating profits by 87.7%.<sup>8</sup> This evidence is consistent with Korean investors believing the very largest conglomerate firms are "too big to fail." The advantageous position of conglomerates in Korean context

# REFERENCES


is widely recognized in the literature from angles of government (Lee et al., 2002) and networking capabilities (Kim, 2005). This is also in consonance with Agarwal et al. (2002)'s study. Her empirical work demonstrates that largeness lowers mortality rates by improving firms' ability to shield themselves from uncertain environments. And the idea of liability of smallness indicates that mortality rates decline with increased size (Hannan and Freeman, 1984). Empirical regularities, in economics, also show a positive relationship between firm size and survival for any given growth rate (Sutton, 1997). This is because the benefits of greater market power (Bain, 1956) and the minimum efficient scale (Mansfield, 1962; Jovanovic, 1982). Thus, the idea that bigger is better contributes to the lower failure rate among conglomerates.

The study has limitations. The empirical analysis was based on listed firms in a single country—Korea—and therefore its generalizability may be limited. The unlisted business in an informal economy shows different behaviors with respect to power trips and thus the consequences will not be consistent with those of listed firms. The causes and outcomes of power tripping can also be found in the cultural context, the social climate, and in psychological factors. According to Hofstede's cultural distance index, Korean culture is known for its high-power distance, high uncertainty avoidance, and short-termism. The Korean decisionmaking process is also heavily influenced by hierarchies. Such a cultural context makes Korean society a place conducive to power trips because people can easily draw on superior power to justify unethical or aggressive behaviors. In this regard, future research could provide valuable insights into the embeddedness of firm behaviors in various cultural and institutional environments.

# AUTHOR CONTRIBUTIONS

Sj-C devised the project, the main conceptual ideas, and was in charge of overall direction and planning. YC developed the theoretical framework and took the lead in writing the manuscript. HP contributed to the design and implementation of the research and to the analysis of the results.

# FUNDING

This work was supported by the research fund of Hanyang University (HY-2016).


<sup>9</sup> In May 2013, Namyang Co., was criticized following an accusation that, for a long time, the firm had strongly pressed local franchises to buy its products, as well as a tape-recording of rough words to local owners. Namyang Co., caused more than about 18 million dollars as of July 2018 of damage through illegal activities such as high-pressure sales and shifting wages of salesmen to local franchises for the past seven years. Angry over the behavior at Namyang Co., consumers started to boycott its products.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Choi, Park and Choi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

### APPENDIX 1 | Abnormal returns for windows surrounding the event day.

### Panel A: Abnormal returns of stocks concerning reported power trip


Panel B: Abnormal returns of stocks on emotional harassment and non-emotional harassment


### Panel C: Abnormal returns of stocks on conglomerate and non-conglomerate firms


Variable definitions: AAR, the average abnormal return in period t; CAR, the cumulative average abnormal return in period t; <sup>∗</sup> , ∗∗ , ∗∗∗ indicate significance at the 10, 5, and 1% levels, respectively (two-tailed).

# The Effect of Emotional Labor of College Administrative Service Workers on Job Attitudes: Mediating Effect of Emotional Labor on Trust and Organizational Commitment

### Sang-Lin Han<sup>1</sup> \*, Hyeon-Sook Shim<sup>2</sup> and Won Jun Choi<sup>1</sup>

<sup>1</sup> School of Business, Hanyang University, Seoul, South Korea, <sup>2</sup> Department of Business Administration, Baewha Women's University, Seoul, South Korea

Service providers working for a service organization are asked to express such positive emotions as joy, pleasure, and politeness required at the organizational level rather than their natural emotions they are experiencing at the moment. They cannot express their emotion they are actually going through and accordingly, their level of emotional labor and emotional dissonance influence on their job commitment and trust toward their organization. This study thus set out to investigate the effects of leading variables of emotional labor on the level of emotional labor and the impact of emotional labor on organizational trust and organizational commitment with a subject group of college administrative staffs. Three underlying factors such as job stress, intimacy, and professionalism were identified as the determinants of emotional labor. Based on the conceptual background and our research questions, five research hypotheses and the proposed research model regarding the effects of emotional labor on organizational commitment and trust were developed. We also tried to include the moderating effects of work environment and gender of service providers on the research model. Given those findings, this study offers theoretical implication that confirms the negative results of emotional labor. Unlike many different studies on emotional labor from the traditional perspective of service, this study offers a practical implication by expanding and applying it to the field of college administrative service, which is an area where service providers are in the different working environment from the traditional company work environment. Finally, the managerial implications and the limitations of the study were also discussed.

Keywords: job stress, intimacy, professionalism, emotional labor, organizational commitment, trust

# INTRODUCTION

Workers working at service organizations are often required to express positive emotions such as joy, pleasure, and politeness (Sutton and Rafaeli, 1988). In particular, in the case of the service industry, emotional expressions of workers are very important because human contacts are frequently made due to the nature of services. The impressions for a few seconds made through the facial expressions and emotions shown by service workers to

### Edited by:

Weon Sang Yoo, Korea University, South Korea

### Reviewed by:

Jaewon Yoo, Soongsil University, South Korea Seong Ho Lee, Hanbat National University, South Korea

> \*Correspondence: Sang-Lin Han slhan@hanyang.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 14 September 2018 Accepted: 21 November 2018 Published: 14 December 2018

### Citation:

Han S-L, Shim H-S and Choi WJ (2018) The Effect of Emotional Labor of College Administrative Service Workers on Job Attitudes: Mediating Effect of Emotional Labor on Trust and Organizational Commitment. Front. Psychol. 9:2473. doi: 10.3389/fpsyg.2018.02473

**41**

customers in the process of service production can be directly connected with the profit of the service company. Nevertheless, there was a lack of interest in emotion studies in the field of service marketing, and interest in emotions in marketing has been mainly shown from the viewpoint of customers. That is, studies have mainly focused on the effects of the emotions experienced by customers in the process of purchasing behavior (Babin and Attaway, 2000; Schaubroeck and Jones, 2000) or how workers should be educated for customer emotion management (Menon and Dubé, 2000; McLaughlin and Carr, 2005). After the studies conducted by Hochschild (1983, 1990), which defined service workers' efforts to control their actual emotions and express certain emotions that seem to be desirable in the service industry as emotional labor, studies on service workers' emotions have been actively conducted (Grandey, 2000; Kruml and Geddes, 2000).

Nonetheless, studies on the emotional labor of service workers in the field of administrative services are insufficient. In particular, despite that customers' evaluation of school administrative services may be more conservative in that the services are delivered to school members, school units' interest in service workers' emotion management can be said to be insufficient.

In addition, in the case of administrative services, following the global stream of administrative reform since the 1990s, workers working at service organizations have been required to express mainly those emotions such as joy, pleasure, and politeness (Sutton and Rafaeli, 1988) because the emotional expressions of the service workers positioned at the front lines of service delivery have great effects on the service quality perceived by customers and overall perceptions of the service organizations by the customers. Therefore, in recent years, even public institutions have been regarding emotions as important and have been making effort to break from the existing authoritative forms and derive changes in behavior into the shape of emotional and customer-oriented service suppliers.

However, in the case of service workers who deliver services at the front lines, the situations where they perceive a sort of emotional labor of being unable to express their actual emotions may lead to stress imposed while they work, which may reduce their trust in their organizations or job satisfaction. Eventually, the core of the efforts of workers at educational institutions and administrative service workers can be said to be improving the quality of service and building customer-oriented organization structures and the accompanying work overload may be a leading variable of emotional labor.

As the emotional labor of service providers is regarded as important at service organizations (Adelmann, 1995; Kruml and Geddes, 2000), the present study was intended to investigate the emotion management of service workers in the field of administrative services, which has been studied relatively less extensively thus far (Diefendorff et al., 2005). In particular, the present study was conducted with service workers working at universities with a view to proposing theoretical or practical measures to improve the emotion management of service workers.

The present study aimed to investigate the effects of administrative service workers' emotion management on the organizational effectiveness perceived at the level of organization dimensions. More concretely, the present study was intended to review variables such as relational effectiveness (job stress, intimacy, and professionalism), emotional labor, and organizational effectiveness (organizational commitment, organizational trust), which are core variables in this study, and examine the causal relations among these variables.

# CONCEPTUAL BACKGROUND AND RESEARCH HYPOTHESES

# University Administrative Staff's Functions and Roles

Although university administrative staff members do not directly engage in educational or research activities, they not only support those activities to ensure that the activities are properly conducted but also play core roles of planning and managing all works in universities. In addition, they not only simply assist professors or students but also are in charge of professional administrative management such as arranging and maintaining all conditions for achievement of the goals of university education. University administrative staff members play the role of a bridge between professors and students while playing pivotal roles in the university administration. Teaching activities are direct instructional activities to achieve the educational goals of school organizations. In contrast, university administration can be said to be service activities intended to support teaching activities and achieve educational goals.

# Emotional Labor

# The Concept of Emotional Labor

Emotional labor is the concept termed emotional work applied to the situation of job performance in organizations. It can be said to be a concept similar to emotional regulation in that it refers to the effort to intentionally express certain emotions beyond suppressing or controlling certain emotions. However, whereas emotional regulation is individuals' intentional behavior to change an excessive positive or negative emotional state into the targeted state of emotions expected by themselves or others, emotional labor is more other-oriented and can be said to be a part of emotional regulation appearing in the process of job performance.

The individuals' efforts to control their actual emotions and express emotions suitable for job performance in organizations as such were conceptualized into emotional labor. In this respect, Hochschild (1983) defined emotion labor as "outward states or facial expressions resulting from induced or suppressed emotions different from actual emotions."

Based on the argument of Ashforth and Humphrey (1993) and Morris and Feldman (1996) presented leading variables of emotional labor and a multidimensional concept of emotional labor for the results of the variables. Unlike the existing viewpoints, they focus on behaviors to express organizationally

desirable emotions with regard to the concept of emotional labor and conceptualized emotional labor based on the diverse dimensions of behaviors to express emotions required by organizations. In this study, the researchers stated that emotional labor consists of four dimensions, which are the frequency of emotional display, the degree of attentiveness required for emotional expressions, the diversity of emotional expressions, and emotional dissonance.

In addition, Grandey (2000) defined emotional labor as a process of expression and regulation of emotions intended to achieve the organizations' goals. Through this process, rather than focusing on occupational classification, observable expressions of emotions, or the disagreement between the characteristics of the situation and emotions, the internal emotional regulation approach is used more frequently. As such, the concept of emotional labor varies slightly among researchers and no agreed concept is available even now.

### Emotional Labor in University Administrative Services

Recently, schools and public institutions have been making considerable efforts in activities for customer satisfaction and kindliness to customers such as implementing kindliness education and recruiting kind staff members. These changes can be regarded as indicating that the emotional labor of employees, which had been receiving attention mainly in service positions in the private sector, has become an important form of labor in the public sector too. Although emotional labor began in the form of kind behaviors demanded from administrative service workers in departments related to students mainly organized for civil services in the early days, it appears in all school personnel now along with the transparency, accountability, and information disclosure of individual departments.

School personnel are implicitly required to be receptive, tolerant, understanding, and willing to help so that they can provide services that meet the needs of students while constantly coming into contact with students. Therefore, when they come up with pessimistic feelings, anger, or unkindness about students in the course of delivering administrative services, they make great efforts in emotional regulation in order to suppress such feelings and induce receptive and friendly emotions. The labor for emotional management as such may lead to the feeling of exhaustion as much as does physical work. This psychological aspect as such will greatly affect the intimacy, professionalism, job stress, and organizational commitment of school personnel thereby acting as a factor that affects organizational effectiveness.

### Previous Studies on Emotional Labor

Hochschild (1983) indicated that emotional labor is directly related to negative behaviors such as drug abuse, alcoholism, and absenteeism and the magnitude of demand for suppression of emotions that must be experienced due to duties was shown to be positively correlated with job stress while being negatively correlated with job satisfaction.

Abraham (1999) conducted a study on customer services (food service industry, telemarketer, apparel retailer, and event company employee) selected from the service industry and indicated that according to the results of the study, job satisfaction (dissatisfaction) mediates the relationship between emotional dissonance and turnover intentions. In addition, job satisfaction, which is affected by emotional dissonance, was shown to be closely related with turnover intentions.

Pugh (2001) studied the process through which the happy emotions expressed by staff members of banks affect the positive emotions of customers, who visit banks, and the positive emotions as such ultimately lead to good perception of the service of banks. Zapf (2002) conducted a study to find what the positive and negative effects of emotional labor on employees who were performing emotional labor were and the results of this study indicated that the positive effect of emotional labor was job satisfaction and the negative effect was emotional exhaustion.

Gountas et al. (2007) examined the effects of flight attendants' emotional expressions on the emotions of customers as airline passengers. The results of this study indicated that flight attendants' positive emotional expressions toward customers affected the positive emotions of customers and the positive emotions of customers eventually affected the customers' overall satisfaction and reuse intentions in relation to the use of the airline.

As such, studies regarding emotional labor have been applied to diverse fields, and discussions on the effects of emotional labor have still produced both positive and negative results. In particular, despite that diverse studies have been conducted in relation to emotional labor, such studies are still insufficient in the field of medical service. The present study was intended to focus on the negative consequences of emotional labor presented by Hochschild (1983).

# Job Stress

Thus far, many scholars have diversely presented concepts related to stress in various fields. They defined stress as an inadequate relationship between environments and individuals and regarded that such an inadequate relationship originates in the desire of humans generated due to the stimulation of environments or regarded the interactions between environments and individuals as being dynamic and the inadequate relationship as a dynamic situation where the solution is perceived to be uncertain in the process of making effort to achieve what individuals desire.

Job stress can be defined as harmful physical or mental responses occurring when what the job demands does not match the employee's ability, resources, or demands. Job stress occurs in situations where environments and individuals disagree with each other, that is, in situations where the environment requires jobs that exceed the individual's ability or does not satisfy the individual's demand.

The first definition of the concept of job stress was made from the viewpoint to see job stress as an external stimulus that affects individuals leading to the destruction of the physical and psychological stability of organization members. From this viewpoint, job stress was defined as "a characteristic of negative environmental factors such as excessive work related to a certain job, role conflicts, role ambiguity, and poor working conditions or job environments that threaten organization members" (Beehr and Newman, 1978). The second definition of job stress was made from the viewpoint of interactions between organisms

and environments. From this viewpoint, job stress was defined as "a state where an individual's ability does not meet the requirements of the job or the job environment provided by the organization does not meet the individual's desire" (French et al., 1974). Since the subjects of the present study are educational administrative service workers, job stress will be defined as a 'state of uncomfortable psychological responses felt by employees in the process of performing jobs due to the incongruity among factors related to the job.' The effects of job stress on the organization are manifested by job dissatisfaction, absenteeism, and turnover, which eventually lead to negative consequences on the organization's performance.

As for the effects of job stress on psychological tension or emotions, it can be said that as job stress increases, the degree of exhaustion of individuals' emotions increases too (Meyer et al., 1993). This is also shown by the findings of a study conducted by Sutton (1984) indicating that job stress can ultimately lead to the deterioration of mental and physical health. In particular, if an individual performs emotional labor during which he/she is personally controlled by emotions despite that his/her job stress due to work increases, he/she may perceive emotional labor more intensively because he/she has no personal vent. Therefore, we developed the following hypothesis.

H1: Job stress increases the level of emotional labor.

# Intimacy

"Intimus," which is the original Latin word for intimacy, means inside or the deepest inside. Therefore, intimacy is related to revealing or sharing one's deep private parts (Price et al., 1995). The largest asset of an enterprise intimate with its customers is naturally customer loyalty. The characteristics of enterprises that pursue customer intimacy as such are reviewed as follows.

First, such enterprises promote long-term relationships with customers. Enterprises intimate with customers recognize their relationships with customers formed in the long term over time as important investments rather than regarding the profits accrued in the first transaction with new customers as important. Second, such enterprises build and maintain a detailed information system for their customers. Third, such enterprises make effort to continuously provide customers with something more than what customers expect.

The intimacy in the present study can be said to be the degree to which the school staff, who are at the forefront of delivery of services, perceive that they are intimate with students or professors, who are their service customers.

Emotional labor can be regarded as a sort of emotional stress experienced in a situation where it is difficult to express one's substantive and negative emotions as they are. In cases where service workers perceive intimate emotions toward their service customers highly, they can express their emotions more frankly and this will ultimately make the service workers perceive the degree of emotional labor as being low. Therefore, we developed the following hypothesis.

# Professionalism in Work

The perspective of professionalism in work examined in the present study can be said to be a concept contrary to 'qualitative work overloading.' School staff members with professionalism in works that correspond to the areas of individuals were inferred to be highly likely to have high self-esteem for the works under their charge thanks to increases in work efficiency. In addition, since it was inferred through the foregoing that the level of perception of emotional labor by school staff members might be low, the following hypothesis 3 was established.

H3: Professionalism decreases the level of emotional labor.

# Organizational Commitment

Organizational commitment is known to be an important variable explaining the characteristics of organizations as with job satisfaction. Sheldon (1971) regarded organizational commitment as an individual's positive evaluation of his/her organization or the individual's will to achieve the organization's goals. That is, an organization member with strong organizational commitment recognizes the goals of the organization as his/her goals and contributes to the development of the organization.

The reason why diverse studies have been conducted in relation to organizational commitment is that organizational commitment is closely related to the effectiveness of the organizations. Accordingly, many studies have been conducted on the influencing factors that determine organizational commitment. Through a multifaceted study on the leading variables and outcome variables of organizational commitment, Meyer and Allen (1990) presented rewards, the number of years of continuous service, goal identity, positions, age, job satisfaction, job stress, education levels, and job environment as leading variables of organizational commitment and productivity, turnover intentions, absenteeism, and job satisfaction as outcome variables and argued that although the leading variables of organizational are not sufficiently consistent because they are quite diverse, the outcome variables are quite clear.

Eventually, organizational commitment refers to an employee's intention to devote his/her energy and loyalty to the organization to which he or she belongs. The organizational commitment as such can be defined as a psychological characteristic related to the attitudes of the employees, and it can be said that elements such as loyalty to the organization, the mind to make best efforts for the organization, and selfidentification of the employees with the organization are contained in organizational commitment.

It can be inferred that as the intensity of emotional labor increases, the level of attachment or commitment to the organization will became lower. Therefore, the following hypothesis 4 was established.

H2: Intimacy decreases the level of emotional labor.

H4: Emotional labor has negative impact on organizational commitment.

Organizational trust is trust in the objective employment relationship between an organization and its members and it is defined as the 'members' overall evaluation and conviction of the organization that the organization will conduct behaviors beneficiary or at least not harmful to the members' (Zapf, 2002).

Concretely, the reason why trust is necessary in organizations is that trust is important for the development of organizations as follows. First, trust promotes the organization to achieve high performance. Second, trust is an important element in achieving the long-term stability of the organization and happiness of the members. Third, trust induces organization members to have a strong sense of community and voluntarily participate in organizations' problems thereby enabling organic and humancentered flexible operation of organizations.

Levering (2000) argued that trust is a complex concept consisting of the factors of veracity, respect for individuals, and fairness, and presented three components of organizational trust; first, trust, which is in the relationship between the employees and the management; second, self – esteem, which is in the relationship between each employee and his/her work; and, third, fun, which is in the relationship between the employees who work together.

As such, trust is a very important element of organization's competitiveness and long-term sustainability in diverse aspects. Therefore, many scholars have promoted studies on interactions between trust and other variables. Such variables include group effectiveness, effective group problem solving, leadership effectiveness, and overall organizational effectiveness.

Trust can be classified by type into emotional trust and rational trust. Lewis and Weigert (1985) conducted a study on trust with a focus on relationships and they classified trust into two axes: emotion and reason. Cognitive trust, which is focused on the aspect of reason, is defined as selecting a person who is trusted based on some good aspects of the person under certain circumstances. This means the formation of subjective trust in relationships with others reflecting aspects that are aimed at pursuing practical interests. On the other hand, emotional trust, which is focused on emotions, is said to be formed through interpersonal emotional ties. That is, trust in others is understood as intimacy and the will to take risks formed through the creation of environments such as emotional ties. In the present study, based on the studies mentioned above, an integrated measurement of trust was made considering both the aspect of emotional trust and the aspect of rational trust.

Service workers who frequently experience emotional labor of being unable to properly deliver actual emotional expressions due to their organizations can be inferred to have low levels of trust in their organizations. Therefore, hypothesis 5 as follows was established.

H5: Emotional labor has negative impact on trust.

# Moderating Effects of Work Environment and Gender

In the present study intended to analyze the effects of emotional labor on university administrative staff members, the effects of emotional labor were regarded to appear differently according to the different work environments of universities or the genders of staff members. We tried to analyze the moderating effects of those two variables. The following research hypotheses were established assuming that the organization culture and working environment of 2 years junior colleges should be different from the researchoriented 4 years universities. Many staff members of researchoriented universities provide services to faculty members and graduate students as well as traditional undergraduate students. However, major role of the staff members of 2 years colleges is focusing on services for students only and in that sense, their working environment could be very different from the each other. Similarly, the perception of working condition and the perceived level of emotional labor might be different depending on gender and, therefore we developed the following hypotheses regarding the effects of moderating variables.

H6: Work environment has moderating effect between the antecedent factors and outcome factors of emotional labor. H7: Gender of service provider has moderating effect between the antecedent factors and outcome factors of emotional labor.

# Research Model

Based on the theoretical background, findings of previous studies, and research hypotheses described thus far, we set up a research model as follows (**Figure 1**).

# RESEARCH METHODOLOGY AND MODEL TESTING

# Questionnaire Survey and Measurement Scale

# Composition of Job Stress Questionnaire Items

To examine the job stress of administrative service workers, which is the purpose of the present study, the model of Matteson and Ivancevich (1987) was used to make questionnaire items for job stress and, job stress by using Likert 5-point scale. The

concrete composition of the questionnaire items for job stress is as follows (**Table 1**).

### Composition of Intimacy Questionnaire Items

fpsyg-09-02473 December 12, 2018 Time: 14:19 # 6

In the present study, the six items utilized in a study conducted by Nielsen (2001) were utilized to measure intimacy. The concrete composition of the questionnaire items is as follows (**Table 2**).

## Composition of Questionnaire Items for Professionalism in Work

In the present study, to examine administrative service workers' professionalism in work, knowledge in related majors, selfimprovement, job performance, and commitment to work were examined (**Table 3**).

## Composition of Questionnaire Items for Emotional Labor

Emotional labor is a unique characteristic of service labor. In service labor, "the quality of interactions" between employees and service practitioners per se constitutes a part of the service delivered to customers. In the present study, emotional labor was measured using a total of 12 items made based on the




TABLE 3 | Composition of questionnaire items for professionalism in work.

### Scale items for professionalism in work


(4) I think I have an ability differentiated from other employees for handling my work.


study conducted by Morris and Feldman (1996). The detailed questionnaire items is as follows (**Table 4**).

## Composition of Questionnaire Items for Organizational Effectiveness

In the present study, organizational commitment and organizational trust were defined as organizational effectiveness. Organizational commitment can be said to be a strong attitude of service workers to identify the goals of the organization with their goals and try to achieve the goals with positive attachment. The organizational commitment as such was measured with the five items used in a study on organizational commitment conducted by Allen and Meyer (1990). The concept of organizational trust was set as the belief about the organization set by the organization members (**Table 5**).

# Survey and Data Collection

The subjects of the present study were limited to staff members who provide administrative services at universities and data were collected through questionnaire surveys conducted with the staff members of four universities and two two-year junior colleges located in Seoul, South Korea. For the questionnaire surveys, the researcher firsthand visited the sites to explain the intent of the questionnaire and collected data through one-to-one questionnaire surveys. A total of 250 copies of the questionnaire were collected over approximately 3 weeks and 235 copies excluding questionnaires with unfaithful responses were used in the final analysis.

# Characteristics of Study Respondents

The demographic characteristics of the 235 respondents who responded to the questionnaire in the present study are as follows. As for sex men out numbered women as the percentages of

TABLE 4 | Composition of emotional labor questionnaire items.

### Scale items for emotional labor




men was 55.3% while that of women was 44.7%. As for ages, the percentage of subjects in their 30 s was the highest at 35.3% followed by those in their 40 s (27.7%), 20 s (24.3%), 50 s (11.5%), and 60 s (1.3%). As for the types of schools where they work, the percentages of universities, two-year junior colleges, and other schools were shown to be 58.3, 40.0, and 1.7%, respectively. As for the number of years of service, the percentage of nine years or more was the highest at 38.7% followed by 1∼3 years (20.0%), less than 1 year (18.3%), 7∼9 years (8.9%), 5∼7 years (7.2%), and 3∼5 years.

# Reliability and Validity of Measurement

Cronbach's α coefficient was used to verify the reliability of the measurement items used in this study. As shown in **Table 6**, Cronbach's α was found to be high, between 0.75 and 0.85. This exceeded the standard (0.70) presented by Nunnally and Bernstein (1994), securing reliability. Confirmatory factor analysis (CFA) was performed on the measurement items to verify the convergent validity of the constructs. The goodnessof-fit index for the CFA was χ <sup>2</sup> = 207.32 (df = 120, p = 0.000), CFI = 0.94, and RMSEA = 0.05, satisfying the guidelines from Hu and Bentler (1999). The results of the CFA indicated that the t-values of all factor loadings are significant at p < 0.001, securing convergent validity. In addition, construct reliability (CR) and average variance extracted (AVE) values that measure internal consistency were examined. The CR value of all variables was 0.70 or higher and the AVE value was 0.5, securing convergent validity.

Two methods were used to verify discriminant validity between constructs. First, CFA was performed on each of the items of the constructs included in the research model (Bagozzi and Yi, 1988) to confirm that the correlation between the constructs was less than 1 (**Table 7**). The results indicated that all correlation values between the constructs were less than 0.85, securing discriminant validity. Second, discriminant validity is secured when the AVE value of the construct is larger than the square value of correlation (AVE > ϕ 2 ), comparing the AVE values of each construct and the square value of the correlation between the two constructs. All standards were met, and discriminant validity was secured.


<sup>∗</sup>All factor loadings are statistically significant at p < 0.001. χ <sup>2</sup> = 207.319; df = 120; p = 0.000; TLI = 0.913; CFI = 0.939; RMSEA = 0.049.

.

TABLE 7 | Correlation matrix of study constructs<sup>∗</sup>

TABLE 6 | Reliability and validity of measurement.


<sup>∗</sup>All coefficients are significant at p = 0.000 level.

# Verification of the Research Hypotheses and Model Testing

Structural equation model analyses were conducted to verifying the research hypotheses (**Table 8**). The goodness-of-fit indices of the overall model were shown to be χ <sup>2</sup> = 303.016 (df = 130, p = 0.000), CFI = 0.878, TLI = 0.964, RMSEA = 0.067 indicating that the model is a good model that satisfies the criteria of Hu and Bentler (1999).

The results of verification of hypothesis 1 indicated that this hypothesis is statistically significant. The results as such can be said to have confirmed the negative consequences of emotional labor that have been shown in many existing studies and also can be said to provide a practical implication that measures to reduce the job stress of service workers should be prepared at the level of organizations.

The results of verification of hypothesis 2 indicated that this hypothesis is statistically significant. The results as such can be said to be interesting ones suggesting that measures to reduce the


TABLE 8 | Results of hypothesis testing and model analysis.

fpsyg-09-02473 December 12, 2018 Time: 14:19 # 8

χ <sup>2</sup> = 303.016; df = 130; p = 0.000; TLI = 0.964; CFI = 0.878; RMSEA = 0.067; ∗∗∗p < 0.000.

variable termed emotional labor should be sought at the level of organizations as well as in terms of personal relationships. These results can be regarded to be very interesting in that whereas many existing studies on emotional labor measured emotional labor mainly from the viewpoint of organizations, the present study suggested that the emotional exchanges of individual service workers are a variable that can reduce emotional labor.

The results of analysis of hypothesis 3 indicated that this hypothesis should not be adopted. In other words, professionalism has no statistically significant effect on the emotional labor. That is, professionalism in work can be regarded to have no statistically significant effect on emotional labor. Although the reduction of work burdens through professionalism in work was predicted to act as a factor to reduce emotional labor in the present study, the results of actual measurement indicated that individuals' professionalism in work has no effect on emotional labor. Consequently, the reason can be found from the fact that the work of university administrative staff does not require a high level of professionalism. The reason why the professionalism does not influence the level of emotional labor is maybe because the service worker's job expertise or experience of working knowledge does not necessarily mean the decrease of emotional labor.

The results of verification of hypotheses 4 and 5 indicated that emotional labor has significant effects on both organizational commitment and organizational trust. As mentioned in the research hypotheses, these results can be said to show that emotional labor is an important factor in determining the organizational effectiveness of universities. The above results of verification of the hypotheses were summarized as shown in the following table.

All hypotheses except for hypothesis 3 were shown to have been satisfied. Therefore, on organizing the study findings, it can be seen that job stress is a variable that has significant effects on emotional labor, and that the perception of high levels of job stress lead to the perception of high levels of emotional labor. In addition, it was shown that the perception of intimacy with students leads to the perception of low levels of emotional labor (Hypothesis 2) indicating that organizational support at the level of university is necessary for administrative service workers to form intimate relationships with students. Finally, since emotional labor has significant effects on the variables of organizational effectiveness (organizational commitment and organizational trust), work environment where low level of emotional labor can be perceived should be adapted to improve the organizational performance of administrative service workers of university.

# Analysis of Moderating Effect

### Moderating Effect of Work Environment

In Hypothesis 6, this study conducted a multi-group analysis to examine the moderating effect of work environment (2- vs. 4-year universities). In order to test the difference between the two different groups, this study analyzed whether there is a χ2 difference between the free model and structural weight constrained model to verify the moderating effects.

In **Table 9**, the χ2 difference between the free and constrained models is 280.664, which is much higher than the effective value of χ 2 at the 0.05 significance level. As a result, the hypothesis 6 was supported because the difference between the two models was significant, and the moderating effect of the university work environment was accepted. The moderating effect was verified and the path coefficients between the two groups were compared (**Table 10**). There is a significant difference between the two

TABLE 10 | Comparison of path coefficients between 2- and 4-year universities.


∗∗p < 0.01 and ∗∗∗p < 0.000.

TABLE 9 | Comparison of free and constraint models for verification of moderating effect of work environment.


groups when the critical ratio (CR) value of the path coefficients of the two groups shows as in the **Table 9**.

## Moderating Effect of Gender

fpsyg-09-02473 December 12, 2018 Time: 14:19 # 9

To test the hypothesis 7, we conducted a multi-group analysis to examine the moderating effect of men and women from the perspective of the level of emotional labor. In order to test the difference between the two different groups (men vs. women), this study analyzed whether there is a χ2 difference between the free model and structural weight constrained model to verify the moderating effects.

In **Table 11**, the χ2 difference between the free and constrained models is 158.18, which is much higher than the effective value of 55.76 at the 0.05 level when the degree of freedom is 30. As a result, the hypothesis 7 was supported because the difference between the two models was significant, and the moderating effect of gender was accepted. The moderating effect was also verified and the path coefficients between the two groups were compared (**Table 12**). There is a significant difference between the two groups when the critical ratio (CR) value of the path coefficients of the two groups shows as in the **Table 12**. Based on this result, we can find that the emotional labor of women is relatively more influenced by the level of intimacy and the emotional labor of men is more influenced by the level of job stress.

# CONCLUSION

# Discussion and Managerial Implications

Since service workers' emotion management is regarded as important in the field of services, the present study was intended to be conducted in the aspect of university administrative service workers among the diverse sectors of emotional labor that have been studied thus far. To this end, the present study was intended to be conducted centering on job stress, which is a leading variable of emotional labor that has been presented in many previous studies, with personal variables termed professionalism in work and intimacy, which are aspects where service workers' individual management is necessary, added to job stress.

In addition, the present study was also intended to be conducted from the perspective about emotional labor believing that organizational effectiveness will be reduced due to the effective emotion management, that is, the management of the norms for emotions expressed by service workers at the level of organization.

To this end, in the present study, the leading variables and outcome variables of emotional labor were extracted from the viewpoint of existing previous studies and for the parsimoniousness of study, it was intended to examine the effects of job stress, professionalism in work, and intimacy, which are leading variables of emotional labor, on the level of emotional labor. Accordingly, interviews were conducted in advance to secure the validity of study subject selection and after securing the validity of study subjects based on the results of the interviews, questionnaire surveys were conducted for empirical analysis with administrative workers working at six universities and school personnel at the points of contacts with students and professors.

The hypotheses were verified based on the results of empirical analysis conducted through a series of processes as such. The results of the analysis are as follows.

First, the job stress of service workers was shown to have increasing (+) effects on emotional labor. It was confirmed that the study findings indicating that the awareness of job stress has significant effects on the emotional labor perceived by service workers in other industries, apply to administrative service workers too. Therefore, it provides a practical implication that conditions under which emotion management can be performed should be prepared at the organization level through the reduction of job stress of service workers.

Second, intimacy was shown to reduce the degree of emotional labor of service workers. This is significant in that the fact that the intimacy perceived by administrative service workers in their relationships with customers affects their personal emotion management was confirmed. In addition, it also provides a practical implication that rather than utilizing only emotion management as a means of control at the level of organization, meetings and events that can form positive relationships with customers (students) should be actively utilized by school organizations to support organization members in relation to their emotional exhaustion with interest.

Third, although the professionalism in work of service workers was thought to have decreasing (−) effects on emotional labor, the results of verification indicated that the decreasing

TABLE 12 | Comparison of path coefficients between men and women groups.


TABLE 11 | Comparison of free and constraint models for verification of moderating effect of gender.


(χ 2 -value = 55.76/df of Dif = 30). effects were not significant. Based on the foregoing, it can be inferred that the sector of personal abilities does not affect service workers' emotion management and that individual service workers perceive emotions and their professionalism in work as separate variables.

# Limitations and Future Research Direction

fpsyg-09-02473 December 12, 2018 Time: 14:19 # 10

The present study examined the effects of administrative service workers' job stress, intimacy, and professionalism on their emotional labor, and the effects of emotional labor on organizational commitment and organizational trust in administrative service situations. Therefore, the present study has the implications as mentioned above but it has the following limitations.

First, in-depth discussions on the significance of emotional labor, which is a central concept of discussion in the present study, for job attitudes in administrative services were insufficient. Therefore, qualitative studies are judged to be necessary as future studies because shortcomings may occur since the results of the present study have not been derived from the opinions of all administrative service workers although the opinions were surveyed through in-depth interviews with administrative service workers present in the vicinity.

Second, the present study considered only intimacy, professionalism in work, job stress, and organizational

# REFERENCES


effectiveness in relation to the effects of emotional labor on job attitudes. However, other underlying factors would affect job attitudes in relation to emotional labor. As follow-up studies, it is hoped that multidimensional studies will be conducted on which factors affect emotional labor and job attitude.

Third, although the present study was conducted in the aspect of administrative service workers, in the future, the effects of emotional labor on the performance of students along with the performance of schools should be examined by conducting surveys on students who are provided with the service along with administrative service workers.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of ethics committee of Hanyang University with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethics committee of Hanyang University.

# AUTHOR CONTRIBUTIONS

S-LH designed and drafted the work. H-SS designed the work and analyzed the data. WC drafted the work and collected the data.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Han, Shim and Choi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Research on How Emotional Expressions of Emotional Labor Workers and Perception of Customer Feedbacks Affect Turnover Intentions: Emphasis on Moderating Effects of Emotional Intelligence

### Young Hee Lee<sup>1</sup> , Suk Hyung Bryan Lee<sup>2</sup> \* and Jong Yong Chung<sup>2</sup>

<sup>1</sup> R&SD Strategy Center, Industry-Academy Cooperation Foundation, Chungwoon University, Incheon, South Korea, <sup>2</sup> School of Integrated Technology and Entrepreneurship, Chungwoon University, Hongseong, South Korea

### Edited by:

Jungkun Park, Hanyang University, South Korea

### Reviewed by:

Jaewook Kim, University of Houston, United States Dongyoup Kim, Kookmin University, South Korea

> \*Correspondence: Suk Hyung Bryan Lee blee@chungwoon.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 13 September 2018 Accepted: 27 November 2018 Published: 04 January 2019

### Citation:

Lee YH, Lee SHB and Chung JY (2019) Research on How Emotional Expressions of Emotional Labor Workers and Perception of Customer Feedbacks Affect Turnover Intentions: Emphasis on Moderating Effects of Emotional Intelligence. Front. Psychol. 9:2526. doi: 10.3389/fpsyg.2018.02526 Previous studies have used various external variables and parameters as well as moderator variables such as emotional intelligence have been to understand emotional labor and its related problems. However, a comprehensive model to study such variables' correlations with each other and their overall effect on emotional labor has not yet been established. This study used a structural equation model to understand the relationship between employees' expression of emotional labor and perception of customer feedbacks. The study also looked at how the perception of customer feedback affects emotional exhaustion in order to understand how emotional exhaustion affects job satisfaction and turnover intentions. Further, in order to fully understand the effects of emotion on emotional labor at the service contact points, this study developed and tested a model of emotional labor with four factors of emotional intelligence as moderating factors. Five hundred and seventy nine emotional labor workers in service industries in the United States were collected and 577 valid survey results have been analyzed. The result shows that there exists moderating effects of emotional intelligence on how employees' Deep Acting and Surface Acting recognize customers' reactions, both positive and negative, that would affect employees' Emotional Exhaustion and Job Satisfaction, and hence, Turnover Intention. The result suggests that employees with better understanding of their own emotions, although they are more likely to recover from emotional exhaustion, experience a greater negative effect when there is a discrepancy between what they feel and how they should act.

Keywords: emotional labor, emotional expression, perception of customer feedback, emotional exhaustion, emotional intelligence

# INTRODUCTION

The growth of the service industry has prompted the importance of employees' emotional labor, and it has received increased attention from various fields including organizational behavior and organizational psychology (Grandey and Gabriel, 2015). A number of studies have found that emotional labor required by organizational norms can negatively affect individual well-being

(Hülsheger and Schewe, 2011). Grandey (2000) stated that emotional labor can negatively affect workers' job satisfaction, and the job satisfaction has strong implications for attendance, turnover, sabotage, job performance, and the mental and physical health of employees (Miao et al., 2017).

Emotional labor also indirectly affects employees by negatively affecting the performance of the organization. Among a range of factors leading to turnover intention, burnout has been found to play a significant role (Kilo and Hassmén, 2016), and surface acting contributes to the burnout, job dissatisfaction, and turnover intention (Lee and Chelladurai, 2018). Departures of skilled labor workers would cause additional expenses in recruiting, hiring, and training of new employees. It will also have negative effects on the service quality of the organization. Therefore, higher turnover rate would negatively affect the organizations business performances. In turn, it will negatively affect the job stability of the remaining employees. Therefore, in order to improve corporate performance, increase employee motivation and enhance employee's job satisfaction, it is important to address employees' emotional labor issues.

As the frontline employees are "barometers of the business" (Dagger et al., 2013, p. 498), especially in service industry, the organizations are trying to provide various supports for their employees at the service contact points (Cohen and Wills, 1985; Ashforth and Humphrey, 1993; Rhoades and Eisenberger, 2002; Kim et al., 2017). Social support is one of the supports that can be provided for the employees as it has long been identified as a resource that enables people to cope with stress (Fenlason and Beehr, 1994; Hobfoll, 2002). Social support is loosely defined as "the availability of helping relationships and the quality of those relationships" (Leavy, 1983, p. 5), and it could be provided by coworkers, supervisors, and the organization.

Lately, a number of studies put on more emphasis on the role of emotional reactions and interpersonal aspects (Mearns and Cain, 2003; Ullrich et al., 2012) as emotions are common among customers in service encounters (Gabbott et al., 2011; Strizhakova et al., 2012). Further, many studies have emphasized the importance of emotional intelligence which refers to employees' ability to control their emotions and empathize with customers. Specifically, Mayer stated that "the ability to perceive emotions, integrate emotion to facilitate thinking, understand, and manage emotions to promote personal growth are paramount to professional achievements and mental health and well-being" (Mayer et al., 2016).

As an individual's success at work is 80 percent dependent on emotional quotient and only 20 percent dependent on intelligence quotient (Goleman, 1998), it is important to address how Emotional intelligence affects the emotional labor. Emotional intelligence is found to be negatively associated with burnout, and it could therefore be a potential protective factor (Chan, 2006; Mikolajczak et al., 2007). Zysberg et al. (2017) found the similar result and stated that emotional intelligence can be inversely associated with burnout, a condition brought about by lack of effective emotional regulation and management. Emotional exhaustion is affected by employees' emotional expression in the service encounter, and emotional intelligence of the individual will play a role in controlling these effects (Morris and Feldman, 1997; Park, 2018). As emotional intelligence is known to be improved by learning, experience, and training (Gardner, 1995), finding ways to enhance employees' ability to control their emotions and empathize will help them.

Due to the importance of emotional intelligence as a personal resource to facilitate processing emotions into effective behavior patterns, Laborde et al. (2016) suggested that the organizations "should implement emotional intelligence training workshops for coaches to help them engage in emotional labor that benefits both themselves and the organizations." However, although there have been various researches that looked into the effects of emotional intelligence on emotional labor, a comprehensive model to study the effects of the sub-dimensions of emotional intelligence has not yet been established.

Wong and Law (2002) extended the concept of emotional intelligence defined by Salovey and Mayer (1990), and refined the framework to categorize the sub-dimensions of emotional intelligence as perceiving, understanding, managing, and using emotions. Each of these sub-dimensions has different characteristics from each other. Further, due to their different characteristics, how each dimension can be strengthened and enhanced would be different from each other as well. Therefore, in order to understand how emotional intelligence affects emotional labor and suggest how emotional labor could be properly managed, how each sub-dimension of emotional intelligence affects emotional labor should be investigated instead of looking at the overall effect of emotional intelligence on emotional labor.

This study aims to suggest how employees' turnover intention could be lowered in order to enhance organizational performance as well as their individual happiness. Therefore, this study adapted two main perspectives. The first is to look for solutions through the governing principles that control workers' emotional expressions. The second is to look for solutions through individually controlled emotional intelligence. In order to accomplish these goals, this study used a structural equation model (SEM) to test the structural relationship among Emotional Exhaustion, Job Satisfaction, and Turnover Intention with Verbal and Non-Verbal Expressions, Customer Feedbacks, Surface Acting, and Deep Acting as observed variables. Further, the SEM was applied to the framework of Emotional Labor Model developed by Grandey and Gabriel (2015) to test the moderating effects of subdimensions of emotional intelligence on each stage from Emotional Labor through Emotional Exhaustion and Job Satisfaction to Turnover Intention. Identifying the moderating effects of each sub-dimension of emotional intelligence at each stage would be helpful in suggesting the solutions to employees' emotional problems which would, in turn, reduce their turnover intention.

The results of this study would suggest that effective management of emotions can increase the satisfaction of the customers on the services, reduce emotional exhaustion of employees, and enhance employees' job satisfaction, thereby lowering their turnover intention.

# THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT

# Service Encounter

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 3

A service encounter occurs every time a customer interacts with the service organization. Specifically, the term 'service encounter' is used to indicate face-to-face interactions between a buyer and a seller in a service setting. These interpersonal exchanges can strongly influence customer satisfaction (Solomon et al., 1985).

Shostack (1985) added that service encounter is not only the interaction between the service provider and the customer, but also the period of interaction. In other words, the concept of service encounter is defined comprehensively by adding factors such as material, convenience, time, and space to human interaction. On the other hand, Keaveney (1995) referred to service encounter as the human interaction between the customer and the service employee, emphasizing that the human factor of the employee and the customer is the most important at that point.

The customer takes a variety of emotions and attitudes through communication with the employee while the service is being provided, and the employees and customers are highly interdependent in the service encounter (McCallum and Harrison, 1985).

# Emotional Labor

Emotional labor is generally defined as the act of expressing organizationally desired emotion during the service transactions, and is caused by the difference between the actual emotional state experienced by the employee and the emotional expression required by the organization's emotional expression norms for effective job performance (Hochschild, 1983). Emotional labor is similar and consistently presented regardless of the differences of various cultures (Grandey et al., 2010).

Emotional labor is a type of labor in which emotional management activities for producing and maintaining a specific emotional state required by the job account for more than 40% of the jobs (Hochschild, 1979), and Emotional labor workers express certain emotions in the workplace, regardless of actual emotional experiences, in order to successfully follow the job demands (Hochschild, 1983). Jobs that require emotional labor include call center representatives, department store sales, flight crew, nurses, etc. (Hochschild, 1983; Brotheridge and Grandey, 2002; Williams, 2003).

Morris and Feldman (1996) defined emotional labor as effort, planning and control needed to express emotions demanded by firms while emotional workers are serving customers, and Grandey (2000) defined emotional labor as an effort to comply with organizational or vocational expression rules. Emotional labor is generated by controlling emotional expression during work (Lam and Chen, 2012), and it has been shown to affect counterproductive work behavior (Raman et al., 2016), service quality and customer satisfaction. Emotional labor workers conduct their duties through face-to-face conversations, voices, and actively expressed emotions in order to influence other people's emotions, attitudes, and behaviors.

The effects of emotional labor within the organization are largely negative. These adverse effects directly affect the employees in the service encounters and spread to customers and organizations. Employees experience stress, depression, panic disorder, psychological distress, and job dissatisfaction. For the organizations, negative effects are exacerbated by employee dissatisfaction, performance reduction, and turnover. Further, emotional labor causes service quality deterioration, which leads to negative consequences such as decreased satisfaction with service and increased complaints by customers. Therefore, the emotions of the employees who represent the company in the closest position to the customers should be managed as an important variable in order to ensure the effectiveness of the organization.

Therefore, better understanding of emotional labor would provide meaningful insights to maintain customer's favorable attitude, improve the service quality of employees, and lower the turnover intention by managing appropriate emotional expressions at the service contact points.

# Employees' Emotional Expressions Surface Acting and Deep Acting

Hochschild (1979, 1983) stemmed her dimensions of the emotional labor based on the approach of performing, and depicted two kinds of emotional labor. The first is Surface Acting. It refers to worker's recreating of feelings, that are not really felt, by changing their outward appearance such as facial gesture, motions, or voice tone, when displaying required feelings. The second is Deep Acting. It happens when representative's sentiments don't fit the circumstance; then they utilize their preparation or past experience to express fitted feelings. Although Ashforth and Mael (1989) added another dimension, Genuine Emotion, to Hochschild's work and suggested three main dimensions, Emotional labor has been mainly studied in two ways: Surface Acting and Deep Acting (Hochschild, 1983; Grandey, 2000; Brotheridge and Lee, 2003; Monaghan, 2006; Johnson and Spector, 2007).

Researchers found that Surface Acting has more negative effects on employees than Deep Acting. For example, employees smile and provide customers with services in kind voice even when they have negative feelings. Such a discrepancy in feelings often leads to emotional conflicts that will have a negative impact on employee job satisfaction. It also leads to emotional exhaustion and burnout for workers (Ashforth and Humphrey, 1993). On the other hand, Deep Acting refers to one's effort to consciously modify feelings in order to express the desired emotion (Hochschild, 1983). It has a positive effect on the individual's job satisfaction as well as the organization's performance, accompanied by a positive result in interaction with the customer.

Both Surface Acting and Deep Acting are artificial rather than natural behaviors. However, while Surface Acting is a false emotional state that faithfully expresses the organization's expression norm but does not actually control the inner emotional state, Deep Acting is the actual emotional state that attempts to match the organization's expression norm with itself

(Diefendorff et al., 2005). As a result, while employees in Deep Acting have a high level of job satisfaction and organizational satisfaction, Surface Acting negatively affects job performance because it is based on surface norms without controlling the inner feelings of employees. For example, Larson and Yao (2005) find that Deep Acting facilitates communication and more genuine forms of empathy toward patients, leading to higher levels of professional satisfaction.

Deep Acting may serve as a buffer against negative moods (Judge et al., 2009; Polletta and Tufail, 2016), and it has been found to reduce emotional exhaustion and emotional dissonance (Brotheridge and Lee, 2003; Kim and Kim, 2018). However, the employees in Deep Acting have negatively affected personal accomplishment, and it leads to conflicting results with surface behavior (Kruml and Geddes, 2000; Brotheridge and Grandey, 2002; Johnson and Spector, 2007). These results suggest that the consequences of emotional labor may depend on the type of emotion strategy utilized as well as the context from which it is performed (Grandey and Gabriel, 2015).

### Verbal Expression and Non-verbal Expression

Communication is an important means of interactions that controls and links one's relationship with the others as a member of the society (Gabbott and Hogg, 2001). In service encounters, Non-Verbal Expression and Verbal Expression both are major factors affecting customer satisfaction and job performance (Gayathridevi, 2013). As communication between employees and customers in service encounters could be a differentiating factor that separates a company from its competitors, it is important to manage the verbal and non-verbal communication performed by employees in order to improve the quality of interaction between employees and customers.

When customers choose products and services, they communicate with the company's front line employees instead of the company itself. Therefore, the emotional response of the customer depends on the communication methods used by the employees at the service contact point where many interactions between the customer and the employee occur (Park et al., 2012). However, verbal and non-verbal efforts of employees to communicate with customers are, by nature, emotional labor from the employees' standpoint. Thus, it is important to understand the interaction between the customer and employee at the service contact point.

From customers' point of view, employees' Verbal and Non-Verbal Expressions are considered as their Emotional Expression. Emotional Labor model developed by Grandey and Gabriel (2015) uses both Emotion Regulation (Surface Acting and Deep Acting) and Emotion Performance (Verbal and Non-Verbal Expression). This study would use such a framework to fully understand Emotional Labor and how its variables are structurally related.

### **Verbal expression**

The linguistic communication is an important factor that can never be overlooked in the service environment, and in the relationship between the service provider and the customer. When delivering products and services, employees communicate information about products and services to customers through language. Language is the center of communication in human relationships and is the transmission and expression of meaning (Sommerville, 1982).

Linguistic communication is one of the systems promised by our society that individuals use to express their thoughts and feelings. Therefore, clear and rapid information transfer is an important clue to customer satisfaction (Kim, 2005), and there is a tendency for intimacy and trust to be increased through the communication process (Kim, 2007). In other words, intimacy and comprehension between employees and customers are increased by sending and receiving the necessary meanings between each other (Jandt, 1998). Employees can only reduce the barriers that may arise in the communication process by fully taking the customer's situation into account (Bugental et al., 1970; Dodd, 1998).

Through linguistic communication, employees clearly communicate their desire for customers, and they understand what they need to do and respond to them, thereby creating mutual trust and intimacy, and eventually lead to a positive response from customers (Kim, 2005). Reliable verbal communication can maximize message delivery (Jones and LeBaron, 2002). Furthermore, Berko et al. (2016) noted that linguistic communication is important in corporate marketing activities and that it is directly linked to corporate trust formation. Marketing is closely related to the delivery of product descriptions and services, and marketing activities are mostly conducted in language.

### **Non-verbal expression**

Non-verbal communication is a means of expressing oneself using space, time, and body movements (Dodd, 1998). Goldhaber (1982) defined non-verbal communication as all the silent messages except the language. In other words, non-verbal communication is the interaction of senders and recipients in a communication context, with the exception of language (Samovar et al., 1981).

Verbal communication is necessarily accompanied by nonverbal elements and sometimes non-verbal communication plays a more decisive role than verbal content. Birdwhistell (1955) argued that, in the communication process, 35% of linguistic messages and 65% of non-verbal messages are delivered. A person reads or expresses emotions or intention through non-verbal cues in situations where there is a limit to expressing them only by the use of language in a communication and a large amount of information is acquired in a short time (Stiff et al., 1994). Non-verbal communication also has the function of controlling social interaction. The beginning and end of conversations, the order of utterances, etc. can be controlled through non-verbal signals (Richmond et al., 1991).

Studies on communications have shown that non-verbal communication delivers more meaning faster than language and conveys feelings and emotions more accurately (Delmonte, 1991). Non-verbal communication narrows the psychological distance between employees and customers, increasing positive emotions such as pleasure and satisfaction, and helping to

interact with linguistic communication (Sundaram and Webster, 2000).

Sundaram and Webster (2000) classified non-verbal communication into kinesics, proxemics, paralanguage, and physical appearances. Employees must maintain facial expressions with a bright smile (Burgoon et al., 1990), since the emotionless expression of the employees can cause discomfort to the customers. In other words, the face is the most important part that can contain a lot of information. It is called a multimessage system (Argyle, 1975) that can read information about personality, interest, reaction and emotion in facial expression. Looking elsewhere or a failure to pay attention to the customer may be conveyed to the customer in a feeling of indifference and neglect (Leigh and Summers, 2002).

On the other hand, the frequent convergence of employees and customers in service encounter increases the intimacy with each other (Wainwright, 1999). In addition, nodding or proper tilting of the customer's words in the body language can be seen as an act of showing interest to the customer. This body language is one of the important forms of non-verbal communication that can convey interest, intimacy, benevolent feelings, and trust to customers (Mehrabian and Williams, 1969; Burgoon et al., 1990). Overall, individuals tend to rely on non-verbal communication in order to understand the emotion, while they rely on verbal communication in order to understand the thoughts (Hall and Schmid Mast, 2007).

# Customer Feedback (Positive/Negative)

We experience emotions, whether they are perceived or not, at every moment in various situations. In general, emotion is generally considered to be composed of two sub-factors that are correlated but independent: positive affect and negative affect. According to Watson et al. (1988), positive affect refers to emotions such as joy, pleasure, satisfaction, and happiness. Negative affect refers to feelings such as anger, fear, sadness, and guilt.

Emotions themselves refer to mental health (Diener, 1984). As static emotion promotes individual's health and has the potential to bring about various physical, psychological and social changes, many scholars actively study the relationship between static emotional and physical health and psychological well-being (Lyubomirsky et al., 2005). In contrast, inadequacy has been widely used as a key indicator of mental health, and many studies have addressed the relationship between such negative emotions and problem behaviors such as psychopathology, suicide, assault, and stress (Sung and Gwon, 2010).

These emotions are also the result of actions or cognitive processes, and they are also the cause of behavior itself as well as cognition (Park and Lee, 2016). According to Pugh (2001), emotions of customers are behavioral responses caused by external environmental stimuli and they cause behavioral change according to the situation. The employees in service encounter can communicate their feelings to the customers during the interaction with the customers, and these feelings that are conveyed to the customers cause the positive or negative feelings of the customers.

In service encounter, it is important for employees to understand whether the customer's response in the form of emotional expression is positive or negative. Because of the differences in experiences and values that customers and employees have, perceived meanings of emotional expressions might be different between customers and employees, which may lead to problems. The service provider could prevent such problems that may arise in communication process by conveying the message in full consideration of the customer's situation (Dodd, 1998).

This study will focus on the Customer Feedback perceived by employees that occur in service encounter. Identifying how emotional labor of the employees is related to Customer Feedback would be helpful in suggesting effective ways to reduce emotional labor in service encounter. In particular, it is expected that more positive results from service encounter could be expected when Customer Feedback, whether it is positive or negative, are properly confirmed. So we propose the following hypothesis to be tested in this study.

Hypothesis 1: Employees' emotional expressions will affect perception of customer feedback.

[H1\_a1] Surface Acting → Customer Feedback (Positive) [H1\_a2] Deep Acting → Customer Feedback (Positive) [H1\_a3] Non-Verbal Expression → Customer Feedback (Positive) [H1\_a4] Verbal Expression → Customer Feedback (Positive) [H1\_b1] Surface Acting → Customer Feedback (Negative) [H1\_b2] Deep Acting → Customer Feedback (Negative) [H1\_b3] Non-Verbal Expression → Customer Feedback

(Negative) [H1\_b4] Verbal Expression → Customer Feedback (Negative)

# Emotional Exhaustion

Emotional exhaustion refers to exhausted and depleted emotions due to work and is a chronic response to work stress situations that are associated with conceptually high levels of human contact (Ryan, 1971; Maslach, 1982; Mikolajczak et al., 2007; Zhao et al., 2018). Maslach (1982) described emotional exhaustion as a negative self-concept, a negative attitude toward work, and physical and emotional exhaustion symptoms including loss of interest or feelings to the customer. It is defined as the negative psychological experience that a person in charge of a job with high interpersonal contact experiences is exposed to a long time stressor (Maslach and Schaufeli, 1993).

Emotional exhaustion is caused by persistent and excessive emotional labor. Hochschild (1983) explained that emotional labor causes negative consequences such as emotional inactivation stress, physical exhaustion, emotional exhaustion, and absenteeism. Grandey (2000) also said that the emotional adjustment and social needs of emotional labor workers cause job exhaustion. In order to alleviate this emotional depletion, a phenomenon of de-personification that seeks psychological distance from others appears. When this de-personalization

becomes severe, it leads to a decrease in the sense of selffulfillment that a person feels no more efficient in performing tasks related to customers or performing their jobs (Maslach, 1982).

The surface behavior follows the norm of the emotional behavior manual given by the organization regardless of the employee's own feelings (Hochschild, 1983). When such a surface behavior is performed and the different emotions than the workers' true feelings should be expressed, emotional laborers experience significant suffering, and, thus, they are more likely to experience emotional exhaustion. Ashforth and Humphrey (1993) argued that when the surface behavior is performed while the disparity between the emotional state of the person and the needs of the organization exists, the worker experiences job exhaustion by dissatisfaction with his job and his own internal feelings. In addition, Brotheridge and Lee (2003) asserted that emotional exhaustion is caused by emotional incongruity. This suggests that emotional incongruity deepened by surface behavior causes emotional exhaustion.

Pugliesi (1999) found that both surface and internal actions negatively affect job satisfaction, while internal actions were less negative than surface activities. Ashforth and Humphrey (1993) suggested that the degree of exhaustion experienced while performing in-depth activities was low. In addition, Abraham (1998) and Morris and Feldman (1996) found that emotional labor is not necessarily negative, and in-depth behavior may be associated with positive outcomes because of positive feedback or self-fulfillment.

On the other hand, inner behavior refers to the effort to consciously modify the emotions of the employees according to their norms to express desirable behaviors (Hochschild, 1983), and emotional inconsistency between their actual behavior and how they actually feel is not large. It may not be easily confirmed that its effect on emotional exhaustion is smaller than the effect of the surface behavior, but the direction of the effect is opposite to the surface behavior. Because the concept of labor is included in the definition of emotional labor itself, internal action must also accompany the intended effort.

In the study of Kammeyer-Mueller et al. (2013), which metaanalyzes the leading and the inferring factors of surface and inner behavior, surface behavior showed a positive correlation with stress and burnout, while inner behavior was estimated to be positively correlated, but not statistically significant.

The influence of employees' emotional exhaustion in service encounter is negative and extends from individuals to organizations. Emotional depletion and emotional exhaustion refer to a state of weakened sense of accomplishment that negatively affects one's self, not just one's emotions (Maslach and Jackson, 1981). Therefore, it is difficult to continuously treat people nicely when the person who is doing the work is emotionally depleted. In this context, Lee and Ashforth (1993) and Maslach et al. (1996) argue that job exhaustion affects job satisfaction and organizational commitment, reduce job satisfaction and increase absenteeism and turnover intentions.

There is also a direct negative relationship between emotional exhaustion and service performance (Wright and Cropanzano, 1998). Employees who have experienced emotional exhaustion will have difficulty making the same effort with their customers (Krishnan et al., 2002). Emotional employees are less likely to be considered genuinely thoughtful or pleasant by their customers, and experience limited patience when they respond to customers (Grandey, 2003). And Emotional Exhaustion affects counterproductive work behavior (Raman et al., 2016).

According to a study by Barger and Grandey (2006), emotions are known to have a contagion among organizational members as well as affecting job performance and job satisfaction of the parties. Emotional labor required as a role within an organization increases emotional exhaustion (Grandey, 2003), and emotional labor induces tension by emotional dysregulation and consumes emotional resources in the process of emotional labor (Hochschild, 1983; Wharton, 1993). In other words, emotional labor, which sells its emotions for pay, exhausts cognitive and emotional resources (Hochschild, 1983).

Previous researches regarding emotional labor have a common opinion that surface behaviors increase emotional exhaustion (Brotheridge and Grandey, 2002; Grandey, 2003), but there is a disagreement about the relationship between inner behavior and emotional exhaustion.

Grandey (2003) argued that both surface and inner behaviors can increase emotional exhaustion. Surface behaviors increase emotional exhaustion by increasing discrepancies between normative emotions and actual emotions, and inner actions can also consume considerable energy and effort to match emotions with normative emotions. In contrast, Brotheridge and Grandey (2002) argued that surface behavior increases emotional exhaustion, but internal action reduces emotional exhaustion. This is because employees can perform inner actions to minimize the state of emotional dissonance and achieve achievement through achieving high performance.

Grandey (2003) hypothesized that inner behaviors would increase emotional exhaustion, but research shows that internal behavior is not significant for emotional exhaustion. These results suggest that inner behavior minimizes emotional dissonance and affirms customer's positive response. Based on the results of previous studies, this study hypothesized that each employee's emotional expression and customer's response would affect employee emotional exhaustion.

Hypothesis 2: Employees' emotional expressions will affect emotional exhaustion.

[H2\_1] Surface Acting → Emotional Exhaustion

[H2\_2] Deep Acting → Emotional Exhaustion

[H2\_3] Non-Verbal Expression → Emotional Exhaustion

[H2\_4] Verbal Expression → Emotional Exhaustion

Hypothesis 3: Perception of Customer feedbacks will affect emotional exhaustion.

[H3\_1] Customer Feedback (Positive) → Emotional Exhaustion

[H3\_2] Customer Feedback (Negative) → Emotional Exhaustion

## Job Satisfaction

There have been various attempts to define Job Satisfaction. Hoppock (1935) defined job satisfaction as a combination

of psychological, physiological, and environmental contexts of organizational members. Locke (1976) defined it as a pleasant and positive emotional state resulting from the evaluation of his or her job or job experience, and described it as being emotional and therefore being discovered and understood only by internal assumptions.

Job satisfaction is determined not only by the difference between the expectation recognized by the employees and the actual experience but also by the relationship between the organizational environment and personal characteristics such as the relationship between the supervisor and the employee. Job satisfaction is a personal response and can be viewed as an emotional response to a task or job (Bhuian and Mengue, 2002). It is known that emotions do not only affect individual performance and job satisfaction but also spread among the members of the organization (Barger and Grandey, 2006). In a recent study, job satisfaction refers to the state of emotions that the members feel in their subjective and relative view of the process of performing their duties and the results of their job performance, based on the individual's values and beliefs (Park et al., 2011).

Job satisfaction is classified into various dimensions such as job, working environment, compensation, administration, position, and promotion depending on purpose and subject of research (Locke, 1976; Farber, 1991; Park et al., 2011). In terms of the society as a whole, the level of job satisfaction of the organizers is the basis for measuring the smooth operation and performance of the organization. Satisfaction with the job also affects the life outside the duty of the organizer; it has a positive side effect (Robbins, 1998). Employees' satisfaction with their duties has a significant impact on the performance as their positive attitude toward the organization would have a great influence on the organization's performance. Job satisfaction of the employees increases the quality of service and directly affects the evaluation of consumers, which is an important factor affecting repurchase (Brown and Peterson, 1993).

Job satisfaction is important because it contributes not only to personal happiness but also to the efficient operation and performance improvement of the organization. In addition to affecting positive mental health through satisfaction of work life on an individual level, job satisfaction is also known to have a direct relationship with exhaustion from work, low morale, and productivity.

In previous studies, emotional exhaustion has been reported as a major variable that negatively affects job satisfaction. The emotions of the employees facing the customers at the service contact point are related to the satisfaction of the customers, the emotions of the employees are related to the relationship with the customers, and they affect the job satisfaction (Macdonald and Sirianni, 1996; Kinnie et al., 1999). Also, a study by Park et al. (2011) on the manager of the retail industry showed that the manager's emotional exhaustion had a negative effect on job satisfaction. Based on the previous findings, this study has established the following hypothesis.

Hypothesis 4: Employees' emotional exhaustion will affect job satisfaction.

# Turnover Intentions

Price (2001) defined the turnover as the departure from the organizational status of the employees, which means disqualification as a member of the organization to which it belongs. The turnover intention, which is the stage immediately before the turnover, is used as an indicator to predict turnover. Employee turnover lowers other employees' willingness and duties to remain in the firm. Further, departures of skilled labor workers would cause additional expenses in recruiting, hiring, and training of new employees. It will also have negative effects on the service quality of the organization. Therefore, higher turnover rate would negatively affect the organizations business performances (Mohsin et al., 2013). Therefore, there is an increasing tendency in interest and effort to decrease the turnover rate of the employees (Cho et al., 2009; Lee and Ryu, 2014).

Although there are various factors that cause the turnover of employees in the service industry, the one of the biggest problems is emotional exhaustion caused by emotional labor resulting from customer service work (Lee and Ashforth, 1996; Choi et al., 2012). Therefore, it is very important for the service organizations to pay close attention to job exhaustion as it would affect their turnover problems.

Based on the results of previous studies, this study has established the following hypothesis.

Hypothesis 5: Employees' emotional exhaustion will affect turnover intention.

Hypothesis 6: Employees' job satisfaction will affect turnover intention.

# Emotional Intelligence

Salovey and Mayer (1990) found that emotional intelligence enables us to observe and look at one's emotions, to properly manage one's emotions, to synchronize one's own emotions, to empathize with other's emotions, and to make good human relations. In other words, emotional intelligence is the ability to lead emotions productively and to synchronize oneself for their own goals in emotional situations. Therefore, emotional intelligence is associated with good job performance (Myers and DeWall, 2017).

In psychology, emotions are interpreted in the same way as emotional intelligence related to the know-hows involved in social situations and the successful management of oneself (Myers and DeWall, 2017). In the process of the perception of external stimuli, emotion has a significant influence on individual decision making according to age, gender, education, psychological state, family relation, economy, customs, religion, race (Prinz, 2007). It is also an interrelated skill related to the ability to create and approach emotions, the ability to understand emotions and emotional knowledge, and the ability to control emotions to promote emotional and intellectual growth, and is distinguished from general intelligence and personality (Mayer and Salovey, 1997; Wong and Law, 2002).

Goleman (1995) showed that emotional intelligence can synchronize individuals and keep oneself from frustration, control impulse and delay satisfaction, prevent emotional

accidents and stressful accidents, and empathize with others. Abraham (1999) defined emotional intelligence as the ability to accurately assess and express emotions of oneself and others, the ability to regulate emotions, and the ability to use emotional knowledge to solve problems. Dulewicz and Higgs (2000) defined emotional intelligence as a concept that persuades and influences others by keeping it and motivating it, away from recognizing emotions.

Salovey and Mayer (1990) defined emotional intelligence as 'the ability to accurately perceive, express and express emotions, the ability to generate and use emotions to promote thinking, the ability to understand emotions and emotional knowledge, the ability to adjust emotions to promote development'. Mayer et al. (2016) added areas of reasoning and claimed that emotional intelligence contained the four major components: (a) the appraisal of emotions; (b) the understanding of emotions; (c) the management of emotions; and (d) the utilization of emotions.

Wong and Law (2002) extended the concept of emotional intelligence defined by Salovey and Mayer (1990), and refined the framework to categorize the sub-dimensions of emotional intelligence as perceiving, understanding, managing, and using emotions. Wong and Law (2002) also developed WLEIS questionnaire, and it has been used in subsequent studies such as Law et al. (2004) to verify reliability and validity.

It is necessary for the individual to recognize the reality and cause of his emotions, as well as to understand and empathize with emotions of others. The ability to express and manipulate controlled emotions is the behavior for problem solving and achievement. It has been found that the ability to self-manage emotions is a key factor in stressful situations and overall emotional management (Benson et al., 2012; Faguy, 2012; Rankin, 2013; Ivcevic and Brackett, 2014; Shanta and Gargiulo, 2014; Conley et al., 2015; Görgens-Ekermans et al., 2015; Ranjbar, 2015; Stelnicki et al., 2015; Fitzpatrick, 2016).

Previously, the effects of individual differences such as emotional intelligence or self-monitoring of emotional workers on emotional labor have been studied (Shin et al., 2008; Baeck et al., 2014). Emotional intelligence, the mental ability related to one's emotional experiences (Mayer et al., 2016), was used to identify the cognitive determinants of emotional labor (Lee and Chelladurai, 2018). Also, the ability to perceive emotions, integrate emotion to facilitate thinking, understand, and manage emotions to promote personal growth are paramount to professional achievements and mental health and well-being (Mayer et al., 2016), and emotionally intelligent workers have positive attitudes toward their job and produce optimal performance (Pau and Sabri, 2015).

Emotional intelligence is known to be improved by learning, experience, and training (Gardner, 1995). People with high emotional intelligence tend to be sociable and well-perceived, and those with high emotional control tend to enjoy a high degree of interaction with others. Individuals with high emotional intelligence are better equipped to select effective emotional regulation strategies than those with low emotional intelligence (Wagstaff et al., 2012). Therefore, emotional intelligence should help minimize the negative impact of Surface Acting and enhance the positive effects of Deep Acting.

Laborde et al. (2016) noted that emotional intelligence was a marker of effective emotional regulation strategy. Lee and Chelladurai (2016) included both emotional labor and emotional intelligence in the study, and emotional intelligence was examined as a moderator between emotional labor and emotional exhaustion, rather than considering the direct relationship between emotional intelligence and emotional labor. Developed from these previous studies, Park (2018) examined the moderating effects of the sub-dimensions of Emotional Intelligence between Emotional Labor and Job Satisfaction. Perceiving Emotion and Understanding Emotion showed moderating effects while Managing Emotion and Using Emotion didn't.

Although the importance of emotional intelligence receives more attention, there hasn't been a comprehensive study to investigate the moderating effects of the sub-dimensions of emotional intelligence on Emotional Labor. Therefore, this study would use the Four-Branch Ability Model (Mayer and Salovey, 1997; Wong and Law, 2002; Mayer et al., 2016) to investigate the moderating effects of Emotional Intelligence.

## Perceiving Emotion

The ability to accurately understand and express one's emotions is used to better recognize and recognize one's emotions. In other words, recognizing that emotion when it occurs is a major part of emotional intelligence.

## Understanding Emotion

It is the core of emotional intelligence, and is the ability to understand and experience the emotions of others who are affected by such emotions. In addition, as a harmony of emotions, it is the source of self-awareness and altruism and the basis of moral judgment and action. Those with social competence carry out the functions of maintaining good interpersonal relationships, influencing others, successfully leading human relationships, and facilitating people.

### Managing Emotion

Regulation of emotions forms the background of all accomplishments. The ability to control emotions for oneself is fast restoration ability in psychological frustration, restoration time from joy or anger to normal state, and self- restraint and control ability to control their emotions so that they do not become excited.

### Using Emotion

The ability to utilize emotions to improve performance, and to control emotions in a positive and productive manner. A person with good self-synchronization is motivated by his inner desire for achievement, and is more efficient and productive in some work.

Based on the findings of previous studies, we propose that the previous hypotheses will show different results based on the level of emotional intelligence.

Hypothesis 7: Hypothesis 1 ∼ Hypothesis 6 will be different according to employee's emotional intelligence (Perceiving emotion, Understanding emotion, Managing emotion, Using emotion).

# MATERIALS AND METHODS

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 9

# Research Model

In previous studies, various external variables and parameters as well as moderator variables have been used to understand emotional labor and its related problems. Emotional Intelligence has been used in many studies in order to understand its effect on Emotional Labor at an individual level as well as at an organizational level (Miao et al., 2017). However, a comprehensive model to study the variables' correlation and their overall effect on emotional labor has not yet been established. This study tries to understand the effects of emotion on emotional labor for both employees and customers at the service contact points, and test the role of emotional intelligence at each stage.

In order to accomplish this goal, a SEM would be used. SEM refers to a multivariate model that includes simultaneous equations, factor analysis, and multilevel models, for single and multi-group data. It is the combination of factor analysis and multiple regression analysis, and it is used to analyze structural relationship among variables. In the analysis using SEM, latent variables are created using multiple observed variables through factor analysis. Then, multiple regression analysis is performed on latent variables level.

Therefore, a SEM with Verbal and Non-Verbal Expressions, Customer Feedbacks, Surface Acting, and Deep Acting as observed variables, and Emotional Intelligence and its subdimensions as a moderator variable will be used to investigate the simultaneous relationship among Emotional Exhaustion, Job Satisfaction, Turnover Intentions, Verbal and Non-Verbal Expressions, Customer Feedbacks, Surface Acting, Deep Acting, and Emotional Intelligence.

# Variables

There exist various emotional labor related problems including Employee Well-being and Organizational Wellbeing (Diefendorff et al., 2011; Grandey and Gabriel, 2015), Self-actualization through the interaction, Job Commitment, Professional Efficacy, Turnover Intentions (Quinones et al., 2016), Job Stressor (Mauno et al., 2017), Job Burnout (Chan, 2006; Zhao et al., 2018), Emotional Burden in the Interaction (Hujala and Oksman, 2018), and Work-life balance and Job Satisfaction (Hofmann and Stokburger-Sauer, 2017).

External Factors that are related to emotional labor can be divided into person characteristics related factors and event characteristics related factors. Person characteristics related factor means person-job congruence, and event characteristics related factor means emotion-goal congruence (Grandey and Gabriel, 2015). Person characteristics refer to personality traits, work motives, and emotional abilities, while event characteristics refer to moods, emotions, and customer mistreatment. Emotional Intelligence (Lee and Chelladurai, 2018), Frequency of Emotional Display, Attentiveness to Required Display Rules, Variety of Emotions Required to be expressed, Emotional Dissonance (Morris and Feldman, 1996; Hayyat et al., 2017; Serebrenik, 2017) have also been used. In addition to these factors, two other variables should be taken into consideration in order to fully understand emotional labor: Emotional Regulations and Emotional Intelligence. The former refers to external factors that dictate how employees should express their emotions, and the latter refers to employees' ability to control their emotions and empathize with customers (Grandey and Gabriel, 2015).

Various control variables have been used in previous studies. Kim et al. (2017) used perceived supports from supervisors, co-workers, and organization as control variables. There are also relational factors and contextual factors (Grandey and Gabriel, 2015). Relational factors include emotional traits/abilities, identification/values, and relational power/intimacy, and contextual factors include job status/autonomy, financial reward, and social support. Emotional Intelligence (Lee and Chelladurai, 2016) and Emotion Regulation Self-Efficacy (Deng et al., 2017) have also been used.

This study will follow Grandey and Gabriel (2015) and use Emotion Regulation (Surface Acting and Deep Acting) and Emotion Performance (Verbal and Non-Verbal Communication) at the same time as factors to analyze Emotional Labor. This study will focus on the structural relationship among Emotional Exhaustion, Job Satisfaction, and Turnover Intention with four factors mentioned above and Emotional Intelligence as a moderator variable.

# Hypotheses

Hypothesis 1: Employees' emotional expressions will affect perception of customer feedback.

[H1\_a1] Surface Acting → Customer Feedback (Positive)

[H1\_a2] Deep Acting → Customer Feedback (Positive)

[H1\_a3] Non-Verbal Expression → Customer Feedback (Positive)

[H1\_a4] Verbal Expression → Customer Feedback (Positive)

[H1\_b1] Surface Acting → Customer Feedback (Negative)

[H1\_b2] Deep Acting → Customer Feedback (Negative)

[H1\_b3] Non-Verbal Expression → Customer Feedback (Negative)

[H1\_b4] Verbal Expression → Customer Feedback (Negative)

Hypothesis 2: Employees' emotional expressions will affect emotional exhaustion.

[H2\_1] Surface Acting → Emotional Exhaustion

[H2\_2] Deep Acting → Emotional Exhaustion

[H2\_3] Non-Verbal Expression → Emotional Exhaustion

[H2\_4] Verbal Expression → Emotional Exhaustion

Hypothesis 3: Perception of Customer feedbacks will affect emotional exhaustion.

[H3\_1] Customer Feedback (Positive) → Emotional Exhaustion

[H3\_2] Customer Feedback (Negative) → Emotional Exhaustion

Hypothesis 4: Employees' emotional exhaustion will affect job satisfaction.

Hypothesis 5: Employees' emotional exhaustion will affect turnover intention.

Hypothesis 6: Employees' job satisfaction will affect turnover intention.

Hypothesis 7: Hypothesis 1 ∼ Hypothesis 6 will be affected by employee's emotional intelligence (Perceiving Emotion, Understanding Emotion, Managing Emotion, Using Emotion).

# Sample and Measures

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 10

The survey has been conducted to collect data on emotional labor and emotional intelligence at the service contact points. Survey questionnaires were created with enhancements and supplemental questions utilizing previous studies. The 7-point response format ranged from 1 (strongly disagree) to 7 (strongly agree). A structured questionnaire was used to collect the data from the employees of the service industry using internet survey. Five hundred and seventy nine data was collected through convenience sampling and 577 valid data were used in the analysis. Two invalid data were disregarded.

Industries that are included in this study are Wholesale or Retail Stores (19.8%), Restaurants (19.4%), Telecommunication (14.7%), Education Service (10.6%), and others (35.5%). Mean age of participants of the survey is 34.7 (SD = 10.7). 59.1% of participants were male and 40.9% were female.

Socio-Demographic Characteristics of participants are provided in **Table 1** and the list of Measured Variables are provided in **Table 2**.

## Overview of Statistical Analyses

Data analysis of this study utilized SPSS 24 and AMOS 18 programs. The hypotheses were tested using frequency analysis and confirmatory factor analysis, structural equation, and statistical analysis method of group comparison.

Firstly, descriptive statistics and the reliability of the variables were computed using SPSS 24.0. Then a structural equation modeling technique, which is available through AMOS 18.0, was used to test both the measurement and structural models. Confirmatory factor analysis (CFA) was conducted on the latent variables to check convergent and discriminant validity of the variables, and then the proposed structural relationship was tested by assessing structural coefficients of the relationship among the constructs in the research model. CFA is used to study the relationships between a set of observed variables and a set of continuous latent variables. The measurement model for CFA is a multivariate regression model that describes the relationships between a set of observed dependent variables and a set of continuous latent variables.

Further, the moderating effect of emotional intelligence was tested. Multisample confirmatory factors analysis (MCFA) allowed us to determine if components of the structural model were equal across different groups (Byrne, 2001). Moderating effects of Emotional Intelligence that comprises four TABLE 1 | Respondents' socio-demographic characteristics.


factors (Perceiving Emotion, Understanding Emotion, Managing Emotion, and Using Emotion) were tested using Multi-Group Structural Equation Model (MSEM).

Cross Validation was performed using Multi-Sample Confirmatory Factor Analysis (MCFA). This method allowed us to conduct analyses at multiple levels at the same time in a single analysis. By decomposing the variance of variables into their between-group and within-group components, MSEM accounts for the fact that relationships might be different on the between-group and the within-group levels. Thus, multilevel mediation analyses with MSEM are less prone to biases than

### TABLE 2 | Measurements of variables.

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 11


(Continued)

TABLE 2 | Continued


other techniques of multilevel mediation analysis (Zhang et al., 2009).

Confirmatory factor analysis on the measurement model was performed using AMOS 18.0. Goodness-of-Fit Test on the model using maximum likelihood estimation showed χ <sup>2</sup> = 844.01, df = 426, p-value = 0.00, GFI = 0.91, RMSEA = 0.04, NFI = 0.94, RFI = 0.93, IFI = 0.97, and CFI = 0.97, indicating that the model shows good fit indices of measured variables on overall latent factor. Further, standardized factor loading is greater than 0.6, indicating that it is statistically significant. Detailed result of CFA is provided in **Table 3**.

Construct Reliability (CR) and Average Variance Extracted (AVE) of each variable satisfy general standards (CR > 0.7, AVE > 0.5) and therefore have Convergent Validity (Bagozzi, 1988).

Discriminant Validity can be tested by using AVE-SE comparison (Fornell and Larcker, 1981) using error-adjusted inter-construct correlation derived from CFA. Results provided in **Table 4** show that SE does not exceed AVE and therefore indicating that all constructs showed sufficient discriminant validity.

Then SEM was analyzed as validity and reliability of the variables have been confirmed.

# RESULTS

# Structural Models

The model showed good fit and all indices satisfy the acceptance criteria (χ <sup>2</sup> = 1282.10, df = 439, p-value = 0.00, GFI = 0.88, RMSEA = 0.06, NFI = 0.90, RFI = 0.89, IFI = 0.93, TLI = 0.93, CFI = 0.93).

Test results showed that Deep Acting (Estimate = 0.22, t-value = 3.95, p-value = 0.00) and Non-Verbal Communication (Estimate = 0.66, t-value = 3.11, p-value = 0.00) have statistically significant effects on Perception of Customer Feedback (Positive) while Surface Acting and Verbal Expression failed to show significant effects. Therefore, [H1\_a2] and [H1\_a3] were supported while [H1\_a1] and [H1\_a4] were not supported.

Surface Acting (Estimate = 0.47, t-value = 10.20, p-value = 0.00), Deep Acting (Estimate = 0.33, t-value = 5.65, p-value = 0.00), and Non-Verbal Communication (Estimate = −0.51, t-value = −2.39, p-value = 0.02) all showed statistically significant effects on Perception of Customer Feedback (Negative). Only Verbal Communication failed to show statistically significant effect on Perception of Customer Feedback (Negative). As a result, [H1\_b1], [H1\_b2], and [H1\_b3] were supported while [H1\_b4] was not supported.

When the effects of variables on Emotional Exhaustion were tested, only Surface Acting showed significant effect (Estimate = 0.40, t-value = 8.24, p-value = 0.00). Deep Acting (Estimate = 0.00, t-value = 0.01, p-value = 0.63), Non-Verbal Communication (Estimate = -0.09, t-value = -0.48, p-value = 0.99), and Verbal Communication (Estimate = 0.01, t-value = 0.05, p-value = 0.96) all failed to show significant effects and, thus, [H2\_2], [H2\_3], and [H2\_4] were not supported while [H2\_1] was supported.

Test results support that customers' reactions and feedbacks affect employees' emotional exhaustion as well. Although Perception of Customer Feedback (Positive) didn't show significance (Estimate = −0.01, t-value = −0.14, p-value = 0.89) and, therefore, [H3\_1] was not supported, Perception of Customer Feedback (Negative) showed significance (Estimate = 0.47, t-value = 9.60, p-value = 0.00) for [H3\_2] to be supported.

Employees' emotional exhaustion showed statistically significant effects on both Job Satisfaction (Estimate = −0.23, t-value = −4.70, p-value = 0.00) and Turnover Intention (Estimate = 0.79, t-value = 18.18, p-value = 0.00) and, therefore, both [H4] and [H5] were supported.

Finally, Job Satisfaction does have a significant effect (Estimate = -0.18, t-value = -5.63, p-value = 0.00) on Turnover Intention and therefore [H6] was supported.

Detailed results of the test on each hypothesis are provided in **Table 5** and **Figure 1**.

# Moderating Effects of Emotional Intelligence

Moderating effects of Emotional Intelligence that comprises four factors (Perceiving Emotion, Understanding Emotion, Managing Emotion, and Using Emotion) were tested using MSEM. Cross Validation was performed using MCFA, then multiple group comparison analysis was conducted in order to test overall moderating effect of emotional intelligence level compared to Unconstrained Model (**Table 6**).

The analysis of MSEM of Perceiving Emotion showed that the paths from Deep Acting to Perception of Customer Feedback (Negative) (Critical Ration = −2.95, p-value < 0.01), Surface Acting to Emotional Exhaustion (Critical Ration = −2.30, p-value < 0.05), and Emotional Exhaustion to Job Satisfaction (Critical Ration = −4.19, p-value < 0.00) were significant. This result could be interpreted as that the effects of Deep Acting on Perception of Customer Feedback

### TABLE 3 | Reliability and validity tests.

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 13


Goodness-of-fit: x<sup>2</sup> = 844.01, df = 426, p = 0.00, GFI = 0.91, RMSEA = 0.04, NFI = 0.94, RFI = 0.93, IFI = 0.97, TLI = 0.96, CFI = 0.97.


SA, Surface Acting; DA, Deep Acting; NVE, Non-Verbal Expression; VE, Verbal Expression; CFP, Perception of Customer Feedback (Positive); CFN, Perception of Customer Feedback (Negative); EE, Emotional Exhaustion; JS, Job Satisfaction; TI, Turnover Intentions.

### TABLE 5 | Structural models results.

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 14


Goodness-of-fit: x<sup>2</sup> = 1282.10, df = 439, p = 0.00, GFI = 0.88, RMSEA = 0.06, NFI = 0.90, RFI = 0.89, IFI = 0.93, TLI = 0.93, CFI = 0.93.

(Negative) is bigger for the group with Low Perceiving Emotion, while the effect of Surface Acting on Emotional Exhaustion and the effect of Emotional Exhaustion on Job Satisfaction are greater for the group with High Emotional Intelligence.

The analysis on MSEM of Understanding Emotion showed that the path from Emotional Exhaustion to Job Satisfaction (Critical Ration = −3.60, p-value < 0.00) is significant. The result shows that the effect of Emotional Exhaustion on Job Satisfaction is greater for a group with High Emotional Intelligence.

The analysis on MSEM of Managing Emotion showed that the paths from Deep Acting to Perception of Customer Feedback (Negative) (Critical Ration = −2.26, p-value < 0.05), Verbal Expression to Emotional Exhaustion (Critical Ration = −2.02, p-value < 0.05), and Job Satisfaction to Turnover Intention (Critical Ration = 2.50, p-value < 0.05) were significant. The effects of Deep Acting on Perception of Customer Feedback (Negative) and Job Satisfaction on Turnover Intention were greater for a group with Low Emotional Intelligence, while the effect of Verbal Expression on Emotional Exhaustion was greater for a group with High Emotional Intelligence.

The analysis on MSEM of Using Emotion showed that the paths from Deep Acting to Perception of Customer Feedback (Negative) (Critical Ration = −2.23, p-value < 0.05), Surface Acting to Emotional Exhaustion (Critical Ration = −2.09, p-value < 0.05), and Emotional Exhaustion to Job Satisfaction (Critical Ration = −5.44, p-value < 0.00) were significant. The effect of Deep Acting on Perception of Customer Feedback (Negative) is greater for a group with Low Emotional Intelligence, while the effects of Surface Acting on Emotional Exhaustion and Emotional Exhaustion on Job Satisfaction are greater for a group with High Emotional Intelligence.

# DISCUSSION

In order to provide solutions to the problems regarding Emotional Labor, especially in service industry, two main


**66**

January 2019 | Volume 9 | Article 2526

Customer Feedback (Negative)

> →

> > →

→

Emotional Exhaustion

Emotional Exhaustion

H1\_b4 Verbal Expression

H2\_1 Surface Acting

H2\_2 Deep Acting

H2\_3 Non-Verbal

H2\_4 Verbal Expression H3\_1 Customer Feedback (Positive) H3\_2 Customer Feedback (Negative)

H4 Emotional Exhaustion H5 Emotional Exhaustion

→ Job Satisfaction

→

Turnover Intention H6 Job Satisfaction

→

Turnover Intention

 Expression

→

Emotional Exhaustion

→

→

Emotional Exhaustion

Emotional Exhaustion

→

Emotional Exhaustion

approaches were used in this study. Firstly, we looked at Emotional Labor in the perspective of Emotional Expression regulated by organizational norms. Then we looked at Emotional Intelligence which is individuals' ability to understand and control his/her own emotions. As looking into these two perspectives simultaneously could provide more in-depth understanding of the nature of Emotional Labor, a SEM with various emotional expressions of the employees and the perception of customer feedbacks was tested in order to test how these variables are related to and affecting emotional exhaustion, job satisfaction, and turnover intention.

The results show that both Deep Acting and Surface Acting are valid influential factors to recognize customer's reactions. Deep Acting of employees is a valid factor to recognize customers' positive and negative reactions. It suggests that employees' effort to understand or experience their true feelings is the behavior intending to understand customers' emotion in service encounter. Also, the employees' Surface Acting is a valid factor to recognize customers' negative reactions. In other words, employees' behavior to hide their true feelings and express their emotions according to organizations' regulations and norms affect employees' ability to recognize customers' negative reactions.

Non-Verbal Expression was shown to be a factor that increases Perception of Customer Feedback (Positive) and reduces Perception of Customer Feedback (Negative). This result is consistent with previous findings regarding the characteristics of Non-Verbal Expression. Body language can convey interest, intimacy, benevolent feelings, and trust to customers (Mehrabian and Williams, 1969; Burgoon et al., 1990), and non-verbal communication narrows the psychological distance between employees and customers, increasing positive emotions (Sundaram and Webster, 2000).

The effects of Non-Verbal Communication might vary according to the characteristics of services provided and of customers. Therefore, more detailed studies on the guideline of Non-Verbal Communication that incorporates the characteristics of the service and the customers would be necessary. Further, as online services grow, in-depth study on non-verbal online communication might provide meaningful insights as well.

In addition, it was shown that customers' negative reaction influences the way Surface Acting leads to Emotional Exhaustion. This result supports the previous findings that the employees' Surface Acting negatively affects employees themselves while Deep Acting positively affects employees. Therefore, more systematic and effective approaches that will enable employees to act and behave based on voluntary Deep Acting at the service contact points should be established rather than coercing employees into unconditional Surface Acting in order to find solutions to the problems related to Emotional Labor.

Emotional Exhaustion caused by Emotional Labor and Perception of Customer Feedback lowers employees' Job Satisfaction and increases Turnover Intention, while Job Satisfaction lowers Turnover Intention, which, again, support previous findings. Therefore, in order to lower Turnover Intention, organizations should find ways to increase Job Satisfaction and reduce factors that lead to Emotional Exhaustion.

This study also tested four factors of Emotional Intelligence in order to verify its moderating effects. The results show that it affects Emotional Exhaustion and Job Satisfaction, and hence, Turnover Intention.

The results of the analysis on Perceiving Emotion show that there exists a difference between the effect of Deep Acting on Perception of Customer Feedback (Negative) and the effect of Surface Acting on Emotional Exhaustion. This result suggests that an employee with better understanding of his/her own emotions experience greater negative effect when there is a discrepancy between what they feel and how they should act. Therefore, organizations should collect employees' opinions prior to establishing organizational guidelines of employee behaviors. Behavioral guidelines that incorporate employees' opinions would help reduce Emotional Exhaustion caused by Emotional Labor.

The result of the analysis on Understanding Emotion also suggests that the effect of Emotional Exhaustion on Job Satisfaction depends on the level of Emotional Intelligence. An employee with higher Emotional Intelligence experiences a greater effect on Job Satisfaction from Emotional Exhaustion.

The analysis on Managing Emotion suggests a similar result. It shows that the effect of Deep Acting on Perception of Customer Feedback (Negative) and the effect of Job Satisfaction on Turnover Intention vary depending on the employees' Emotional Intelligence. This result could be interpreted as that an employee with a high Emotional Intelligence is more likely to recover from psychological frustration and emotional depletion and therefore would have lower Turnover Intention.

Lastly, the analysis on Using Emotion showed that the effect of Deep Acting on Perception of Customer Feedback (Negative), the effect of Surface Acting on Emotional Exhaustion, and the effect of Emotional Exhaustion on Job Satisfaction vary depending on Emotional Intelligence. This result confirms that the ability to utilize his/her own emotions would enhance Job Performance of employees.

# Contributions and Implications

There are two main perspectives of this research. The first is to look for solutions through the governing principles that control workers' emotional expressions. The second is to look for solutions through individually controlled emotional intelligence. Utilizing these two perspectives that are independent yet interrelated, the more thorough and effective solutions to the problems regarding Emotional Labor could be suggested.

The findings of this study regarding the effects of Emotional Expression are consistent with previous findings. The study showed moderating effects of the sub-dimensions of Emotional Intelligence on Emotional Labor and how they affect structural relationships among variables of Emotional Labor. Based on the findings of this study, two suggestions can be made.

Firstly, organizations in the service industry should provide training programs for their employees to improve their communication skills with the customers at the service contact points. The program should be divided into two categories:

Verbal and Non-Verbal. As the importance of Non-Verbal communication was emphasized in the result, the training program to improve employees' non-verbal communication skills would be more important. Further, it should be noted that communication within the organization would be as important as the communication between employees and customers.

Secondly, it would be important to actively include customers in the communication process at the service contact points. The characteristic of non-separability in service could be utilized to include customers in the communication process. Redesigning the service process and introducing various motivating incentives may enhance customer participation as well. By including customers more actively in the service process, organizations could build more positive and trusting emotions with customers.

The findings of this study would suggest the solutions that could enhance job satisfaction of the workers in the service industry with higher turnover rate, and in turn will increase the job security. The sustainability of workers' lifestyle as well as the quality of their lives are, therefore, expected to be enhanced. Further, more effective management of the emotional labor would help organizations establish more positive organizational culture and enhance effectiveness of job performances, which will eventually increase the organizations' profitability.

Lastly, this research shall provide meaningful results that confirm the importance of individual emotional intelligence. These results will help the organizations and individual workers understand how enhancing emotional intelligence will strengthen their adaptiveness to rapidly changing socioeconomic environments.

## Limitations and Future Directions

This study has a few limitations. Firstly, as this study focused on showing the moderating effects of the sub-dimensions of Emotional Intelligence, the data used in this study didn't take industry-specific characteristics into account. Future studies may collect industry-specific data, such as banking or hospitality industry, in order to provide more in-depth findings of the moderating effects of emotional intelligence on how emotional expressions and perception of customer feedbacks affect employees' Turnover Intention.

Further, the data used in this study was collected in the United States only and hence make it difficult for us either to generalize our findings or to conduct cross-country comparisons. Future studies may replicate the findings of this study with data collected from other countries or regions to enable cross country comparisons of the moderating effects of emotional intelligence.

Lastly, only the existence of the moderating effects of emotional intelligence was examined in this study. How

## REFERENCES


Argyle, M. (1975). Bodily Communication. New York, NY: Methuen.

employees' emotional intelligence can be enhanced and thus help them manage emotional exhaustion should be more thoroughly studied, maybe in cultural contexts as well, in order to provide meaningful solutions to increase organizational effectiveness and sustainability. In order to do this, future studies may examine how emotional intelligence affects a person's ability to empathize and to communicate.

# CONCLUSION

The result of this study provides the support for moderating effect of emotional intelligence on how employees' Deep Acting and Surface Acting recognize customers' reactions that would affect employees' Emotional Exhaustion and Job Satisfaction, and hence, Turnover Intention. The result suggests that an employee with better understanding of his/her own emotions experience greater negative effect when there is a discrepancy between what they feel and how they should act.

Emotional Intelligence affects employees and the organizations in various ways. Although an employee with higher Emotional Intelligence experiences a greater effect on Job Satisfaction from Emotional Exhaustion, the one with a high Emotional Intelligence is more likely to recover from psychological frustration and emotional depletion and therefore would have lower Turnover Intention.

In addition, the ability to utilize his/her own emotions would enhance Job Performance of employees.

## ETHICS STATEMENT

An ethics approval was not required as per institutional guidelines and national laws and regulations because no unethical behaviors existed in this study. We just conducted paper– pencil test and were exempt from further ethics board approval since this research did not involve human clinical trials or animal experiments. All subjects gave written informed consent in accordance with the Declaration of Helsinki. Research respondents were ensured confidentiality and anonymity. All participation was voluntary.

# AUTHOR CONTRIBUTIONS

YL and SL contributed to design models and hypotheses. YL and JC acquired, analyzed, and interpreted the data. YL and SL wrote and revised the article. YL, JC, and SL approved the final version.



Hoppock, R. (1935). Job Satisfaction. New York, NY: Harper.

fpsyg-09-02526 December 22, 2018 Time: 12:37 # 19




**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Lee, Lee and Chung. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Do Emotional Laborers Help the Needy More or Less? The Mediating Role of Sympathy in the Effect of Emotional Dissonance on Prosocial Behavior

### Yun-na Park<sup>1</sup> , Hyowon Hyun<sup>1</sup> and JiHoon Jhang1,2 \*

<sup>1</sup> School of Business, Hanyang University, Seoul, South Korea, <sup>2</sup> Spears School of Business, Oklahoma State University, Oklahoma, United States

Despite the growing body of research on emotional labor, little has been known about

the social consequences of emotional labor. Drawing on emotional dissonance theory, the authors investigate the relationship between the felt emotional dissonance and prosocial behavior (e.g., donation to a charity). Findings from multiple studies suggest that higher emotional dissonance serially influences perceived lack of control, emotional exhaustion, lowered sympathy for others' feeling, and subsequently lower willingness to help others. When individuals are asked to recall their past experiences of emotional dissonance, they expressed lack of control and emotional exhaustion (Study 3), lower sympathy for others' feeling (Studies 1, 3), and subsequently become less willing to help others both in their intention (Studies 2A and 3) and with actual money (Study 2B). Further, this negative relationship is moderated by the display rules (i.e., surface acting vs. deep acting, Study 3). Managerial and public policy implications are discussed.

Keywords: emotional labor, emotional dissonance, surface acting, deep acting, sympathy for others' feeling, emotional exhaustion, prosocial behavior

# INTRODUCTION

According to the Bureau of Labor Statistics (2016), about 80% of total U.S. workers are employed by the service sector, and employment in the service sector is projected to increase over 12% between 2008 and 2018 (Bureau of Labor Statistics, 2016). Situations in other developing countries do not differ much (The World Bank, 2016). As the proportion of service-sector employees increases, it becomes more important to understand and manage these service-sector employees.

One characteristic peculiar to service-sector employees is their frequent contacts with customers. Therefore, previous research has emphasized the importance of managing these employees because evaluation of a service firm often depends on how customers perceive the interaction they have with the frontline employees (Bitner, 1990; Abraham, 1998). For instance, Babakus et al. (2009) showed that frontline employees play a crucial role not just in the service delivery but also in customer relationship building. More important, research shows that the expressed emotions of service providers heavily influence customers' perceptions of service quality (Abraham, 1998). For this reason, many organizations in the service industry attempt to control the way their employees display their emotions. Consequently, the service-sector employees are very often required to

### Edited by:

Weon Sang Yoo, Korea University, South Korea

### Reviewed by:

Jin Ho Jung, Ohio Northern University, United States Jungkeun Kim, Auckland University of Technology, New Zealand

\*Correspondence:

JiHoon Jhang jihoonjhang@hanyang.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 18 October 2018 Accepted: 14 January 2019 Published: 07 February 2019

### Citation:

Park Y-n, Hyun H and Jhang J (2019) Do Emotional Laborers Help the Needy More or Less? The Mediating Role of Sympathy in the Effect of Emotional Dissonance on Prosocial Behavior. Front. Psychol. 10:118. doi: 10.3389/fpsyg.2019.00118

**73**

express certain emotions different from their true ones. The term 'emotional labor,' first coined by Hochschild (1979, 1983) was originated from describing a situation wherein employees are required to display emotions that may differ from what they truly feel at the moment especially in the context of client contacts.

Despite its positive impact on customers' perceptions of service quality, the enforced emotional display rules are known to have negative impacts on both individual and organizational well-being (Grandey, 2000). To name a few, a body of research in emotional labor consistently found that emotional laborers tend to have lower identification with the organization, lower job involvement, lower job satisfaction, higher work stress, and lower well-being (Pugliesi, 1999; Schaubroeck and Jones, 2000; Cho et al., 2013). Because of its significance, much research efforts have been devoted to examine the antecedents, dimensions, and consequences of emotional labor (e.g., Rafaeli and Sutton, 1987; Morris and Feldman, 1996; Groth et al., 2009).

Common in the aforementioned works is that most research has focused only on the outcome variables related to organizations' performance. That is, the majority of works has been done from the perspectives of organizations. In spite of its practical and theoretical significance of this topic, there is a dearth of research on this topic from consumers' perspectives. To our best knowledge, no research efforts have been devoted to examine the behaviors of emotional laborers as consumers. Although this may not be directly related to organizational well-being such as job performance or job retention rate, we argue that it would have even bigger implications to the society given that the majority of employees are to some extent considered as emotional laborers (Bureau of Labor Statistics, 2016). Therefore, it is very timely and important to investigate how these emotional laborers behave as consumers. For instance, how these people would treat other service workers (e.g., emotional laborers) when they become customers? Would they be sympathetic for others' feeling more or less? why? And what would be the consequences of that? These are the questions of interest the present research would address.

The purpose of this research is to understand the psychology of emotional labor and its consequences beyond the organizational boundary. Specifically, we examine how emotional laborers would differ in their degree to feel sympathy toward others (or others' feeling) and how this difference results in their pro-social intention and actual pro-social behavior. By examining these, we would like to shed light on the societal consequences of emotional labor, which has been neglected in the previous literature.

# THEORETICAL BACKGROUND

# Emotional Labor as Regulatory Behavior

Since Hochschild (1983) first coined the term 'emotional labor' as "the management of feeling to create a publicly observable facial and bodily display" (p. 7), many researchers have attempted to grasp the concept of emotional labor by providing their own definition for the term. For instance, Ashforth and Humphrey (1993), focusing more on the observable behavior rather than the management of feeling, defined emotional labor as "the act of displaying appropriate emotions with the goal to engage in a form of impression management for the organization" while Morris and Feldman (1996) defined it as "the effort, planning, and control needed to express organizationally desired emotion during interpersonal transactions." Grandey (2000) defined emotional labor as the process of faking and suppressing their true emotions to follow the guidelines imposed by organizations.

Although the exact wording varies by scholars, what is common among various definitions is the notion of emotion regulation for the organization; emotional laborers must regulate their emotions to achieve bigger goals (i.e., organizational goals). Therefore, Diestel and Schmidt (2011) think emotional labor as a goal-directed regulatory behavior.

# Deep Acting and Surface Acting

One important (and implicit) assumption implied by the term 'emotion regulation' is that some emotions have to be regulated because, without being regulated, they might not be appropriate from the perspectives of organizations. For instance, if there is a conflict between the emotions an employee wants to express and the emotions an organization wants their employees to express, the employee must regulate emotions.

According to Gross (1998), people can regulate emotions in two ways; focusing on either the precursors of emotions (e.g., the situation) or the observable signs of emotions (e.g., facial expressions). In the first emotion regulation technique, employees adjust their emotional responses to the situation by modifying the way they perceive the situation. By either thinking about events that call up specific emotions needed in the situation or reappraising the objects or situations, employees try to modify their thoughts and feelings with the goal to make the expression more genuine. This type of emotion regulation process corresponds to the emotional labor concept of 'deep acting' (Hochschild, 1983).

By contrast, in the second emotion regulation technique, employees just manipulate emotional expressions of their responses to the situation without adjusting the perception of the situation. By either faking the expressions entirely or adjusting the intensity of the displayed emotions, employees try to modify their expressions without changing their thoughts or feelings. This type of emotion regulation process corresponds to the emotional labor concept of 'surface acting' (Hochschild, 1983).

When an employee is engaged in this second type of emotion regulation technique (i.e., surface acting), the employee recognizes that the emotion he/she expressed differs from the emotion he/she actually felt, and consequently feels emotional discrepancy. This perceived discrepancy is called 'emotional dissonance' (Hochschild, 1983; Abraham, 1998; Kruml and Geddes, 2000). In the literature on emotional labor, it is a widely held notion that surface acting is linked with emotional dissonance (e.g., Ashforth and Humphrey, 1993; Deng et al., 2017).

# Emotional Dissonance and Perceived Control

Emotional dissonance, in many aspects, is analogous to cognitive dissonance (Festinger, 1957). Specifically, the following three propositions are noteworthy. According to the cognitive dissonance theory (Festinger, 1957), (1) people are sensitive to the discrepancy between their belief and action, (2) people, when recognizing the discrepancy, are motivated to resolve the discrepancy, and (3) people, in attempting to resolve the discrepancy, usually trace back to why the discrepancy arose (e.g., Ben, 1967). When we apply these to the context of emotional labor, an employee engaged in surface acting can recognize the discrepancy between their felt and expressed emotions and be motivated to resolve the discrepancy. In attempting to resolve the discrepancy, the employee would trace back to why the perceived emotional discrepancy arose. The reason is obvious; they had to follow the display rule enforced by the organization. Therefore, we argue that this reasoning leads the employee to perceive that she has no control over the way she expresses her true feeling, which may result in a generalized lower sense of control<sup>1</sup> (or perceived lack of autonomy).

Organizational behavior literature on the perceived control has equivocally reported positive relationship between high levels of perceived control and job-related variables (e.g., Spector, 1986; O'driscoll and Beehr, 2000; Thompson and Prottas, 2005). For instance, Spector (1986) showed that employees' high levels of perceived control positively affect not only organizational (e.g., job satisfaction, commitment, involvement, performance, and motivation) but also individual facets (e.g., emotional distress, role stress, absenteeism, and turnover rate). In a similar vein, Thompson and Prottas (2005) found that job autonomy and perceived control are positively associated with numerous organizational outcomes. These findings imply that lower (vs. high) perceived control that results from emotional dissonance, would have negative impacts on job-related variables.

More relevant to our theorizing are the findings by Muraven et al. (2007) that demonstrate the relationship between the lowered sense of control and resource depletion. In their studies, Muraven et al. (2007) randomly assigned employees to either controlling situation (i.e., working environment where everything was structured so that no autonomy is allowed) or autonomy supportive situation (i.e., working environment where maximum autonomy is allowed) and examined employees' degrees of depletion. The results show that people tend to be depleted more when they were working in the controlling situation. This suggests that when people perceive lack of control over the situation they are working in, they are more likely to be depleted. Taken together, we argue that emotional dissonance leads to resource depletion via lowered perceived control.

# Emotional Exhaustion and Sympathy for Others' Feeling

Although the role of perceived control has not been explicitly highlighted in the previous literature, the relationship between emotional dissonance and resource depletion is not a new notion. Rather, emotional dissonance has been regarded as a major source of ego-depletion (Baumeister et al., 2007; Deng et al., 2017). The reason behind this relationship is as follows. According to the self-regulation resource theory (Muraven et al., 1998; Vohs and Heatherton, 2000), people use some types of resources to regulate themselves. If an individual is involved in any kinds of regulatory behavior, this uses the resources and leads to depletion (e.g., Schmeichel et al., 2003; Vohs and Schmeichel, 2003). In the context of emotional labor, faking or suppressing emotions requires considerable self-regulatory resource expenditure, which leads to depletion.

Many emotional labor researchers have also examined the relationship between emotional dissonance and a construct very similar to resource depletion under different names such as emotional exhaustion, burnout, fatigue, and energy depletion (e.g., Gross, 1998; Grandey, 2000). Largely, two lines of research stream are noteworthy. In one line of research, emotional dissonance has been found to mediate the effect of emotional labor on emotional exhaustion (e.g., Lewig and Dollard, 2003). In another line of research, emotional exhaustion has been found to mediate the effect of emotional dissonance on job-related outcomes (e.g., Karatepe and Aleshinloye, 2009). These findings together lend a strong support to the relationship between emotional dissonance and emotional exhaustion.

One subtle difference between self-regulation and emotional regulation theory lies in at which level it defines resource depletion (or emotional exhaustion). The latter theory defines emotional exhaustion as the physiological responses while the former theory includes not only physiological but also psychological states. In this research, we take the perspectives of the former.

Then, what would be the consequences of emotional exhaustion? We argue that when people are emotionally exhausted, they have no emotional resources they can use in the interaction with others. As a result, they would be less likely to feel sympathy for others' feeling. Taken together, we propose that if an individual experiences emotional dissonance, they would feel less sympathy toward others. It is formally stated as,

H1: Those experiencing emotional dissonance (vs. not) would be less likely to feel sympathy for others' feeling.

# Sympathy and Prosocial Behavior

What, then would be the consequences of this lowered sympathy? A body of literature on prosocial behavior shows the strong linkage between individual's sympathy (or empathy) and prosocial behavior (Eisenberg et al., 1989; Eisenberg and Fabes, 1990). For example, Eisenberg et al. (1989) measured the degree to which one feels sympathy toward others using various ways including facial expression, physiological index (e.g., heart-rate), as well as self-report. Each measure was shown to be a strong predictor of one's prosocial behavior, supporting the notion that

<sup>1</sup> It should be noted that this lowered sense of control would be observed only in the employees engaged in surface acting. If an employee is engaged in deep acting, he/she would not recognize any discrepancy between the felt and expressed emotions (i.e., no emotional dissonance). Therefore, they would perceive that they have control over the way they express their true feeling, which results in enhanced sense of control. In fact, previous research has suggested that deep acting promotes a sense of control (e.g., Brotheridge and Lee, 2002; Scott and Barnes, 2011).

sympathy and prosocial behavior is strongly linked. Further, Jolliffe and Farrington (2006) administered the 20-item Basic Empathy Scale (BES) to 363 K-10 adolescents (aged about 15) and found that empathy was positively related to prosocial behavior and a lack of empathy associated with aggressive and antisocial behavior. A corollary hypothesis of these finding, along with our H1, is that emotional dissonance would have negative impacts on prosocial behavior. Therefore, we argue that, if people experience (or recall their experiences of) emotional dissonance, they would be less likely to act pro-socially because of lowered sympathy from feeling emotional dissonance. This argument is formally hypothesized as,

H2: Those experiencing emotional dissonance (vs. not) would be less likely to act pro-socially.

H3: The effect of emotional dissonance on prosocial behavior will be mediated by sympathy for others' feeling.

# Differential Effects of Surface and Deep Acting on Sympathy and Prosocial Behavior

If our proposed effects are driven by emotional dissonance, emotional laborers engaged in deep acting (vs. surface acting) would not exhibit the same pattern of behavior (i.e., lowered sympathy and lowered willingness to act pro-socially). For deep acting, by definition, is a strategy to adapt one's inner thoughts and feeling (Ashforth and Humphrey, 1993), it is less likely to trigger emotional dissonance because, for those engaged in deep acting, the discrepancy between felt and expressed emotions would be minimal. This would subsequently lead to higher sense of control, and less emotional exhaustion. Hence, we predict that those taking deep acting strategy would be more likely to feel sympathy for others' feeling (than those taking surface acting strategy) and thus results in higher prosocial behavior intention. This hypothesis is formally stated as,

H4: Different emotion regulation strategies would have differential effects on prosocial behavior intention; those engaged in deep (vs. surface) acting would be more (vs. less) likely to feel sympathy for others' feeling, and subsequently exhibit higher (vs. lower) willingness to act pro-socially

# Overview of Studies

In a series of multiple studies, we tested our hypotheses. We first report the results of a study that examines the relationship between emotional dissonance and the degree to which one feels sympathy for the needy (Study 1). Then, we report two studies that establish the basic effect. We either manipulated (Study 2A) or measured (Study 2B) individual's emotional dissonance and found that, the greater dissonance one feels, the less one is willing to help others (Study 2A) or the less money one actually donates to a charity (Study 2B). Finally, we tested if different emotion regulation strategies (i.e., deep acting vs. surface acting) would have differential impacts on prosocial intention (Study 3). Additionally, more refined process measures (e.g., sense of control and emotional exhaustion) are tested with serial mediation tests (Study 3).

# STUDY 1 DOES FEELING EMOTIONAL DISSONANCE LEAD TO LESS SYMPATHY?

The purpose of Study 1 is to examine the relationship between emotional dissonance and sympathy for others' feeling. As we stated in our theorizing, previous research predicts that those experiencing emotional dissonance would become less sympathetic for others' feelings. We test this hypothesis in this study.

# Methods

A total of 201 participants (Mage = 36.06, 41.8% Women) who claimed themselves as service workers were recruited through Amazon's mTurk. We divided the participants into two groups and varied the degree to which one feels emotional dissonance. Thus, the experimental design was a simple factor (Emotional dissonance vs. Control) between-subjects design. Participants' emotional dissonance was manipulated with a writing task. Specifically, we asked half of the participants to vividly recall and write about their own experiences where they had to hide their true feelings because of the guidance imposed by their employer (see **Appendix A**). This instruction was carefully crafted by combining the definitions of both emotional dissonance and emotional labor proposed by previous literature (Hochschild, 1983). For the other half of the participants, we asked them to write about the place in which they were at the moment (Nikolova et al., 2017; see **Appendix A**). This group of participants serve as a control group. There were no requirements for or limits on the length or content, but we forced the participants to write at least for one and half minutes to proceed to the next page in the survey by hiding the next button. All participants wrote at their pace, and the average time they spent on this writing task was 2 min and 21 s.

After completing the writing task, all participants were informed that the first study was done. Then, they completed a filler task described as an unrelated second study. Following the filler task, participants answered to a battery of demographic questions into which we ostensibly inserted the six-item 'Sympathy for the Feelings of Others' scale (Lee, 2009, see **Appendix B** for more details). This scale was measured on 7 point Likert scale (1 = Strongly disagree, 7 = Strongly agree) and serves as our dependent variable. Any negatively worded items (e.g., I really don't get emotional when I see people crying) were reverse-coded in all of the analyses throughout the paper. Thus, the higher score means that people feel more sympathetic for others' feeling.

# Results and Discussion

Independent t-test results showed that there is a significant difference in the sympathy for others' feeling scores (Cronbach's α = 0.76) between the two groups (emotional dissonance vs. control). Participants in the emotional dissonance group reported a lower sympathy score (M = 3.94, SD = 1.30) than those in the control group [M = 4.50, SD = 1.28, t(199) = −3.06, p < 0.01]. This result is consistent with and thus support H1

that people experiencing emotional dissonance would be less likely to feel sympathy for others' feeling. The result of Study 1 provides initial evidence for the notion that emotional dissonance negatively influences sympathy for others' feeling. In the next study, we examine the consequence of this lowered sympathy – prosocial behavior.

# STUDY 2A DOES FEELING EMOTIONAL DISSONANCE LEAD TO LOWER WILLINGNESS TO HELP OTHERS?

The objectives of Study 2A are twofold. First, we examine if there exists a hypothesized relationship between emotional dissonance and prosocial behavior. We predict that if participants feel emotional dissonance, they would be less likely to act prosocially. Second, we provide process evidence. We propose that lowered sympathy for others' feeling would mediate the effect of emotional dissonance on prosocial behavior.

# Methods

Two hundred and five participants were recruited from Amazon's mTurk (Mage = 38.31, 52.7% Women). As in Study 1, we randomly assigned participants to two experimental groups (Emotional Dissonance vs. Control). Thus, the design of Study 2A is again a single factor between-subjects design. However, different from Study 1 where only service workers were allowed to participate, we imposed no restrictions on the participants' qualifications in Study 2A for convenience.

Prior to the study, all the participants were informed that they would be participating in several short studies. The first study was introduced as a writing task. We used the same writing task as in Study 1 to manipulate emotional dissonance. We did not, however, hide the next button nor did measure the time spent on the writing task.

Following the writing task, participants proceeded to the next stage framed as an unrelated second study. In this stage, participants were asked to view two fictitious posters of a nonprofit organization. Each poster highlights different causes. One poster depicts a group of children in the classroom, soliciting for help on educating children in underdeveloped countries. The other poster portrays people looking over the area seriously damaged by the earthquake, appealing for help on recovering from a disaster. To make them look real, we put the UNICEF logo in the corner of both posters (see **Appendix C**).

Participants were presented with one poster at a time. Along with the poster, we measured an individual's willingness to act pro-socially using the following two questions; 'how much would you be willing to donate?' and 'how much would you be willing to share the poster on social media like Facebook or Instagram?' (1 = Very unlikely, 7 = Very likely). Given that prosocial behavior includes a broad category of acts such as helping, sharing, donating, cooperating, and/or volunteering (Brief and Motowidlo, 1986; Penner et al., 2005), there are various ways to measure prosocial behavior. We, however, intentionally chose the above two items for reasons. Among many different forms of prosocial acts that maintain or produce the well-being of others, donation (i.e., giving one's monetary resources to benefit others) is considered as a way that often entails the highest monetary cost to the self (Twenge et al., 2007). In contrast, sharing or click 'like' on social media is regarded as a way to help and support others, with minimal monetary, emotional, and temporal cost to the self (see Wang et al., 2017). Therefore, with these two measures, we hope to investigate the effect on emotional dissonance on the full span of prosocial behaviors.

Finally, we collected the same six-item sympathy for others' feeling scale (Lee, 2009) within the demographic questions as we did in Study 1. We predicted that, compared to the control condition, participants in the emotional dissonance condition would exhibit lower sympathy for others' feeling and lower willingness to act pro-socially (i.e., less willing to donate and less willing to share the posters on Social Media).

# Results

Participants in the emotional dissonance condition reported lower sympathy for others' feeling (M = 4.59, SD = 1.62) than those in the control condition [M = 4.96, SD = 1.43, t(203) = 1.74, p < 0.05, one-tailed], replicating the results of Study 1. Next, we separately analyzed the willingness to donate and the willingness to share scales, but no significant differences emerged. Thus, we averaged the two items to create a composite 'willingness to help others' measure (r = 0.723). Results showed that, compared to the control condition (M = 4.04, SD = 1.93), participants in the emotional dissonance condition indicated lower willingness to help others [M = 3.53, SD = 2.01, t(203) = 1.81, p < 0.05, onetailed]. The results do not statistically vary when we analyzed the data separately for each poster or using the repeatedmeasures ANOVA. In either case, when participants recalled their experiences of emotional dissonance, they expressed lower willingness to help providing children with better education and recovering the areas damaged by a natural disaster (see **Table 1**). Taken together, the results are consistent with our prediction and thus support H2.

Finally, a mediation analysis (PROCESS model 4 with 5,000 bootstrapped samples, Hayes, 2018) with emotional dissonance as the independent variable, 'sympathy for others' feeling' as the mediator, and the composite measure of 'the willingness to help others' as the dependent variable yielded significant mediation via 'the sympathy for the feeling of others' (The Bootstrapped Indirect Effect = −.2174, 90% CI = 0.0101, 0.4464) (See **Figure 1**). This suggests that the negative effect of emotional dissonance on the willingness to help others is mediated by (lowered) sympathy for others' feeling, which provides support for H3.

# Discussion

The results of Study 2A provide initial evidence for the notion that people experiencing emotional dissonance are less willing to help others. Specifically, we used two different causes (i.e., providing children with better education and recovery from a natural disaster) and two different ways to help (i.e., donation and sharing on social media) to examine the willingness to help others, and the results did not vary by the causes nor by the types of acts, which together increases the generalizability of our findings. Moreover, in Study 2A, we examined the underlying

### TABLE 1 | Sympathy for others' feeling.

fpsyg-10-00118 February 5, 2019 Time: 17:13 # 6


Higher number means higher sympathy. All p-values were calculated with one-tailed test.

mechanism and found that this emotional dissonance effect is mediated by the lowered sympathy.

It should be, however, noted that the effect sizes reported here are rather weak. We conjecture that this might be due to the fact that we did not impose any restrictions on the qualifications of potential participants. That is, we allowed anyone to participate in our study. Therefore, if a non-service worker was asked to recall their experiences of emotional dissonance, he/she might not be able to vividly recall them, which subsequently may affect our manipulation of emotional dissonance. The data on employment status shows that among 205 participants, only 79% (162/205) were the paid employees (not all of them were service workers), while 14.1% (29/205) and 6.9% (14/205) of the participants were self-employed and not working, respectively. Although we have no data on specific industries in which the participants are working, we posit that the proportion of service workers was lower than 79%, which might contribute to our weak effect sizes. Therefore, hereafter throughout the remaining studies, we collected the data only from the service-workers.

Although the results suggest that emotional laborers (vs. nonemotional laborers) would be less likely to help others, our results in Study 2A are solely based on a self-reported willingness rather than actual behavior. Of course, it is not always required to examine actual behavior in order to generalize the findings on willingness to real behavior because there is a widely accepted notion that attitudes and intention are the precursors of behavior (e.g., Ajzen and Fishbein, 1977). Nevertheless, especially in a prosocial behavior domain, there exists a reason to be more careful in drawing conclusion. Many researchers found that the link between people's willingness to help and their actual helping behavior is weak (Yoon and Kim, 2018). This implies that nonemotional laborers (i.e., participants in the control condition) who expressed greater willingness to help others actually may not help others more than emotional laborers (i.e., participants in the emotional dissonance condition). Then, the observed between-group difference in the willingness to help others might not be that evident in the real environment, thus practical implications are limited. Therefore, it would be beneficial to examine if the observed emotional dissonance effect can be demonstrated not just in prosocial intentions but also in actual prosocial behavior.

# STUDY 2B. DO EMOTIONAL LABORERS ACTUALLY HELP OTHERS LESS?

The purpose of Study 2B is to examine actual prosocial behavior of emotional laborers. For this, we developed a simple experimental paradigm where participants are unexpectedly provided with an opportunity to earn a small amount of money and then determine what percentage of that money they would like to donate to a charity. We employed actual donation behavior because it would be a strong test; the money is real, and the participants have economic incentives not to donate. Therefore, using this experimental paradigm, if we find a difference in the amount of donation by the degree to which one feels emotional dissonance, it would lend more support to our proposition that emotional dissonance negatively affects prosocial behavior.

# Methods

To test if our proposed emotional dissonance effect holds in a situation where the real money is involved, we piggybacked our study on another unrelated survey. We covertly asked three sets of questions at three locations of the unrelated survey; the eligibility questions at the beginning, emotional dissonance questions in the middle, and an extra money offer question at the end. We described these questions more in detail below.

First, the eligibility questions consist of two items about employment status and the specific job functions. If respondents are paid-employees and work in the service-related industry,

they are the potential subjects qualified for our study. Second, emotional dissonance is measured with the two-item emotional dissonance questions (Kruml and Geddes, 2000, see **Appendix B** for more details). Only the qualified respondents saw and answered to the questions on a 5-point scale (1 = never, 5 = always). Finally, the extra money offer question was presented only to them when they reached the last page of the survey after getting the mTurk completion code for the original unrelated survey. The message was titled as 'Chance for additional \$1!' and asked if respondents want to participate in a 1-min short survey for an extra \$1. It was completely their choice to accept the offer. If one wants to participate, they can continue by clicking the 'yes' button. Otherwise, they can just opt out by clicking the 'no' button.

Once participants have agreed to continue, they were instructed that the purpose of the short survey is to examine how much people would like to donate to a certain non-profit organization. Specifically, they read that they can pledge to donate any percentages of their \$1 bonus payment that will be actually matched by us, and their bonus payment will be determined by deducting their pledged amount from the original \$1. For instance, if a participant decides to donate 40% of the bonus money (i.e., \$.4), a total of \$.8 (after we match the amount) will be donated to the charity, and the participant will be receiving \$.6 (i.e., \$1–\$.4). It was emphasized that they can freely choose any percentages between 0 and 100%. Moreover, they read, not to be compelled to look nice, and it would be completely their choice. On the following page, participants viewed a modified version of the UNICEF poster used in Study 2A that solicits donations for children in the underdeveloped country. Participants then indicated what percentages they would like to donate to the charity and what percentages they would like to keep to themselves, respectively, which should sum to 100%.

Participants were recruited through Amazon's mTurk. Among 1133 mTurkers who accepted the HIT, 301 respondents claimed themselves as the paid-workers in the service industry, thus passed our predetermined criteria. Only these respondents answered to the two-item emotional dissonance questions in the middle of the unrelated survey. Subsequently, it was only to these respondents that the extra money offer question was shown at the end of the unrelated survey. Two hundred forty-four respondents agreed to participate in this extra short survey. Therefore, all of the analyses in Study 2B hereafter are based on these 244 participants (Mage = 31.36, 31.1% Women).

# Results

We averaged the two emotional dissonance scales (r = 0.704) to create a composite measure. Then, it was reverse-coded so that the higher score means the higher felt emotional dissonance. To test our hypothesis on the relationship between emotional dissonance and actual donation behavior, we regressed the pledged percentages to donate (0%∼100%) on emotional dissonance. The result of our regression analysis was consistent with H2. Specifically, those who reported experiencing higher emotional dissonance pledged lower percentages to donate [Standardized b = −0.189, t(242) = −2.998, p < 0.01]. Thus, H2 was supported again in a setting where actual prosocial behavior is engaged.

Although we clearly understand the problems of dichotomizing a continuous independent variable (Fitzsimons, 2008), given that our regression analysis was significant, for illustrative purpose only, we median-split our independent variable (i.e., emotional dissonance) to run ANOVA. A singlefactor between-subjects ANOVA yielded the same result; those experiencing high (vs. low) emotional dissonance actually donated less [Mhigh = 26.45%, Mlow = 35.95%, F(1,242) = 6.945, p < 0.01].

# Discussion

In Study 2B, we replicated the findings of Study 2A that emotional dissonance decreases prosocial behavior. The study design of Study 2B, however, is quite different from Study 2A in several aspects. First, the sample was carefully selected. In contrast to Study 2A where the sample was drawn from the general population, in Study 2B, we drew our sample from the specific group of people that fits our purpose better; paid-employees working in the service-related industry.

Second, different from the previous two studies where we manipulated emotional dissonance, we measured emotional dissonance in Study 2B. In Studies 1 and 2A, we used the recall-based writing task to manipulate emotional dissonance. Despite our efforts to carefully craft the instructions for emotional dissonance manipulation, the ad-hoc data analysis shows that many participants failed to recall and write about their experiences in the work context. Specifically, only 31.2 and 32.7% of participants in the emotional dissonance conditions in Studies 1 and 2A, respectively, described their experienced emotional dissonance in the work place, but the remaining 68.8 and 67.3% described general emotional dissonance experienced in the context of social interaction. Although we believe that it may not debunk our arguments nor change the implications of our findings, we should admit that some constructs (e.g., emotional dissonance) are harder to manipulate than other constructs, and thus manipulation may not be the best way to conduct research on emotional dissonance.

Finally, we used the real money in Study 2B to test our hypothesis. By demonstrating that the results still hold when participants were asked to donate their real money, we showed that our hypothesized effect is very robust. One thing to note here is, the final set of subjects in our Study 2B (n = 244) was those who had higher desire for additional money compared with those who had opted out. Therefore, it is reasonable to assume that our sample consists of those the least willing to donate real money. Nevertheless, the result was consistent with our hypothesis; people experiencing higher emotional dissonance pledged to donated 26% less amount of money to the needy.

So far, in three studies, we have demonstrated that those experiencing emotional dissonance are less willing to help the needy, and this effect is mediated by lowered sympathy for others' feeling. Then, can we conclude that emotional laborers help the needy less? What would happen if an emotional laborer does not experience emotional dissonance much? Does he/she still help others less or more? Our theory suggests that only those

experiencing emotional dissonance would be less likely to help others. Therefore, if one is engaged in an emotion regulation strategy that does not lead to emotional dissonance (i.e., deep acting), he/she would not exhibit lowered sympathy nor lowered willingness to help others. We test this idea in the next study.

# STUDY 3. DIFFERENTIAL EFFECTS OF DIFFERENT EMOTION REGULATION STRATEGIES ON PROSOCIAL BEHAVIOR

The objectives of Study 3 are twofold. First, as we discussed above, we examine if different emotion regulation strategies would have differential effects on prosocial behavior. We expect that the negative effect of emotion regulation on prosocial behavior would be attenuated in those engaged in deep acting. Second, we test additional mediators between emotion regulation and sympathy for others' feeling. Following our theorizing, we examine if different emotion regulation strategies affect one's sense of control and emotional exhaustion differently. By explicitly measuring and testing this theoretically proposed mechanism, we would like to provide with clearer picture of how emotional labor shapes an emotional laborers' prosocial behavior.

# Participants

We recruited our participants through Amazon's mTurk. As in Study 2B, we asked two-item eligibility questions at the beginning of the survey, and only those who passed our predetermined qualification (paid-employees working in the service-related industry) participated. We targeted to collect 300 subjects. A total of 2021 mTurkers answered to the eligibility questions until 301 subjects completed our survey (Mage = 36.11, 46.5% Women).

# Independent Variables

Although emotional dissonance is a concept very closely related to surface acting, and thus they are often interchangeably used in the emotional labor literature, emotional dissonance is distinct from surface acting in the sense that the former arises as resulting experiences from performing the latter (Van Dijk and Brown, 2006). Given that the purpose of Study 3 is to examine the effects of different emotion regulation strategies (not emotional dissonance) on prosocial behavior, what we have to measure is the degree to which an employee uses a certain emotion regulation strategy not the degree to which an employee perceives emotional discrepancy between the felt and expressed emotions. Therefore, the independent variable in Study 3 (i.e., surface and deep acting) should be different from those in previous studies (i.e., emotional dissonance). For this reason, we moved away from the two-item emotional dissonance scales used in Study 2B to surface and deep acting scales in Study 3. We used five items from Grandey (2003) to measure surface acting (Cronbach's α = 0.91), and four items from Brotheridge and Lee (2003) and Grandey (2003) to measure deep acting (Cronbach's α = 0.88). They were all measured on 5-point scales (1 = Never, 5 = Always, see **Appendix B** for details).

# Mediators

Another purpose of Study 3 is to measure additional mediators. According to our theorizing, whether people are engaged in surface or deep acting, their sense of control will be influenced. This perceived control then influences emotional exhaustion, sympathy for others' feeling, and subsequently prosocial intention. Therefore, the potential mediators we could measure include 'sense of control,' 'emotional exhaustion,' and 'sympathy for others' feeling.'

Participants' perceived 'sense of control' was measured with Yoon and Kim's (2018) one item question on a 10-point scale (1 = none at all, little, 10 = a great deal, a lot; see **Appendix B** for details). 'Emotional exhaustion' (Cronbach's α = 0.95) was measured with nine items adopted from Maslach and Jackson's (1981) Burnout scale (1 = Never, 7 = Always). This burnout scales consist of three subscales; emotional exhaustion (nine items), depersonalization (five items), and personal accomplishment (seven items). Although we were not interested in the other two subscales (i.e., depersonalization and personal accomplishment), we just used the whole burnout scale for exploratory purpose<sup>2</sup> . 'Sympathy for others' feeling' (Cronbach's α = 0.83) was measured with the same six items used in Studies 1 and 2A (Lee, 2009) on 7-point Likert scales (1 = Strongly disagree, 7 = Strongly agree).

# Dependent Variables

We measured our dependent variables using two scales; prosocial behavior (Caprara et al., 2005) and charitable behavior (Charities Aid Foundation, 2012). The former consists of 16 items (Cronbach's α = 0.95) that encompass a variety of prosocial behaviors from sensing friends' discomfort ("I immediately sense my friends' discomfort even when it is not directly communicated to me.") to lending money ("I easily lend money or other things."). The latter consists of eight items (Cronbach's α = 0.89) that also cover various charitable behaviors an individual can do for a non-profit organization. The former was measured on 5-point scales (1 = Never/Almost never true, 5 = Almost always/Almost true) while the latter on 7-point scales (1 = Very unlikely, 7 = Very likely). Each scale was presented to participants with specific instructions (see **Appendix B**).

# Procedure

All of our independent variables, mediators, and dependent variables were measured in the reverse order to avoid demand artifact. That is, participants answered to the questions about prosocial and charitable behavior first, then about sympathy for others' feeling, emotional exhaustion, sense of control, and finally about surface and deep acting followed by demographic questions.

# Results

# Prosocial Behavior

To test our hypothesis, we first ran two simple regressions. When we regressed prosocial behavior on surface acting, surface acting

<sup>2</sup>Later we analyzed the data including the other two subscales and found that those two subscales are also significant mediators in some model specifications, however, we did not include these measures and would not discuss any further.

was found to be negatively associated with prosocial behavior [Standardized b = −0.118 t(299) = −2.057, p < 0.05]. This result is conceptually consistent with the findings of Studies 2A and 2B that emotional dissonance leads to lower willingness to help others. In contrast, when we regressed prosocial behavior on deep acting, the opposite pattern emerged. The result shows that deep acting is positively associated with prosocial behavior [Standardized b = 0.401 t(299) = 7.562, p < 0.001]. Taken together, these results show that different emotion regulation strategies (i.e., surface and deep acting) have differential effects on prosocial behavior. Thus, H4 was supported. When we analyzed each item of prosocial behavior separately, the pattern of the results did not vary. Hence, we just report the results with the composite measure hereafter.

### Charitable Behavior

We ran the same two regressions with charitable behavior as the dependent variable. When we first regressed charitable behavior on surface acting, different from our prediction and findings of Studies 2A and 2B, no significant associations were found [Standardized b = 0.046 t(299) = 0.799, p = 0.425]. We address this unexpected null effect in discussion section below. In contrast, when we regressed charitable behavior on deep acting, deep acting was found to be positively related to charitable behavior [Standardized b = 0.390 t(299) = 7.321, p < 0.001]. Taken together, the results again support the idea that different emotion regulation strategies differentially affect prosocial behavior.

### Mediational Analyses

To test if the theorized variables serially mediate the effect of emotion regulation strategies on prosocial behavior, we fit the data with model 6 in SPSS PROCESS macro with 5,000 bootstrapped samples (Hayes, 2018). When we ran a mediation analysis with surface acting as the independent variable, 'sense of control,' 'emotional exhaustion,' and 'sympathy for others' feeling' as the mediators by the order, and the composite measure of 'prosocial behavior' as the dependent variable, the indirect effect of 'surface acting' on 'prosocial behavior' via three mediators was significant (The Bootstrapped Indirect Effect = −0.0029, 95% CI = −0.0077, −0.0003).

The signs of beta coefficients were also all consistent with our theory. Specifically, 'surface acting' was negatively associated with 'sense of control' [b = −0.363, t(299) = −2.855, p < 0.005]. 'Sense of control' was negatively associated with 'emotional exhaustion' [b = −0.191, t(298) = −5.060, p < 0.001]. 'Emotional exhaustion' was negatively associated with 'sympathy' [b = −0.152, t(297) = −2.643, p < 0.01]. 'Sympathy' was positively associated with 'prosocial behavior' [b = 0.273, t(296) = 8.886, p < 0.001]. These results suggest that when people are engaged in surface acting, they would perceive lowered sense of control, and experience more emotional exhaustion. Emotionally exhausted employees become less sympathetic toward others and subsequently less willing to behave prosocially. This result again conceptually replicates our main premise that emotional dissonance decreases prosocial behavior via lowered sympathy.

In addition to this big mediational model, two other mediational paths were also significant. Specifically, the indirect effect of surface acting on prosocial behavior via 'sense of control' (The Bootstrapped Indirect Effect = −0.0367, 95% CI = −0.0741, −0.0067), and via 'emotional exhaustion' and 'sympathy for others' feeling' (The Bootstrapped Indirect Effect = −0.0387, 95% CI = −0.0732, −0.0102) were significant (see **Figure 2A** – Any solid lines mean significant paths). The former path indicates that surface acting lowered one's sense of control, and this lowered perceived control directly influences prosocial behavior. The latter path implies that surface acting (in addition to sense of control) can directly increases emotional exhaustion, and lowered sympathy, then lower willingness to help others.

When we ran a serial mediational analysis with the same dependent variable and mediators but with deep acting as the independent variable, the indirect effect of 'deep acting' on 'prosocial behavior' via three mediators was again significant (The Bootstrapped Indirect Effect = 0.0056, 95% CI = 0.0009,0.0127). The direct effect of deep acting on prosocial behavior remains significant [b = 0.324, t(296) = 7.695, p < 0.001].

The signs of beta coefficients were also all consistent with our theory. Specifically, 'deep acting' was positively associated with 'sense of control' [b = 0.381, t(299) = 2.897, p < 0.005]. 'Sense of control' was negatively associated with 'emotional exhaustion' [b = −0.271, t(298) = −6.070, p < 0.001]. 'Emotional exhaustion' was negatively associated with 'sympathy' [b = −0.214, t(297) = −4.380, p < 0.001]. 'Sympathy' was positively associated with 'prosocial behavior' [b = 0.256, t(296) = 9.155, p < 0.001]. These results suggest that when people are engaged in deep acting, they would perceive enhanced sense of control, and be less likely to experience emotional exhaustion. Without emotionally exhausted, employees can be sympathetic toward others and subsequently become more willing to behave pro-socially.

In addition to this big mediational model, one additional mediational path was significant. Specifically, the indirect effect of deep acting on prosocial behavior via 'sense of control' (The Bootstrapped Indirect Effect = −0.0279, 95% CI = 0.0035,0.0630) was significant (see **Figure 2A**). This significant indirect effect indicates that deep acting enhanced one's sense of control, and this enhanced perceived control directly influences prosocial behavior.

When we ran the same mediational analyses with charitable behavior as the dependent variable, we found very similar results (see **Figure 2B**), which suggest that surface acting and deep acting influence one's sense of control in opposite directions and subsequently prosocial and charitable behavior. In summary, the results of Study 3 provide strong support for H4 and theorized mechanism.

# Discussion

The results of Study 3 are consistent with our theorizing as well as the prediction of H4. The effect of different emotion regulation strategies on prosocial behavior is serially mediated by perceived control, emotional exhaustion, and sympathy for others' feeling. When an individual uses surface acting strategy, the pattern is the same as when one feels emotional dissonance; lowered sympathy

and lowered prosocial intention. However, when an individual uses deep acting, the pattern is different from when one feels emotional dissonance; enhanced sympathy and higher prosocial intention. This opposite pattern results from the fact that deep acting does not trigger emotional dissonance, which lends more support for our proposed hypothesis that emotional dissonance is the driver of the effects.

In Study 3, two things are noteworthy. First, when we measured the mediators not explicitly hypothesized but believed to mediate the effect of emotional regulation strategies on prosocial behavior, an interesting pattern emerged; the indirect path from surface or deep acting to prosocial behavior via 'sense of control' was significant. It is a very interesting result that could lead to a totally different interpretation and model configuration.

When people feel high (vs. low) perceived control due to deep (vs. surface) acting, they would be more (vs. less) likely to help others because they feel powerful (e.g., Inesi et al., 2011). However, another line of research also suggests that people, when feeling lack of control, could donate and help more in order to enhance self-image or to restore illusory control by helping others (e.g., Fiske et al., 1996). We therefore cannot make predictions in one or the other way. Nevertheless, it is possible that sense of control exerts direct influence on prosocial behavior independent of emotional exhaustion. Further, among three mediators tested in Study 3, sense of control seems to be closer to cognitive process while the other two mediators affective. From this reasoning, we explored a different model shown in **Figure 3**. This model assumes that the effect of emotion regulation strategies on prosocial behavior is mediated by dual paths (i.e., cognitive and affective). When we estimated the model, some of fit indices did not pass the threshold (GFI = 0.84, AGFI = 0.81, CFI = 0.94, RMSEA = 0.05). Hence, we dropped this model.

Second, when we tested the effect of surface acting on charitable behavior, we found no associations between two variables. This result is not consistent with our findings that surface acting negatively influences prosocial behavior. We have gone extra miles to come up with plausible explanations, but our best conjecture is that participants might be influenced by the instructions. That is, when we asked the questions regarding charitable behavior, we asked our participants to think about a non-profit organization, and answer if they would do various charitable behavior. Therefore, people may think about their favorite organization, and probably answer more positively. But there exists no ancillary data to test this conjecture, thus we leave this issue as it is.

# GENERAL DISCUSSION

Findings of four studies provide consistent support for the emotional dissonance effect. That is, individuals experiencing emotional dissonance tend to be less sympathetic for others' feeling, and subsequently less willing and likely to help others. This interesting phenomenon appears to hold true across various operationalizations. For instance, the pattern of results remains the same when emotional dissonance was manipulated (Studies 1 and 2A) or measured (Studies 2B and 3), and when the decision was hypothetical (Studies 2A and 3) or real (Study 2B), and when participants were considering children's education or recovery from a disaster (Studies 2A and 2B). Moreover, this effect is found to be dependent on one's emotion regulation strategies (Study 3). When people are engaged in deep (vs. surface) acting, the observed emotional dissonance effect disappeared. Taken together, the proposed emotional dissonance effect seems very robust.

Although we repeatedly replicated our key findings over multiple studies, statistical significance of support for our claim varied across studies. Especially, the results of Study 2A were significant only with a one-tailed test. We therefore conducted meta-analysis to examine (1) if the effect is on average statistically significant and (2) whether the effect sizes across studies are homogeneous. Following the recommendations by Lipsey and Wilson (2001), we tested the first by a Z statistic and the second by a Q statistic with df = number of studies compared minus one (see **Table 2**). **Table 2** shows the effect of emotional dissonance on a) sympathy for others' feeling (Studies 1, 2A, and 3) and b) prosocial behavior (Studies 2A, 2B, and 3). Z statistics for both variables indicate that our effects are very robust and statistically significant. Q statistics for both variables are non-significant, which indicates that the null hypothesis of homogeneity cannot be rejected. In summary, the results of meta-analysis suggest that our effects replicated with multiple operationalizations and measures are very robust.

Despite the results of meta-analysis, some may argue that our effects are driven not by the felt emotional dissonance but by different income levels. That is, people are less likely to donate if their income level is low while they are more likely to donate if their income level is high. If the significant results we demonstrated so far were not just spurious correlations (Simon, 1954), it should remain significant when we control for the income. Thus, we tested this possibility for Studies 2A, 2B, and 3. For Study 2A, when we regressed the willingness to help others on both sympathy for others' feeling and participants' income, only sympathy for others' feeling remains significant while income was not. For Study 2B, the unrelated survey did not collect the income data. But we were able to get some data similarly utilized – the

last month's salary. It was sensitive question and thus not forced to answer. As a result, there were many missing values in the Study 2B's dataset. But when we ran the same regression analysis with only those who answered to the question, the results are basically the same; the salary data was not statistically significant. For Study 3, when we ran the same four regression models in **Figure 2**. While controlling for income, the patterns of the results were the same and income was also significant positive (0.03 < b's < 0.08, all p's < 0.01). Taken together, we think that the alternative explanation of the income level is not viable.

# Theoretical Implications

To our best knowledge, the present research is the first attempt to examine the effect of emotional labor from the perspectives of consumers. Our research suggests that there exist many unexplored variables beyond the boundary of organizations. For instance, prosocial behavior that we introduced in this research is the variable that has never been examined along with emotional labor. We, however, proposed and demonstrated the significant linkage between emotional labor and prosocial behavior.

Additionally, we proposed multiple theoretical mediators including 'sense of control.' By showing that perceived control mediates the effect of emotional dissonance on emotional exhaustion, we unveiled one important construct between emotional dissonance and emotional exhaustion.

Despite not without limitations, we not only measured but also manipulate the degree to which emotional dissonance is perceived.

Research on emotional laborer from the perspectives of consumers has just begun. That means there exist numerous research opportunities in this topic. Especially, it would be fruitful to examine some consumer behavior that can be benefited from perceived emotional dissonance. For instance, if emotional dissonance makes people less emotional, it might be beneficial to consumers in a certain environment where consumers may easily fall prey to strong emotions (e.g., impulsive buying, addiction). The burgeoning area of financial decision making is one of the future areas to which this emotional dissonance research can contribute.

# Practical Implications

Our research suggests that the negative consequences of emotional labor may extend beyond organizational outcomes like Job retention rate or job satisfaction. Although we examined prosocial behavior as our main dependent variable, it is possible that emotional laborers act in ways more detrimental to the society. For instance, if emotional labor makes people to be less sympathetic toward others, it implies that people become more selfish and opportunistic. When more people become selfish, it is obvious that their behavior would incur societal costs. Additionally, it would be more difficult for government or policy makers to encourage people to join the force to solve many problems that should be solved by the society together (e.g., global warming, find dust problems in Asia). Therefore, policy makers may want to establish regulations guiding the display rules imposed by organizations in the service-sector.

TABLE 2 | Meta-analysis of Key Effect of Emotional Dissonance in Studies

 1–3.


Effect of dissonance

interest Sympathy for others' feeling

(Studies 1, 2A, and 3) Prosocial Behavior (Studies

2A, 2B, and 3)

 on DVs of

emotional

Weighted mean

SE of the mean

Z

95% CI Upper

95% CI Lower

Q

 df

p-value of Q

bound

bound

effect size

effect size (based on

Fisher's Zr)

−0.231 −0.144

0.043

−3.387

−0.061

−0.228

0.835

2

0.659 (n.s.)

0.053

−4.318

−0.126

−0.336

1.324

2

0.516 (n.s.)

Although we are not sure how long the effect of experienced emotional dissonance would last, we found that even briefly reminding of past experiences of emotional labor made people become less sympathetic and less willing to help others. This implies that even after leaving the job, the negative impact of working as a emotional laborer might last for a long time. Therefore, both legislators and policy makers need to think about devising the remedy for this. Guaranteed opportunities for consultation might be one possible solution.

Our findings suggest that deep acting does not lead to lowered prosocial behavior. One quick solution, thus, might be to train employees how to change the way they perceive the situation. Additionally, we found that it is perceive sense of control resulting from emotional dissonance that leads to lowered sympathy and willingness to help others. Thus, if managers could somehow induce this sense of control by various ways, we can expect to reduce the negative consequences of emotional labor.

Companies can improve the situation as well. For instance, increasing number of call centers in South Korea are nowadays allowing customer service workers to end a phone call at their discretion when experiencing verbal abuse (Song, 2017). As many companies in the service sector adopted this policy, it is reported that the well-beings of both companies and employees drastically improved (e.g., number of abusive calls dropped, increased job satisfaction, etc.).

# CONCLUSION

Our work reported here is, to our best knowledge, the first academic investigation to examine the effect of emotional labor on prosocial intention and actual behavior. We believe that this paper along with our findings shed light on the area that has been neglected; emotional laborers' behaviors as consumers. As we stated in our introduction, this has significant impact on the

# REFERENCES


society. Our finding suggests that organizations have to carefully design the display rules so that the negative impact of being involved in emotional labor can be minimized. Very often, the focus of organizations can be on just their customers, but their employees need to be considered together for better outcomes for the society.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of IRB committee at Oklahoma State University with online informed consent from all subjects. The protocol was approved by the IRB committee at Oklahoma State University.

# AUTHOR CONTRIBUTIONS

Y-nP developed the basic ideas and concepts. All authors designed and ran the experiments together. All analyses were done by Y-nP under the supervision of JJ. Y-nP drafted the manuscript and HH and JJ revised together.

# FUNDING

This work was supported by the research fund of Hanyang University (HY-2018).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg. 2019.00118/full#supplementary-material



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Park, Hyun and Jhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Is Deep Acting Prevalent in Socially Responsible Companies? The Effects of CSR Perception on Emotional Labor Strategies

Se Hyung Oh\*, Yein Hwang and Hwayoung Kim

School of Business, Hanyang University, Seoul, South Korea

This study examines the relationship between corporate social responsibility (CSR) perception and emotional labor strategies, and the effects of the interaction between CSR perception and moral identification on emotional labor strategies via affective organizational commitment. Our data from 352 frontline employees in the service industry show the main effect of CSR perception on emotional labor strategies. We find that CSR perception is positively (negatively) related to deep acting (surface acting). Affective organizational commitment mediates the relationship between CSR perception and surface acting but not between CSR perception and deep acting. Moral identification moderates the effects of CSR perception on surface acting through affective organizational commitment. This paper reveals that the employees' views on their organization's social responsiveness and morality affect their emotional labor strategies.

Keywords: CSR perception, deep acting, surface acting, moral identification, affective organizational commitment

# INTRODUCTION

Service workers often confront emotional challenges. Although employees frequently encounter demanding or even aggressive customers, they need to display the organizationally required emotions irrespective of their inner feelings. Presenting organizationally mandated emotions in this way is called emotional labor (Hochschild, 1983). To meet these emotional display rules and norms, employees usually engage in two types of emotional labor strategies: deep acting and surface acting. Deep acting is regulating the inner feelings to match the required emotional displays, whereas surface acting is superficially displaying the required feelings without modifying one's inner feelings (Hochschild, 1983). Deep acting is considered to be a better emotional display strategy than surface acting since the resulting emotional expressions are perceived as authentic by customers (Grandey, 2003).

In the early studies of emotional labor, scholars have focused on the influence of organizations. In the seminal work of Hochschild (1983), emotional labor is seen as an occupational requirement which is a mean to achieve organizational objectives. In another early influential work by Rafaeli and Sutton (1987), emotional work is described as the behaviors obligated by organizations. Including these two studies, the focus in early works in emotional labor was the effects of organizational practices (e.g., selection, training, performance evaluation, socialization, culture,

### Edited by:

Weon Sang Yoo, Korea University, South Korea

### Reviewed by:

Young Kyun Chang, Sogang University, South Korea Likuo Sung, Shanghai University of Finance and Economics, China

> \*Correspondence: Se Hyung Oh ohsh1@hanyang.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 15 October 2018 Accepted: 31 January 2019 Published: 18 February 2019

### Citation:

Oh SH, Hwang Y and Kim H (2019) Is Deep Acting Prevalent in Socially Responsible Companies? The Effects of CSR Perception on Emotional Labor Strategies. Front. Psychol. 10:308. doi: 10.3389/fpsyg.2019.00308

**87**

etc.) on service employees (e.g., Rafaeli and Sutton, 1987; Van Maanen and Kunda, 1989; Ashforth and Humphrey, 1993). However, since the early works, most emotional labor studies have focused on the effects of individual traits and dyadic relationships (see Pugh et al., 2013; Grandey and Gabriel, 2015). As a result, relatively little has been known about the effects of organizational-level factors on emotional labor (Pugh et al., 2013).

One important way to investigate the effects of organizational factors is to explore employees' perceptions about the organization. Micro organizational behavior researchers have been using this way (e.g., Moorman et al., 1998; DeConinck, 2010; Alfes et al., 2013; Rupp et al., 2018). Although perceptions are different from objective organizational reality, they are more proximal predictors of individual attitudes and behaviors (e.g., Khilji and Wang, 2006; Nishii et al., 2008). Therefore, perceptions about the organization are a key to understand how organizational factors influence employee minds and behaviors.

This study investigates the effects of employees' perceptions about their employer's involvement in corporate social responsibility (CSR) activities, called CSR perception (Rupp et al., 2013), on their emotional labor strategies. Drawing on social identity theory (Tajfel and Turner, 1986), we expect that the positive perceptions and emotions about the employer generated from CSR perception induce strong motivation to engage in genuine rather than superficial emotional labor strategies. Through examining this hypothesis, this study intends to reveal that employees' evaluation of the employer is an important determinant of emotional labor strategies.

Another objective of this study is to show the individual concerns about their employers' morality can be a boundary condition of the effects of CSR perception. Employees' CSR perception reflects the employer's involvement in ethical and philanthropic activities (Maignan and Ferrell, 2000). Therefore, CSR perception would affect employees with high desire to belong to an ethical organization (May et al., 2015) more strongly than those with low desire. Through testing this proposition, we aim to show how employees' moral desires interact with the employers' perceived involvement in morally right activities (CSR) to affect the employees' emotional labor strategies. The results of this study thus will demonstrate the importance of employer's morality and employees' moral desires on frontline service workers' effectiveness and morale.

The current study fills the theoretical gap in both emotional labor and CSR literature. As mentioned, there is a lack of studies about the influence of organizational factors in the emotional labor literature (Pugh et al., 2013). Through investigating frontline service employees' emotional and behavioral reactions to their perceptions about the employer, this study will provide insight toward the better understanding of the organizationlevel influences on employee emotional labor. This study will also contribute to CSR literature. Different from emotional labor research, CSR studies have mainly focused on the macrolevel of analysis (Glavas, 2016) although CSR is a construct bridging macro and micro levels (Lindgreen and Swaen, 2010). Consequently, little has been known about how CSR directly influences employee behaviors (Glavas and Kelley, 2014). Through examining the effects of CSR perceptions on frontline employees, this study expands the current knowledge of mediation mechanism between corporate CSR activities and employee outcomes.

# THEORETICAL BACKGROUND AND HYPOTHESES

Corporate social responsibility refers to all company activities demonstrating the inclusion of social and environmental concerns in business operations, and in interactions with stakeholders, also according to the ambition levels of corporate sustainability (van Marrewijk, 2003). CSR has gained significant attention from researchers and practitioners (Aguinis and Glavas, 2012; Gond et al., 2017) because it generates various types of positive organizational outcomes such as corporate credibility in product (Lin et al., 2012), enhanced financial performances (Orlitzky et al., 2003), and positive relationships with customers (Matute-Vallejo et al., 2011). These positive effects are mainly due to enhanced organization's image and attractiveness for external stakeholders (e.g., customers, shareholders, and society) (Minor and Morgan, 2011).

Researchers have also found that CSR activities induce positive outcomes from internal stakeholders. For example, Jones (2010) actually found that volunteering program through which employees perform community service enhances employee organizational citizenship behaviors (OCBs), and in-role performance. Turker (2009) also found that CSR were the significant predictors of employees' organizational commitment. CSR activities for employees such as providing career growth opportunities or family-friendly policies affected the level of employees' organizational commitment (Turker, 2009). Glavas and Piderit (2009) revealed that CSR induced positive employee outcomes such as engagement and creativity. These results confirm Rupp et al. (2006) proposition that CSR activities can trickle down to affect employees' subsequent attitudes and behaviors in a positive way.

Despite the growing number of studies and interests on the effects of CSR activities on incumbent employees, due to the lack of the investigations about the mediational processes (Ng et al., 2018), how the macro-level CSR activities affect employees' attitudes and performances are still vastly unknown (Glavas, 2016; Aguinis and Glavas, 2017). As a way of overcoming this theoretical problem, the concept of CSR perception is getting increasing attention from researchers. CSR perception is "employees' perceptions of the extent to which their employer engages in various CSR activities" (Rupp et al., 2018, p. 560). It reflects how employees make sense of their organizations' CSR activities and help explain the process of the individuallevel consequences of CSR activities (Aguilera et al., 2007). Similar with objective CSR, CSR perception produces various types of positive employee outcomes such as identification (Kim et al., 2010), engagement (Caligiuri et al., 2013), performance (Jones, 2010), and creativity (Hur et al., 2016). Compared with objective CSR activities, CSR perception actually better predicted employee behaviors such as job satisfaction and

turnover intentions (Valentine and Fleischman, 2008; Valentine and Godkin, 2016). It is because perception is a more immediate predictor of the employee attitudes and behaviors than the objective organizational reality (Rupp et al., 2013).

Our first hypothesis deals with the relationship between service employees' CSR perceptions and their emotional labor strategies. Social identity theory maintains that an individual's self-description is significantly influenced by the membership of social groups (Tajfel and Turner, 1986). According to this theory, because of self-enhancement needs, if a group is perceived as having characteristics positively distinguishing it from other groups, members increasingly identify with the group and accept its rules and norms (Tajfel and Turner, 1986; Ashforth and Mael, 1989). More importantly, when individuals strongly identify with a group, they are motivated to engage in behaviors that are beneficial for the group because they tend to identify the group's wellbeing with their own (Tajfel and Turner, 1986).

Deep acting involves the motivation to alter the inner feelings to conform to the organization's rules and norms (Gosserand and Diefendorff, 2005). This motivation is enhanced when employees like their organization (Mishra et al., 2012), accept its emotional display rules and norms (Gosserand and Diefendorff, 2005), and identify its success with their own (Ashforth and Humphrey, 1993). CSR activities enhance an organization's image and attractiveness not only for external stakeholders (e.g., customers, shareholders, society) but also for incumbent employees (Rupp et al., 2006; Brammer et al., 2007; De Roeck et al., 2014). This means that if employees perceive that their employer actively engages in CSR activities, they increasingly identify with it and accept its rules and norms because of the employer's enhanced images and attractiveness. Consequently, employees are motivated to match their inner feelings with the organization's emotional display rules and norms to provide good customer service, enhancing the organization's success. Based on this reasoning, we presume that employees' CSR perceptions about their organizations are positively related to their deep acting.

Different from deep acting, surface acting reflects employees' disengagement from the organizational norms and rules (Ozcelik, 2013). When employees are dissatisfied with their organization, they tend to dissociate themselves from it (Farrell, 1983). This type of employee is typically demotivated and follows organizational rules and norms superficially (Farrell, 1983). However, as mentioned, CSR perception increases employees' satisfaction and identification with their organization. CSR perception also motivates employees to regulate their emotions to follow the organization's emotional display rules and norms. Hence, we predict that employees' CSR perceptions about organizations are negatively related to their surface acting.

H1: Employees' CSR perceptions are positively related to their deep acting.

H2: Employees' CSR perceptions are negatively related to their surface acting.

The next set of hypotheses deals with the mechanism of the influence of CSR perception on emotional labor strategies. In this study, we argue that affective organizational commitment is an important mediator between CSR perception and emotional labor strategies. Affective organizational commitment refers to the employee's emotional attachment to, identification with, and involvement in the organization (Mowday et al., 1979). Affective commitment positively affects important employee outcomes such as job satisfaction (Boles et al., 2007), OCB (Ng and Feldman, 2011), and in-role performance (Meyer et al., 2002).

As mentioned, when employees are attracted to and proud of their organization, they are more likely to identify with and affectively committed to it (Ashforth and Mael, 1989; Brammer et al., 2007). This increased affective commitment, in turn, enhances employees' motivation to follow the organization's rules and norms (Mowday et al., 1979). When employees perceive that their organization is socially responsible, they increasingly identify with and affectively commit to it because of the organization's distinctive positive characteristics (Aguilera et al., 2007; Brammer et al., 2007; Glavas and Kelley, 2014). This increased affective commitment would, in turn, intensify employees' motivation to modify their inner feelings to provide good and professional customer service to support the achievements of the organization. In addition, the increased affective commitment due to CSR perception would reduce the inclination to provide superficial and shallow customer service. Hence, our hypotheses are as follows.

H3: The positive relationship between employees' perceptions of their company's CSR activities and their deep acting is mediated by their affective organizational commitment. H4: The negative relationship between employees' perceptions of their company's CSR activities and their surface acting is mediated by their affective organizational commitment.

Recently, May et al. (2015, p. 682) developed a new construct called moral identification, which is defined as "the perception of oneness or belongingness associated with an organization that exhibits ethical traits (e.g., care, kindness, compassion), which also involves a deliberate concern of the membership with an ethical organization." This construct reflects employees' judgment about their organization's morality as well as their desire to be part of an ethical organization. Employees with high moral identification tend to avoid unethical behaviors and have low turnover intentions when they believe that their organization is ethical (May et al., 2015).

In this study, we argue that moral identification moderates the relationship between CSR perception and affective organizational commitment. People with high moral identification care about the morality of their organizations and have a strong desire to be part of ethical organizations (May et al., 2015). Given that CSR activities are thought to be the ethical behaviors of companies (Carroll, 1991; Joyner and Payne, 2002), employees with high moral identification would show more positive attitudinal reactions to the CSR activities of their organizations. Empirical evidence shows that people with high moral concerns are more likely to react positively to organization activities focusing on helping external and internal stakeholders. For example, Rupp et al. (2013) found that a positive relationship between CSR perceptions and OCB

was more pronounced among employees with high (vs. low) moral concerns (Rupp et al., 2013). Kolodinsky et al. (2010) also found that business students who have strong ethics of caring others showed more favorable attitude about CSR. Additionally, Wang et al. (2017) found that employees' moral concerns amplified the positive effects of CSR perception on organizational identification. Consequently, we hypothesize that moral identification moderates the positive relationship between CSR perceptions and affective organizational commitment. More specifically, the influence of employees' CSR perceptions on their affective organizational commitment would be stronger (weaker) when employees have high (low) moral identification.

H5: The positive effects CSR perception on affective organizational commitment is moderated by moral identification such that the effect is stronger (weaker) when moral identification is strong (weak).

Thus far, we have hypothesized that (a) the relationship between CSR perception and emotional labor strategies is mediated by affective organizational commitment and that (b) the relationship between CSR perception and affective organizational commitment is moderated by moral identification. These hypotheses collectively suggest a moderated mediation model. As suggested, when employees perceive their organizations are socially responsible, they have enhanced affective organizational commitment, leading to increased deep acting and decreased surface acting. This indirect relationship will be stronger (vs. weaker) when the employees have strong (vs. weak) moral identification, given that CSR activities are thought to be the ethical behaviors of companies (Carroll, 1991; Joyner and Payne, 2002). **Figure 1** depicts our research model. Our model predicts the first stage moderation mediation model where moral identification moderates the indirect effect of CSR perception on emotional labor strategies via affective organizational commitment.

H6: The positive indirect effect of CSR perception on deep acting via affective organizational commitment is moderated by moral identification such that the indirect effect is stronger (weaker) when moral identification is high (low).

H7: The negative indirect effect of CSR perception on surface acting via affective organizational commitment is moderated by moral identification such that the indirect effect is stronger (weaker) when moral identification is high (low).

# METHODS

To test our hypotheses, we collected data from 352 service workers through an online research company that has more than one million panels. All respondents were frontline workers working as full-time regular employees in the organizations mostly in service (hotel, tourism, airline, etc.) (51%) and retail (department store, mall, etc.) (40%) industries. Rest of the respondents worked in hospitals, insurance companies, and transportation companies. Respondents' average age was 37.8 years (SD = 9.8; min = 20; max = 72) and average tenure was 73.7 months (SD = 67.7; min = 1; max = 362). The average education level was between 2-year college and 4-year college. Over half (52%) of the respondents were women.

# Measures

We used a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree) throughout the questionnaire. The questionnaires were originally constructed in English but were translated into Korean. We used a translation and backtranslation procedure (Brislin, 1980) to ensure the accuracy of the translation.

# Corporate Social Responsibility Perception

Corporate social responsibility perception was measured with the five-item scale developed by Maignan and Ferrell (2000). A sample item is "Flexible company policies enable employees to better coordinate work and personal life." The Cronbach's alpha is 0.86.

# Affective Organizational Commitment

Affective organizational commitment was measured with Meyer et al. (1993) six-item scale. However, exploratory factor analysis revealed that three items of the scale were double-loaded with CSR perception and these were excluded. The three remaining items used to measure affective organizational commitment are "I do not feel a strong sense of 'belonging' to my organization" (reverse coded), "I do not feel 'emotionally attached' to this organization" (reverse coded), and "I do not feel like 'part of the family' at my organization" (reverse coded). The Cronbach's alpha is 0.94.

# Deep Acting and Surface Acting

Deep acting and surface acting were measured by using the sixitem scale developed by Brotheridge and Grandey (2002). Three items measured deep acting and the other three items measured surface acting. A sample item of deep acting is "when doing your job, how often do you try to actually experience the emotions you must show to customers?" The Cronbach's alpha of deep acting is 0.86. A sample item of surface acting is "when doing your job, how often do you fake a good mood when interacting with customers?" The Cronbach's alpha of surface acting is 0.81.

# Moral Identification

Moral identification was measured by the five-item scale developed by May et al. (2015). This scale first requests respondents to imagine a person who is caring, compassionate, fair, friendly, generous, helpful, hardworking, honest, and kind. Then, it asks five questions based on the image of this person, such as "I strongly desire to be a member of an organization whose members have these characteristics." The Cronbach's alpha is 0.86.

# Data Analysis

In order to test Hypothesis 1 through 5, we conducted a series of regression analysis. The mediation hypotheses (Hypotheses 3 and 4) were examined by the methods suggested by Sobel (1982) and Baron and Kenny (1986). Finally, the moderated

mediation hypotheses (Hypotheses 6 and 7) were tested through the bias-corrected bootstrapping techniques with 5,000 bootstrap samples using the PROCESS macro for SPSS (Hayes, 2013). This bootstrap technique involves the computation of the 95% bias-corrected bootstrap confidence intervals and it tests for a non-zero weight of the moderator in the indirect effect process. If the 95% bias-corrected confidence interval does not include zero, the moderation mediation is present.

# RESULTS

# Preliminary Analyses

**Table 1** presents the descriptive statistics, reliabilities, and correlations. CSR perception has a positive correlation with affective organizational commitment and moral identification. Affective organizational commitment and surface acting are negatively correlated.

We conducted a series of confirmatory factor analyses to verify the plausibility of our hypothesized factor structure. The proposed five-factor model demonstrated an acceptable fit to the data (χ <sup>2</sup> = 458.89, df = 142, p < 0.001, CFI = 0.91, RMSEA = 0.08, TLI = 0.90, SRMR = 0.08). Furthermore, the five-factor model provided a significantly better fit than the other models, namely the four-factor model that combined CSR perception and moral identification into a single factor (χ <sup>2</sup> = 1264.57, df = 146, p < 0.001, CFI = 0.69, RMSEA = 0.15, TLI = 0.73, SRMR = 0.09), the three-factor model that combined CSR perception, moral identification, and affective organizational commitment into a single factor (χ <sup>2</sup> = 1725.66, df = 149, p < 0.001, CFI = 0.556, RMSEA = 0.18, TLI = 0.49, SRMR = 0.14), the two-factor model that merged CSR perception, moral identification, affective organizational commitment, and surface acting into a single factor (χ <sup>2</sup> = 2260.89, df = 151, p < 0.001, CFI = 0.41, RMSEA = 0.208, TLI = 0.37, SRMR = 0.17), and the onefactor model that merged all five variables into a single factor (χ <sup>2</sup> = 2581.2, df = 152, p < 0.001, CFI = 0.32, RMSEA = 0.22, TLI = 0.23, SRMR = 0.19). These results confirm the validity of our suggested factor structure.

When all data are self-reported and collected through the same questionnaire during the same period of time, common method variance is of specific concern. To assess the potential impact of common method variance, we conducted Harman's single-factor test using the un-rotated principal component analysis and principal component analysis with varimax rotation was conducted. One-factor model of the un-rotated principal component analysis explained only 27.1% of total variance. The result of principal component analysis with varimax rotation showed that the first five eigenvalues were greater than 1.15. The five factors accounted for 73.5% of the total variance; however, no one single factor accounted for more than 25% of the variance, and the highest variance explained by any single factor was 17.6%, suggesting that the common method bias was mitigated.

# Hypothesis Testing

**Table 2** reports the results of a series of regression analyses to test Hypotheses 1 to 5. In our analyses, we included age, sex, education, tenure, and company size. Sex affects individual differences in detecting emotional cues and managing own emotions (Domagalski, 1999). Age is influential on emotion control abilities (Kruml and Geddes, 2000) and tenure affects the experience of emotional dissonance (Hochschild, 1983; Kruml and Geddes, 2000). Education-level was found to affect emotional experience (Hamid and Cheng, 1996). Company size is influential on customers attitudes toward the company and its employees (Jarvenpaa et al., 1999), potentially affecting the frontline workers' customer experience. All independent and control variables were mean-centered prior to entering the regression analyses.

Hypothesis 1 proposed that CSR perception is positively related to deep acting. As shown in model 1, after accounting for the effects of the control variables, we found that CSR perception significantly and positively affects deep acting (β = 0.136, p < 0.01), supporting Hypothesis 1. Hypothesis 2 predicted the

negative influence of CSR perception on surface acting. Model 4 shows that CSR perception significantly and negatively affects surface acting (β = −0.106, p < 0.05), supporting Hypothesis 2. Hypotheses 3 and 4 predicted the mediating effects of affective organizational commitment between CSR perception and the two emotional labor strategies, deep acting (H3) and surface acting (H4). To test these hypotheses, we conducted a series of regression analyses suggested by Baron and Kenny (1986). As shown in model 1, the direct effect of CSR perception on deep acting is significant. Model 7 shows that CSR perception has a significant positive effect on the mediator variable, affective organizational commitment (β = 0.301, p < 0.01). However, as shown in model 2, the effect of the mediator variable, affective organizational commitment, on deep acting is not significant (β = 0.004, p = n.s.). In fact, compared with model 1, the effect of CSR perception on deep acting does not reduce even though affective organizational commitment is included in model 3. Therefore, Hypothesis 3 is not supported. Hypothesis 4 suggests the mediating effect of affective organizational commitment between CSR perception and surface acting. As seen in model 4, CSR perception has a significant negative effect on surface acting (β = −0.106, p < 0.05). Model 7 shows the CSR perception has a significant effect on affective organizational commitment (β = 0.301, p < 0.01). Model 5 shows that the mediator variable, affective organizational commitment, has a significant negative effect on surface acting (β = −0.161, p < 0.05). Finally, compared with model 4, the significant negative effect of CSR perception becomes insignificant when affective organizational commitment is added into model 6 (β = −0.061, p = n.s.), demonstrating the full mediating effect of affective organizational commitment. The result of the Sobel test (Z = −2.95, p < 0.01) confirms that CSR perception has an indirect effect on surface acting via affective organizational commitment. Therefore, Hypothesis 4 is supported. Hypothesis


N = 352. Numbers in parentheses on the diagonal are Cronbach's alphas of the measures used in the study. <sup>∗</sup>p ≤ 0.05, ∗∗p ≤ 0.01.

### TABLE 2 | Regression test results.


N = 352. <sup>∗</sup>p < 0.05 ∗∗p < 0.01.

5 is about the moderating effect of moral identification on the relationship between CSR perception and affective organizational commitment. Models 7, 8, 9 show the sequence of hierarchical regression analysis. In model 9, the interaction effect between CSR perception and moral identification on affective organizational commitment is significant (β = 0.135, p < 0.05). **Figure 2** shows the patterns of interaction effects. Compared with employees with low moral identification, those with high moral identification showed a more positive relationship between CSR perception and affective organizational commitment. Therefore, Hypothesis 5 is supported.

Hypotheses 6 and 7 suggest that the indirect effects of CSR perception on the two emotional labor strategies (deep acting and surface acting) through affective organizational commitment are moderated by moral identification. We tested the moderated mediation model suggested in Hypotheses 6 and 7 by using the PROCESS macro (Hayes, 2013). **Tables 3A,B** summarize the results of the two moderated mediation tests. The 95% bias-corrected confidence intervals for the bootstrapped indirect effects at the two levels of the moderator, moral identification, were generated to test the significance of the conditional indirect effects. **Table 3A** shows the results of the test for Hypothesis 6. When moral identification is both high (+1 SD) and low (−1 SD), The 95% bias-corrected confidence intervals for the bootstrapped indirect effects from CSR perception to deep acting contain zero, indicating that moral identification does not moderate the indirect effect (see **Table 3A**). Therefore, Hypothesis 6 is rejected. **Table 3B** shows the results of the test for Hypothesis 7. When moral identification is low (−1 SD), the negative indirect effect of CSR perception is insignificant (b = −0.02, boot SE = 0.02, CI95% = [−0.05, 0.02]). However, when moral identification is high (+1 SD), the negative indirect effect of CSR perception is significant (b = −0.05, boot SE = 0.03, CI95% = [−0.11, −0.02]), showing that the significance of the indirect effects of CSR perception on surface acting via affective organizational commitment depends on the values of moral identification.

In order to probe the tested effect further, we used the Johnson-Neyman technique (Preacher et al., 2007; Hayes and Matthes, 2009). As seen in **Figure 3**, when moral identification is greater than 4.2, the negative indirect effect different from zero. This means that the negative effects of CSR perception on surface acting via affective organizational commitment are significant and amplified by moral identification when moral identification is greater than 4.2. Since the indirect effects of CSR perception on surface acting via affective organizational commitment are moderated by moral identification and the moderating effect occurs in the predicted direction, Hypothesis 7 is supported.

# DISCUSSION

The purpose of this study was to examine whether employees' perceptions about the employer's involvement in CSR activities affect their emotional labor strategies. Another purpose was to explore the mechanism and a potential moderator of the suggested relationship. Our results show that CSR perception significantly affects employees' emotional labor strategies. We found that CSR perception is positively associated with deep acting and negatively associated with surface acting. It seems that employees' evaluation of their employer's responsiveness to social matters and needs does affect their choice of emotional labor strategies. The perceived positive distinctiveness of the employer (CSR activities) seems to motivate employees to provide professional customer services with genuine emotional displays while reducing their inclination to serve customers superficially.

We also found that CSR perception significantly influences affective organizational commitment. Our data revealed that CSR perception has a significant positive impact on affective organizational commitment. It appears that employees have an increased sense of emotional attachment to the employer when they perceive that it is socially reputable. However, this increased affective organizational commitment accounts for the TABLE 3A | Conditional indirect effects of CSR perception on deep acting at values of moral identification.



95% bias-corrected confidence interval.

fpsyg-10-00308 February 14, 2019 Time: 19:8 # 8

mechanism between CSR perception and surface acting only. Our results showed that affective organizational commitment fully mediates the relationship between CSR perception and surface acting. However, affective organizational commitment does not mediate the relationship between CSR perception and deep acting. Employees working for companies actively engaging in CSR activities enjoy the socio-emotional benefits from customers and society due to the company's enhanced images and reputations (Minor and Morgan, 2011; Lii and Lee, 2012; Hur et al., 2014). This extrinsic rewards might account for the positive relationship between CSR perception and deep acting. Further investigation of the mediating mechanism of the CSR perception-deep acting relationship is worthwhile.

We found that moral identification moderates the positive effects of CSR perception on service employees' organizational attitudes and emotional labor strategies. The positive effects of CSR perception on affective organizational commitment were stronger (weaker) when employees had a strong (weak) desire to be part of an ethical organization. Furthermore, moral identification moderated the indirect negative effect of

CSR perception on surface acting via affective organizational commitment. Our data showed that the negative indirect effect of CSR perception on surface acting through affective organizational commitment is stronger (weaker) when employees have strong (weak) moral identification. It appears that employees who have a strong desire to be a member of an ethical organization tend to react to employer's CSR activities more positively, resulting in a weaker tendency to provide superficial customer service.

# Theoretical Implications

fpsyg-10-00308 February 14, 2019 Time: 19:8 # 9

This study offers several contributions to current emotional labor literature. First, it reveals a novel determinant of emotional labor strategies, CSR perception. Given that emotional labor is identity-challenging experiences (Ashforth and Humphrey, 1993), perceptions about an employer, which affects employees' identity significantly (Hogg and Terry, 2000), should generate various types of emotional and behavioral consequences from service employees. This study provides one example. Our data show that how employees see the employers' responsiveness to social needs (CSR perception) indeed had significant positive effects on deep acting and negative effect on surface acting. Similar with other studies about the effects of the perceptions about the employer (Duke et al., 2009; Janssen et al., 2010; Kraemer and Gouthier, 2014; Hur et al., 2016), enhanced organizational attitudes seem to account for the mediating mechanism. Future studies need to investigate more specific mediating mechanism between CSR perception and emotional labor strategies. For example, given that emotional labor has three components [emotion requirements (e.g., emotion display rules), emotion regulation (e.g., modifying feelings), and emotion performance (e.g., observable emotional expressions congruent with requirements)] (Grandey and Gabriel, 2015), how CSR perception affect each component of emotional labor is worthwhile to explore. Also how employee identity interacts with the three component of emotional labor is another important and interesting research topic.

This study provides implication on organizational moralityemotional labor relationship. CSR is an ethical organizational behavior in nature (Carroll, 1991). Therefore, the results of the current study imply that morality of the employer influences employee's mindset and behaviors when they serve customers. Previous emotional labor researchers have found that the degree of organizational efforts on service ethics are positively associated with service employee emotional performances (e.g., appropriate emotional expression) (Ashforth and Humphrey, 1993). For example, Dietz et al. (2004) found that service workers provided enhanced customers services when excellent service is an important theme in an organization. Lam et al. (2010) also found that the service climate has a positive impact on positive emotional display and buffers negative influences from supervisors. Additionally, Pugh et al. (2013) pointed out that the organization-level utilization of human resource management (e.g., recruitment, training, performance monitoring, etc.) stressing customer service quality should have positive influences on employees' emotional labor practices. This line of research is about the positive relationship between an

organization's efforts on service ethics and employees' emotional performances. However, the current study suggests that not only the organization's work- or service-related ethics but its ethics in general (e.g., responsibility about external and internal stakeholders in general) can generate important emotional labor consequences. Although there are emotional labor studies about the effects of ethics-related organizational factors such as organizational justice (Moliner et al., 2005) and perceived organizational support (Duke et al., 2009; Hur et al., 2013), we could not find any study explicitly focusing on the organizational morality and employee emotional labor. Given that individual and organizational morality is one of the major factors of employee identity, decision making and behaviors (Trevino et al., 2006), organizational morality is more likely to affect employee emotional performance. Our data provide initial evidence. Future researchers need to investigate more about the organizational morality-emotional labor relationship.

Finally, the current study provides implications on the conservation of resources (COR; Hobfoll, 1989) theory-based emotional labor studies. COR theory, originally introduced as a motivation theory, claims that "individuals strive to obtain, retain, protect, and foster those things that they value" (Hobfoll, 2011, p. 341). COR theory is one of the main theoretical frameworks in explaining the processes and consequences of emotional labor (Grandey and Gabriel, 2015). It suggests that involuntary emotional regulation to match the organization's emotional requirements consumes psychological resources, resulting in emotional exhaustion (Brotheridge and Lee, 2002). It also suggests that gains in motivational resources such as task significance and social support can help service workers to prevent resource losses and enable them to cope with stresses from service works (Grandey and Diamond, 2010). There are many emotional labor studies based on COR theory (e.g., Chau et al., 2009; Grandey and Diamond, 2010; Goodwin et al., 2011; Bhave and Glomb, 2016) and most evidence fit well with the theory. The current study provides novel insight on this resource-based perspective. The results of this study imply that CSR perception can be a source of motivational resources that service workers consume when they interact with customers. Although we could not find any studies framing perceptions about employer as a source of resources, we noticed that there are several studies resonating with this idea. For example, Duke et al. (2009) demonstrated that perceived organizational support can provide a psychological buffer from the frustration and dissatisfaction experienced from service works in retail stores. Gouthier and Rhein (2011) showed that frontline employees' pride about organization affects their motivation to commit to customer service positively. Kraemer and Gouthier (2014) also revealed that service workers' organizational pride alleviate the emotional exhaustion and turnover intention. This line of research suggests that positive perceptions about the employer such as CSR perception provide psychological and motivational resources to cope with negative experiences from service works. Whether CSR perception actually works as other resources (e.g., helping employees to tolerate stresses and emotional exhaustion) is a worthwhile topic to investigate for future researchers.

# Managerial Implications

fpsyg-10-00308 February 14, 2019 Time: 19:8 # 10

This study provides clear practical implications. First, it is important for service organizations to become socially responsible. Our results show that when employees see that their organizations have high responsibility on social matters and needs, their organizational attitudes become positive and their emotional performance toward customers is enhanced. CSR activities have long been considered to be only as a tool for public relations (Middlemiss, 2003; Kim and Park, 2011). However, practitioners should also consider them to be an instrument to enhance employee morale and organizational performance.

Second, management needs to understand the individual differences in the reaction to CSR activities. Indeed, CSR activities can generate both positive and negative impacts on the employees' morale (McShane and Cunningham, 2012). Based on our results, we can speculate that individual differences such as moral identification could be one of the reasons. Therefore, for example, recruiting employees for CSR activities based on their individual differences in values and personalities should increase the positive consequences of CSR activities.

Finally, managers need to be concerned about employees' emotional attachment to the organization. Our results suggest that employees' affective organizational commitment, which rises with their CSR perceptions, plays an important role in service workers' attitudes and emotional performance. Given that affective organizational commitment is the direct precursor of emotional labor strategies, a manager who wants to enhance employees' emotional performance should care about their emotional attachment to the organization.

# Limitations and Future Research

The current study is not without limitations. First, our results are based on cross-sectional data. Although cross-sectional data are still widely used in emotional labor studies (e.g., Molino et al., 2016; Indregard et al., 2018a,b), using time-lagged data would increase the persuasiveness of the suggested causal relationships in this study. Future research would thus benefit from using time-lagged data to analyze the dynamics of employees' CSR perceptions on their organizational attitudes and behaviors. Second, our data depend on self-reports. We used self-report data because all our variables deal with employees' feelings, which are difficult for others to measure correctly. Although Harman's single factor results showed that our data do not have a common method variance problem (Podsakoff et al.,

# REFERENCES


2012), future studies can confirm our results by conducting a study collecting data from multi sources (e.g., customers, coworkers, etc.). Third, our data were collected only from Korean subjects. Since cultural orientation affects employees' attitudes and behaviors in organizational settings (Hofstede and Hofstede, 2005), our collectivistic Koreans data may show different patterns from individualistic Westerner data. Further research using data from subjects with different cultural orientation is thus required to confirm the external validity of our results.

# CONCLUSION

This study examines the influence of employees' CSR perceptions on their emotional labor strategies. Our results demonstrate that CSR perception encourages deep acting and discourages surface acting. Affective organizational commitment explains why CSR perception discourages the inclination to provide superficial customer services. Finally, employees' desire to be a member of an ethical organization amplifies or diminishes the effect of CSR perception on surface acting via affective organizational commitment. This study suggests that management should be concerned about employees' perceptions of an emotional attachment to the organization in order to enhance their emotional performance to customers.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Hanyang University Institutional Review Board with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Hanyang University Institutional Review Board.

# AUTHOR CONTRIBUTIONS

SO conceived the original idea, developed the theoretical models, and took the lead in writing the manuscript. YH helped developing theories and wrote theories for Hypotheses 1, 2, 3. HK carried out data analyses, interpreted the results, and wrote the "Materials and Methods" and "Results" section. SO, YH, and HK, all three of them, significantly contributed to manuscript.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Oh, Hwang and Kim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 1

# Relationship Bonds and Service Provider's Emotional Labor: Moderating Effects of Collectivism

Myoung-Soung Lee, Sang-Lin Han\*, Suji Hong and Hyowon Hyun

School of Business, Hanyang University, Seoul, South Korea

Since service providers directly conduct emotional labor to customers, it is important to identify the factors influencing emotional labor of service providers. Even though the studies identifying the predisposing factors influencing emotional labor are taking place, there is no empirical evidence confirming how relationship bonds, which have been established between corporations and service providers, are related to emotional labor. This study examined the influences of relationship bonds on emotional labor through person-organization fit (P-O fit) and the moderating effects of collectivism between P-O fit and emotional labor. Analysis was conducted by performing questionnaire surveys targeting 350 employees in the financial industry. As a result of the analysis, it has been found that financial bonds, social bonds, and structural bonds enhanced P-O fit and P-O fit improved deep acting. In addition, this study identified that collectivism of service providers strengthened the influence of P-O fit toward deep acting. This study not only suggested the empirical evidence identifying the process of relationship bonds influencing emotional labor but also expanded the scope of study by examining moderating roles of collectivism in cultural psychology aspect.

### Edited by:

Ki-Joon Back, University of Houston, United States

### Reviewed by:

Juan M. Madera, University of Houston, United States Gilbert Ernest Franco, Beacon College, United States

> \*Correspondence: Sang-Lin Han slhan@hanyang.ac.kr

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 14 September 2018 Accepted: 06 February 2019 Published: 26 February 2019

### Citation:

Lee M-S, Han S-L, Hong S and Hyun H (2019) Relationship Bonds and Service Provider's Emotional Labor: Moderating Effects of Collectivism. Front. Psychol. 10:370. doi: 10.3389/fpsyg.2019.00370 Keywords: relationship bonds, emotional labor, person-organization fit, collectivism, service provider

# INTRODUCTION

Emotional labor is a service provider's effort to present emotions in a way that is desired by the organization (Meier et al., 2006) and a form of emotion control which creates a publicly visible facial and bodily display in the workplace (Hochschild, 1983). Therefore, service providers engaged in emotional labor try to induce customers' positive responses with the display of appropriate emotions such as smile by controlling personal emotion (Wang, 2014). Such emotional labor is closely related with service providers as it happens mainly during the interaction between customers and service providers on behalf of an organization (Ashforth and Humphrey, 1993).

Many of the preceding studies on emotional labor focused on customer response to the emotional labor (Hennig-Thurau et al., 2006; Groth et al., 2009), or the influence of emotional labor on the employee's job outcomes (Ashforth and Humphrey, 1993; Van Dijk and Brown, 2006). Not a few researchers confirmed that service provider's display of positive emotion enhances customer evaluation about service quality (Pugh, 2001), increases reuse intentions (Wang, 2009), and leads to positive word-of-mouth (Wang, 2014). From the perspective of corporations, emotional labor of the service provider leads to positive results. From the service provider's viewpoint, however, it can cause negative outcomes such as an increase of job stress and sick leave and high turn-over rate (Hatzinikolakis and Crossman, 2010).

Affective delivery means the delivery of emotion to the customer by service provider (Christoforou and Ashforth, 2015). Affective delivery is associated with emotional exhaustion fpsyg-10-00370 February 23, 2019 Time: 17:41 # 2

(Trougakos et al., 2008). Emotional labor requires more emotional resources than affective delivery and causes a higher level of emotional exhaustion (Smith et al., 2009). If corporations neglect emotional labor and job stress of service providers, they are exposed to a bigger risk of emotional exhaustion which can lead to higher turnover and lower commitment to the organization (Huang and Dai, 2010). In that sense, it is important for firms to understand the factors influencing emotional labor of service providers but there have been only few studies regarding the antecedent factors of emotional labor. Furthermore, previous studies have been conducted with regard to the factors influencing emotional labor based on the factors related to job and individual characteristics (Hur et al., 2014). Job autonomy (Diefendorff and Gosserand, 2003), emotional requirement (Grandey, 2003), time pressure (Ashforth and Humphrey, 1993) have been identified as the leading factors related to job influencing emotional labor. Some important issues with regard to individual characteristics, the effects of emotional intelligence (Delpechitre and Beeler, 2018), chronological age and work experience (Hur et al., 2014), demographic factor (Schaubroeck and Jones, 2000) of service providers have been identified. Even though scope of study on emotional labor is being expanded, researches on predisposing factors of emotional labor are very limited.

One of the factors influencing service provider's emotional labor is related to the organization which they belong to. As their perception about the organization influences the emotional labor (Yoo and Arnold, 2016), businesses need to make an effort to have a positive relationship with the service providers working for them. Studies on relationship marketing suggest relationship bonds as strategy for a corporation to build a relationship with customers which proved crucial for better relationship performance (Dash et al., 2009). Most of the existing research on relationship bond focused on strategies to build ties between corporations and customers, or between corporations, and studies from the perspective of employees has not been adequate. Accordingly, this study aims to identify the influences of the relationship bond on the emotional labor of service providers.

Relationship between a corporation and service providers positively influences the service provider's affective commitment and normative commitment to the corporations (Wang, 2014), which in turn enhances person-organization fit. Conversely, this person-organization fit can influence the emotional labor of the service providers. When individuals perceive organizational characteristics or values consistent with their own, they tend to show positive emotions and attitude toward the organization but when they perceive the opposite, they are more likely to be negative in their emotions and attitude due to conflicting values (Cable and Judge, 1996; Lauver and Kristof-Brown, 2001). As value conflicts between individuals and organization tend to require more emotional resources, poor person-organization fit affects negatively emotional labor. On the other hand, cultural propensity of an individual is another factor influencing emotional labor. Employees with collectivistic tendencies are more likely to be cooperative as they put more emphasis on the relationship with the group than on personal needs (Ozdemir and Hewett, 2010). This indicates that service providers with higher collectivist tendencies are more actively engaged in emotional labor to achieve corporate goals. Therefore, collectivism can moderate the influence of person-organization fit (P-O fit) on emotional labor.

This study aims to examine the influences of relationship bonds on emotional labor through person-organization fit: The influence of the three elements of relationship bond (financial bonds, social bonds, and structural bonds) on P-O fit of service providers, the influence of P-O fit of service providers on their emotional labor, and moderating effects of collectivism between P-O fit and emotional labor.

# THEORETICAL BACKGROUND AND RESEARCH HYPOTHESES

# Emotional Labor

Emotional labor, using employees' emotions to enhance goals of firms (Grandey and Melloy, 2017), is the effort to expose the emotions actually felt by employees as the emotions wanted by customers after regulating and managing them (Grandey, 2000; Molino et al., 2016). Corporations set employees' emotions and the expression modes toward customers and employees moderate their own emotions by conforming to emotional display rules (Ashforth and Humphrey, 1993). Such an emotional labor has a lot to do with service providers who interact with customers instead of corporations (Ashforth and Humphrey, 1993). Since service providers not only perform work while interacting with customers but also become a part of outcomes of expressing emotions, moderating one's emotions is direct and essential. Thus, emotional labor of service providers is higher than that of other office workers or physical laborers (Hochschild, 1983; Grandey and Melloy, 2017).

Previous studies divided emotional labor strategies into surface acting and deep acting (Grandey, 2000; Yoo and Arnold, 2016). Surface acting is expressing emotions after matching them with emotional display rules required by corporations (van Gelderen et al., 2017) and deep acting is matching individual's emotions with emotional display rules required by corporations (Hennig-Thurau et al., 2006). In other words, with regard to surface acting, emotions are suppressed and expressed deceivingly in order to conduct emotional display rules (Ashforth and Humphrey, 1993) and with regard to deep acting, emotions to be expressed to customers are internalized and efforts are made to actually feel them (Yoo and Arnold, 2016). Since service providers must express emotions according to emotional display rules, they have to follow the rules after choosing either surface acting or deep acting (Hochschild, 1983). Therefore, corporations should induce service providers to do deep acting. If service providers do express the emotions required by corporations but feel different emotions, they are stressed from emotional discord. On the other hand, if service providers feel the emotions which must be expressed to customers, emotional discord is removed and positive psychological result is brought about (Brotheridge and Grandey, 2002). In that sense, the factors influencing emotional labor of service providers must be identified.

# Relationship Bonds

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 3

Relationship bonds refers to corporations more effectively reaching stakeholders and forming long-term and continuous relationships (Copulsky and Wolf, 1990; Wu and Lin, 2014). Through such relationship bonds, stakeholders make a commitment to the relationships with corporations since they can get continuous outcomes (Wang, 2014).

Previous studies divided the relationship bonds into financial bonds, social bonds, and structural bonds based on characteristics of relationship bonds, customization, and degree of relationship sustainability (Berry, 1995; Kim and Kim, 2018). Financial bonds refer to providing financial benefits through the most fundamental relationship formation strategy (Berry, 1995) and some examples are corporations providing incentives to customers and striving to form relationships through discounts (Hsieh et al., 2005). Similarly, relationships can be formed with stakeholders other than customers through financial dependence. Financial incentives such as salary become financial bonds to service providers (Wang, 2014) and degree of financial bonds is decided based on level of salary.

Social bond is based on establishing close ties among stakeholders through social interaction or friendship (Liang and Wang, 2005) and they can acquire psychological gains through it (Gwinner et al., 1998). Moreover, establishing relationship through interaction creates emotional engagement among stakeholders and makes no replacing relationship by another target (Lee et al., 2015). Service providers establish and progress relationships through the interaction with other colleagues (Gremler and Gwinner, 2000). Also, interaction among service providers provides psychological benefits by forming intimacy and trust (Rodríguez and Wilson, 2002). Service providers who have established positive social relations become emotionally immersed in that corporation and have the will to belong to it continuously (Wang, 2014). Therefore, from the standpoint of service providers, social bond is the corporate effort to establish social relations.

Structural bond refers to create the value stakeholders need (Wang, 2014). This means providing unique and customized values which cannot be obtained from others and these customized values are generated when the right to make decisions is given to stakeholders (Bolton et al., 2003). Giving the right to stakeholders to make decisions by themselves – the arrangement of the opportunities to receive the services tailored to the needs of stakeholders – can create the values of customization (Nath and Mukherjee, 2012). In the standpoint of service providers, structural bond refers to corporate efforts such as the introduction of policies and programs to design customized jobs. The right to make decisions regarding jobs such as business hours and business schedule by service providers is receiving customized values tailored to the needs of service providers from the corporation (Wang, 2014).

# Relationship Bonds and P-O Fit

Individuals' attitude or behavior may change within organization based on whether there is a high similarity between individuals and organization or the degree of individual's value and organization's values coinciding with each other (Meglino and Ravlin, 1998). P-O fit, a part of person-environment fit, may be defined as the degree of compatibility between individuals and organization (Kristof, 1996). This is determined through how similar organization's values and objectives are with individuals or how far organization can support individual's values and objectives (Downes et al., 2017). When individual's values and organization's values are judged to be similar to each other, individual thinks that he/she is a part of organization (Saks and Ashforth, 1997), perceives organization's goal as his/her goal and, pursues it with the strong sense of unity with organization (Cable and DeRue, 2002). On the contrary, when P-O fit is low, individual experiences psychological discomfort due to the discrepancy between the two different values and shows negative emotion and attitude (Lauver and Kristof-Brown, 2001). For this reason, previous studies identified P-O fit as the predisposing factor influencing job attitude of employees. P-O fit improves job satisfaction and organizational commitment by having positive influence on engagement of employees (Biswas and Bhatnagar, 2013) and lowers turnover intention (Liu et al., 2010). In particular, P-O fit not only lowers emotional exhaustion of service provider but also makes positive influence on occupational behavior of service provider (Yoo et al., 2014). Thus, it is important to understand some underlying factors which influence P-O fit of current employees.

Financial incentives not only motivate individuals but also become an element of values (Miao and Evans, 2007). Individuals interact with organization to obtain financial values. Individuals accomplish the goal organization wants and organization satisfies individuals' financial needs. Through this continuous interaction process, individuals match their values and goals with organization (Abdalla et al., 2018). If individuals do not receive financial compensation which is proper to their performance in interaction with organization, individuals are likely to show negative response after feeling that it is unequal (Newman and Sheikh, 2012). On the other hand, individuals who have received plenty of financial values through interaction with organization feel guilty about the behaviors going against the interests of organization (Wang, 2014). This is the powerful evidence which shows that the values of service providers and the values of firm coincide with each other owing to financial values.

H1. Financial bonds will increase the P-O fit of service providers.

Powerful social interaction is helpful for work performance through relationship network between individuals or enables approaching essential information (Biong and Ulvnes, 2011). Furthermore, one can make joint interpretation with regard to certain phenomenon or situation by sharing and understanding one another's vision, goal, and meaning to a high level (Tsai and Ghoshal, 1998). When this is expanded, the values of organization and the values of individuals coincide with each other through interaction between individuals. Service providers can make specific contribution through values and jobs of organization by sharing information through communication with colleagues. As examined so far, the socialization process within organization fpsyg-10-00370 February 23, 2019 Time: 17:41 # 4

plays an important role in forming values of service providers so that they match corporate values (Kilroy et al., 2017).

H2. Social bonds will increase the P-O fit of service providers.

Structural bonds, providing customized organization system to resolve occupational problems of service providers, improves the relationship between corporation and service providers. Designing corporate system so that it enables flexible business schedule setting have service providers be immersed in corporation by providing customized values which are difficult to get from other corporation (Wang, 2014). Furthermore, structural bond plays the role of the sign of delivering corporate values and goals to service providers. High autonomy with regard to job is perceived as the message that corporation is convinced of individuals' capabilities (Saragih, 2011). Thus, participation in corporate objective development increases and ownership and sharing of corporate values increase (Kilroy et al., 2017). Eventually, corporation can enhance P-O fit by providing the parts necessary in job to service providers and sharing similar characteristics through structural bonds (Kristof, 1996).

H3. Structural bonds will increase the P-O fit of service providers.

# P-O Fit and Emotional Labor

Previous studies identified that P-O fit not only had positive influence on individuals' attitude and performance with regard to job (Downes et al., 2017; Abdalla et al., 2018) but also reduced negative factors such as burnout (Kilroy et al., 2017). Other emotional labor strategies are used in customer service context based on the degree of service providers perceiving P-O fit. When service providers perceive that their own values coincide with corporate values, they enthusiastically immerse themselves into job (Kim et al., 2013). In other words, since service providers take conducting emotional labor to accomplish corporate goals as one's own role, service providers consciously try to revise their own emotions and internalize the emotions required by corporation. Thus, service providers earnestly express emotions through deep acting. On the contrary, service providers who experienced discord with organization undergo less positive or even negative emotions (Lauver and Kristof-Brown, 2001; Gabriel et al., 2014). Since service providers do not perceive their own values and corporate values as separate things, service providers are not motivated by the role corporations demands from them. Therefore, service providers can make positive expressions by suppressing or deceiving their emotions but do not revise emotions or make the effort to internalize display rules (Kilroy et al., 2017). Based on what has been examined so far, the following hypotheses have been set.


# Moderating Effect of Collectivism

The studies on collectivism either examine the difference of value between countries with different cultures (Takano and Osaka, 1999) or are taking place on the aspect of exploring different tendencies of individuals (Arpaci et al., 2018). The purpose of this study is not comparing nations with different cultural values but examining how individual tendencies have varying influence on occupational aspects. Thus, this study aims to explore the effect of collectivism in terms of personal aspect. Collectivism has the characteristic of binding individuals through strict social framework and dividing them into in-group and out-group. Individuals display high loyalty toward in-group and expects that in-group will look after them (Hofstede, 1980). Service providers with strong collectivism put greater emphasis on relational aspect than their needs and make responses which are cooperative toward organization after perceiving relational behaviors as important (Ozdemir and Hewett, 2010). Thus, if service providers cannot match their objectives to organizational goals (Triandis and Gelfand, 1998) and cannot abide by the norms between organization and themselves, service providers come to have greater will to accomplish the goal desired by organization since they feel guilty (Li and Dant, 1997). Since individuals with collective values have interest toward harmonizing with organization (Mulki et al., 2015), they can match their own values with organization's values faster and stronger. Service corporation aims to satisfy customers by delivering intangible services and emotional display rules of service providers are performed at this time. If service providers have a high collectivism value, they strive to more actively internalize emotional display rules required by corporation since individual values and corporate values are matched stronger. Thus, surface acting will be weakened and deep acting will be strengthened.


# METHODOLOGY AND MODEL TESTING

# Participants and Procedure

This study collected data from 350 employees working in the financial industry in South Korea through online questionnaire survey panel and analyzed them. The employees are the leading service providers performing emotional labor while frequently coming into contact with customers. The corresponding panel has a lot of participants who agreed to take part in the study. The authors requested survey participants to choose their own occupation among several job categories and those who were not service providers were excluded in the survey. Then, respondents were asked about the industry they belong to and the survey was conducted for those who chose the financial industry.

The gender of the respondents was comprised of 195 men (55.7%) and 155 women (44.3%). When it came to age bracket, those in their 30s was the highest with 120 employees (34.3%) and it was followed by those in their 40s (109 employees, 31.1%), those in their 20s (80 employees, 22.9%), and those fpsyg-10-00370 February 23, 2019 Time: 17:41 # 5

in their 50s (41 employees, 11.7%). When it came to work period, less than 3 years were 103 employees (29.5%), 3 years and longer and less than 6 years were 84 employees (24.0%), 6 years and longer and less than 9 years were 47 employees (13.4%), 9 years and longer and less than 12 years were 47 employees (13.4%), and 12 years and longer were 69 employees (19.7%). When it came to position, 178 respondents were staff (50.9%), 70 respondents were administrative managers (20.0%), 75 respondents were section chiefs (21.4%), and 27 respondents were general managers (7.7%).

# Measures

Relationship bond was assessed with ten items developed by Wang (2014). Financial bond was measured by three items.


The bolded value in the diagonal is the square root of the AVE; <sup>∗</sup>p < 0.05, ∗∗p < 0.01.

An example item is "My company provides satisfactory total income." Cronbach's alpha was 0.96. Social bond was measured by four items. An example item is "My company supports me so that I communicate well with co-workers." Cronbach's alpha was 0.88. Structural bond was measured by three items. An example item is "My company allows me to flexibly manage work schedules." Cronbach's alpha was 0.87. Correlations among the study constructs are displayed in the **Table 1**.

P-O fit was assessed with four items developed by Saks and Ashforth (2002). An example item is "To what extent are the values of the organization similar to your own values?" Cronbach's alpha was 0.88. Collectivism was assessed with three items developed by Wu (2006). An example item is "Group success is more important than individual success." Cronbach's alpha was 0.87.

Emotional labor was assessed with eight items developed by Diefendorff et al. (2005). Surface acting was measured by four items. An example item is "I fake the emotions I show when dealing with customers." Cronbach's alpha was 0.85. Deep acting was measured by four items. An example item is "I try to actually experience the emotions that I must show to customers." Cronbach's alpha was 0.86.

All items were scored on a five point scale, ranging from 1 = strongly disagree to 5 = strongly agree. Measurement scales of the study constructs are shown in the **Appendix**.

# Convergent and Discriminant Validity of Constructs

Prior to verifying hypotheses, this study identified convergent validity and discriminant validity of constructs through confirmatory factor analysis (Nunnally, 1978; Fornell and Larcker, 1981). Convergent validity was identified by composite reliability (CR) and average variance extracted (AVE). The evaluation standard of CR is 0.7 and higher and the evaluation standard of AVE is 0.5 and higher. In a bid to identify discriminant validity, whether the square root value of AVE exceeded the correlations between constructs was examined. Since the fit of confirmatory factor analysis was shown as χ <sup>2</sup> = 574.59(df = 254, p = 0.00), GFI = 0.88, CFI = 0.95, NFI = 0.91, RMSEA = 0.06, they were generally acceptable. Also, since the standards which could evaluate convergent validity exceeded the standard level, convergent validity was identified. As a result of identifying discriminant validity, the square root value of AVE with regard to constructs had been found to exceed the correlations between constructs. Therefore, discriminant validity had been found to have no problems.

# Common Method Bias

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 6

Since self-reported measures are utilized, there is a danger of common method bias. In their argument of the causes and complications of this bias, Podsakoff et al. (2003) suggested both procedural antidotes, such as acquiring measures from different respondents or methodologically disconnecting measures, and statistical antidotes to diagnose the bias. This study adopts Harman's single factor test that Podsakoff et al. (2003) announced as a statistical remedy. Adopting exploratory factor analysis, the number of factors and variances inferred from the unrotated factor solutions are tested. The analysis identifies seven factors with an eigenvalue greater than 1; their variances are 30.92, 12.30, 10.27, 7.83, 6.02, 5.69, and 4.05%. As the analysis identifies more than one factor and the factor with the highest variance elucidate less than 50% of the total variance, it can be considered that the study does not have a serious common method bias.

# Test of Hypotheses

Since the fit of this study model was χ <sup>2</sup> = 529.94(df = 201, p = 0.00), GFI = 0.88, CFI = 0.94, TLI = 0.93, NFI = 0.90, RMSEA = 0.07, it is generally at an acceptable level. **Figure 2** is the result of verifying hypotheses through structural equation model. This study expected that relationship bonds will have positive influence on P-O fit of service providers. As a result of identifying the hypotheses, relationship bonds elements – financial bonds (β = 0.31, p < 0.01), social bonds (β = 0.44, p < 0.01), and structural bonds (β = 0.28, p < 0.01) – had positive influence on P-O fit. Thus, H1, H2, and H3 were supported. Next, this study expected that P-O fit will have negative influence on surface acting and positive influence on deep acting. As a result of identifying the hypotheses, P-O fit did not have any influence on surface acting (β = 0.04, p > 0.05) but had positive influence on deep acting (β = 0.49, p < 0.01). Therefore, Hypothesis 4 was rejected but Hypothesis 5 was supported.

# Test of Moderating Effect

This study aimed to identify the moderating effects of collectivism between P-O fit of service providers and emotional labor. Since P-O fit had been found to not have any influence on surface acting, the moderating effect toward deep acting was identified. In order to identify moderating effects, interaction effect was verified through the process model using bootstrap (Preacher and Hayes, 2008; Hayes, 2013). Hayes (2013) used No.1 model out of the suggested process models, Y inserted deep acting, X inserted P-O fit, and M inserted collectivism. In addition, education, work period, position and income of service providers were controlled. As a result of the analysis (**Table 2**), the interaction effect of collectivism was statistically significant with regard to deep acting (β = 0.1197, CI = [0.0040, 0.2354], p < 0.05). In a bid to specifically identify moderating effect,


fpsyg-10-00370 February 23, 2019 Time: 17:41 # 7


Process model 1; N = 350; Bootstrap samples 5,000.

post hoc test was conducted. For this aim, low level simple slope (−1 SD) and high level simple slope (+1 SD) were identified from the average of collectivism (Aiken and West, 1991). As a result of the post hoc test, as shown in **Figure 3**, the influence by P-O fit on deep acting increased as collectivism of service providers went higher. Thus, Hypothesis 7 was supported.

# CONCLUSION AND DISCUSSION

This study examined relationship bonds for corporation to form relationship with service providers. This study developed and tested the model which identified the influence on emotional labor by relationship bonds through P-O fit and the role of collectivism. This study discovered the following findings. Three elements of relationship bonds – financial bonds, social bonds, and structural bonds – are positively related to deep acting through the increase of P-O fit. The influence on deep acting by P-O fit is moderated by collectivism of service provider. Specifically, as collectivism of service provider goes higher the effect of P-O fit increasing deep acting becomes stronger.

## Theoretical Implications

There are several theoretical contributions. First, it contributed to relationship marketing literature by applying the concept of relationship bonds into service providers. Owing to the influence by service providers regarding customers, the importance of forming relationship between corporation and service providers is being emphasized (Tortosa et al., 2009). Relationship marketing literature found that relationship bonds is the strategy for forming relationships and the antecedent factor for accomplishing relationship performance (Dash et al., 2009). Previous studies identified relationship bonds as the strategy to maintain relationship between corporation and customers or corporations (Wang, 2014) but the studies on service providers merely stay at initial level. Moreover, the studies have identified that performances changing owing to the ways corporation handle service providers but academic discussion on forming relationship between corporation and service providers is not taking place (Herington et al., 2006). This study aimed to overcome the limitation of previous studies by examining relationship bonds from the aspect of service providers.

Second, this study suggested the empirical evidence identifying the process of relationship bonds influencing emotional labor. Since service providers directly perform emotional labor to customers, it is important to identify the factors influencing emotional labor of service providers. The previous studies on emotional labor treated individual characteristics – personality traits, emotional abilities, and so on – and event characteristics – moods, customer mistreatment,

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 8

and so on – as major predisposing factors (Grandey and Gabriel, 2015). However, there was no empirical evidence identifying how relationship bonds, forming relationship between corporation and service provider, was related to emotional labor. This study suggested relationship bonds as the major previous study having influence on emotional labor.

Third, owing to relationship bonds, this study identified the roles of P-O fit when emotional labor response is occurring. This study provides the understanding of the mechanism explaining how environmental factors of service providers affect emotional labor. Emotional labor is an essential element for service providers and this implies that it is crucial for corporations, which have to induce positive emotional labor strategies from service providers, to give support so that individuals feel conformity with organization. Almost all service sites require emotional labor (Tschan et al., 2005). Therefore, understanding of P-O fit roles in service jobs is a good way of carrying out positive emotional labor strategies in service industry which frequently interacts with customers.

Fourth, this study was expanded since the moderating effect of collectivism between P-O fit and emotional labor was identified. Previous studies suggested the moderating role of collectivism from cultural psychology aspect with regard to work behavior and performance of employees (Hui et al., 2015; O'Neill et al., 2015). However, specific role between fit and emotional labor was not mentioned. This study suggested clear and specific viewpoint explaining emotional labor from cultural psychology aspect by identifying the moderating effect of collectivism between P-O fit and emotional labor.

# Practical Implications

Through the results of this study, several practical implications can be discussed. First, corporations can improve P-O fit of service providers through relationship bonds. Since P-O fit enhances performance by being positively applied to service providers (Downes et al., 2017; Abdalla et al., 2018), it is important to manage relationship bonds. Corporate financial support becomes the element of value to individuals and acts positively toward accomplishing organization's goals since individuals equate themselves and organization (Miao and Evans, 2007). Therefore, corporations need to form financial bonds by establishing financial compensation system which service providers perceive as sufficient. In addition, corporations need to come up with the measures to improve the social relations among service providers. Communication among service providers improves the ability or skills to resolve problems by promoting information sharing (Felício et al., 2014). Thus, corporate efforts to establish positive relationship among service providers by invigorating corporate communities and organizing horizontal corporate culture are necessary. Through these efforts, social sense of belonging by service providers can also be fulfilled. Structural bond is a positive factor toward P-O fit. The system of service providers flexibly adjusting one's own work schedule assists the balance between work and life and immerses workers into jobs (Martin and MacDonnell, 2012). Since corporate efforts to form structural bonds plays the role of resources in service providers doing work, the efforts not only appear as performance by being connected to work efficiency improvement but also reduces the expenses arising from stress (Zablah et al., 2012).

Second, P-O fit can improve deep acting of service providers. Corporation needs to interact with service providers to deliver the goals and values pursued by corporations. Corporations can induce positive behaviors by enhancing similarity between corporation and service providers (Meglino and Ravlin, 1998). Since deep acting by service providers lets customers to experience high service quality after delivering sincere emotions (Pugh, 2001; Hennig-Thurau et al., 2006), relationship bonds strengthens deep acting through P-O fit and is positively related to performance. Meanwhile, the influence by P-O fit on surface acting was not significant. Surface acting, service providers expressing their emotions to customers after controlling and adjusting them, appears in the interaction with customers. Therefore, the negative factors related to customers such as customer incivility have direct influence on surface acting (Sliter et al., 2010). Since P-O fit is not the factor related to direct interaction with customers, it cannot influence surface acting.

Third, collectivism increases the influence by P-O fit with regard to deep acting of service providers. This study suggested that corporation managers must take into account cultural values. In particular, since corporate global management increases, managers must consider countries' cultural backgrounds and values. This can provide better insight and understanding regarding employment, arrangement, and education of service providers when it comes to service firms entering new markets. While Asian countries like Korea can use the method of hiring collective individuals as the corporate measure to manage emotional labor, U.S. has limited number of applicants with collective values (Astakhova, 2016). Therefore, service firms must apply management styles suiting cultural norms so as to accomplish maximum results by hiring local talents.

# Limitations and Future Direction of Study

Even though this study suggested theoretical and practical implications, it has several limitations. First, this study examined the factors of relationship bonds into three aspects: financial bonds, social bonds, and structural bonds. Service providers border between corporation and customer owing to occupational characteristics. Thus, social bonds has to be examined by dividing into social bonds of internal aspect related to corporation and social bonds of external aspect related to customer. However, this study examined social bonds only in the aspect related to corporation. The scope of study will be expanded further if social bonds is divided into two aspects.

Second, this study identified that relationship bonds act positively in deep acting through P-O fit of service providers. It can be expected that performance will improve if deep acting increases but this study did not reflect performance aspect. Future studies need to suggest more completed study model by taking into consideration performance aspect of service corporations.

Third, this study collected data targeting service providers in the financial industry. The employees in the financial industry have the characteristics distinct from other industries. For instance, services dealt by employees in the financial industry are more complicated than other service industries. This may make applying this study results into the entire service industries difficult. Therefore, the generalization of this study results has to be induced by conducting studies targeting diverse industries.

Fourth, this study identified which organizational level factors influence emotional labor through P-O fit of current employees. It is also important to identify the effects of job characteristics on emotional labor through P-O fit and therefore, future study needs to investigate how P-O fit can be influenced by job characteristics of current employees.

# ETHICS STATEMENT

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 9

This study was carried out in accordance with the recommendations of ethics committee of Hanyang University with

# REFERENCES


written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethics committee of Hanyang University.

# AUTHOR CONTRIBUTIONS

M-SL designed the work and analyzed the data. S-LH designed and drafted the work. SH collected and analyzed the data. HH collected and analyzed the data.

# FUNDING

This research was supported by the research fund of Hanyang University (HY-2018).


fpsyg-10-00370 February 23, 2019 Time: 17:41 # 10

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 11

and positive affective displays. Acad. Manag. J. 51, 131–146. doi: 10.5465/amj. 2008.30764063


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Lee, Han, Hong and Hyun. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# APPENDIX

TABLE A1 | Items and CFA results.

fpsyg-10-00370 February 23, 2019 Time: 17:41 # 12


<sup>a</sup>All factor loadings are significant (p < 0.01).

<sup>b</sup>CR, composite reliability.

<sup>c</sup>AVE, average variance extracted.

# The Role of Emotional Service Expectation Toward Perceived Quality and Satisfaction: Moderating Effects of Deep Acting and Surface Acting

### Ji Youn Jeong<sup>1</sup> , Jungkun Park<sup>2</sup> \* and Hyowon Hyun<sup>2</sup>

<sup>1</sup> School of Ecological Environment and Eco-tourism, Kyungpook National University, Sangju, South Korea, <sup>2</sup> School of Business, Hanyang University, Seoul, South Korea

A conceptual model articulating the nature of customer expectations and satisfaction over services was proposed with emotional factors. Five propositions about consumer emotional service expectations as a primary antecedent toward confirmation, perceived quality, and satisfaction were provided. As moderators, two dimensions of consumer detection of emotional labor (i.e., detecting deep acting and surface acting) were imposed on each of the relationships. Evidence demonstrated the roles of emotional service expectation in service confirmation and satisfaction. The moderating effects of consumer detections of employees' emotional strategies were limited to the relationship between emotional service expectation and confirmation; its relationship was weakened by detections of surface acting while the other relationships were not moderated by detections of deep nor surface acting. Structural equation modeling analyses were conducted using online survey data targeting consumers in the hotel industry.

### Edited by:

Monica Gomez-Suárez, Autonomous University of Madrid, Spain

### Reviewed by:

Teresa Villacé-Molinero, Universidad Rey Juan Carlos, Spain Jorge Pelegrín-Borondo, University of La Rioja, Spain

### \*Correspondence:

Jungkun Park viroid2007@gmail.com

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 11 September 2018 Accepted: 01 February 2019 Published: 12 March 2019

### Citation:

Jeong JY, Park J and Hyun H (2019) The Role of Emotional Service Expectation Toward Perceived Quality and Satisfaction: Moderating Effects of Deep Acting and Surface Acting. Front. Psychol. 10:321. doi: 10.3389/fpsyg.2019.00321 Keywords: emotional service, customer expectation, satisfaction, emotional labor, deep/surface acting

# INTRODUCTION

Having motivated employees who proactively and professionally engage in customer interactions is undoubtedly critical to service firm's success (Menguc et al., 2013; Schepers et al., 2016). In the hospitality industry, employees are deeply involved in everyday interactions with customers, such interactions being regarded as essential to the delivery of quality service experiences to customers (Farrell and Oczkowski, 2009). As a result, establishing effective strategies for employee–customer interactions, is recognized as the most important step in enhancing firms' competitiveness (Lam et al., 2018). Indeed, huge resources are invested in training employees to enhance their interaction performance in the hospitality industry (Schepers et al., 2016).

Some researchers argue that service components are categorized into two attributes: (i) the core attributes that consist of functional quality factors; and (ii) the relational attributes that describe the interpersonally recognized factors during the service delivery (Babin et al., 2004). On the other hand, other researchers suggest that such a classification is artificial and fuzzy because the nature of services involves intangible and experiential consumption (Gaur et al., 2014; Koenig-Lewis and Palmer, 2014). They argue that factors perceived interpersonally by customers can be a core factor that significantly influences service quality, increasing consumer satisfaction and loyalty

**111**

behaviors. These conflicting arguments lead to a main question of this study: does employees' emotional service delivery during customer interactions, that is one of the relational attributes, significantly influence service outcomes, such as service confirmation, quality perceptions, and satisfaction? The second research question is as follows: if so, do different detections of employees' emotional service strategies lead to different customer reactions? Typically, service employees are expected to express positive emotions and to suppress negative emotions in their interactions with customers (Diefendorff and Richard, 2003), but it is inherently impossible always to feel genuine positive emotions for customers (Groth et al., 2009). Consequently, employees attempt to utilize surface acting, faking a positive emotional display, to comply with their job requirements and meet customer needs for positive emotional services (Hochschild, 1983; Lam et al., 2018). However, customers may detect surface acting, and show less positive responses (i.e., service satisfaction and perceived quality) than when detecting deep acting (Lam et al., 2018).

To address these two research questions, we developed a theoretical model consisting of customer emotional expectations, overall confirmation, perceived quality, satisfaction, and two moderators of consumer detection of deep acting and surface acting. Affective elements have been well-reported in relation to the expectation – performance discrepancy link (Burns and Neisner, 2006) and customer satisfaction (Oliver, 1980). Nevertheless, the relationships between customer's emotional service expectations and consumer satisfaction have not been fully addressed in the service literature. Furthermore, it has not yet been concluded how customer detections of employees' emotional service strategies influence such relationships among emotional expectation, service quality perception, and satisfaction (Groth et al., 2009). Thus, this study would fill the void in the literature by addressing emotional service expectation and customer detection of the emotional display strategies in the hospitality service context.

# LITERATURE REVIEW AND HYPOTHESES

# Emotional Service Expectation and Confirmation

Emotional labor can be defined as service employees' efforts to demonstrate and express desired emotions at work by managing feelings (Hochschild, 1983; Ashforth and Humphrey, 1993; Grandey, 2000). For example, Hochschild (1983) defined emotional labor as "the management of feelings to create a publicly observable facial and bodily display" which is "sold for wage and therefore has exchange value" (p. 7). Similarly, Grandey (2000, p. 97) defined emotional labor as the "process of regulating both feelings and expressions for the organizational goals." Previous literature has indicated critical roles of emotional labor in the service context. Tsai (2001) stated that employees' emotional services are all part of the service itself. Hochschild (1983) also pointed out that "the emotional style of offering the service is part of the service itself " (p. 5). The importance of emotional services in confirming customer expectation is unanimously accepted among scholars and practitioners in the service field (Park et al., 2018).

The notion of confirmation has been explained in relation to expectation, which is defined as belief probabilities of the consequences of an event (Roch and Poister, 2006). Burgoon and Walther (1990, p. 236) identified expectation as "what is predicated to occur rather than what is desired." According to expectation violation theory (Burgoon, 1993), an individual interacts with others with expectancy that refers to what will happen in a given situation and so, the disconfirmation of such expectancy negatively influences outcomes of the interaction. In the hospitality service context, customers seem to have a certain level of emotional service expectation and want to confirm it during the interaction; they expect employees to express positive emotions, such as friendliness, positivity, compassion, and/or warmth, while suppressing negative emotions, such as anger, indifference, or frustration (Beal et al., 2006; Grandey and Gabriel, 2015).

The literature has suggested that customer expectation positively influences confirmation. For example, Park et al. (2018) argued that high expectation increases customers' involvement levels, leading to easy confirmation. Groth et al. (2009) also empirically demonstrated that positive attitude with high expectation results in high confirmation. Consistent with the results in the existing studies, this study hypothesized that customers' expectation of emotional services will positively influence overall confirmation of a service outcome.

**H1:** Emotional service expectation has a positive effect on overall confirmation.

# Emotional Service Expectation Toward Perceived Quality and Customer Satisfaction

In their seminal work on SERVQUAL, Parasuraman et al. (1985) indicated that service quality is determined by the evaluation of a service provider by comparing the service provider's performance with the customer's expectations of how similar service providers should perform. In other words, customers set the range of expected outcomes of emotional services as a quality standard in which a given service quality is likely to be evaluated (Tsai, 2001; Humphrey et al., 2007). Therefore, quality of services depends on customers' expectation of how effectively and positively employees express positive emotions. In addition to service expectation, overall confirmation is likely to directly influence quality perception. Parasuraman et al. (1988) stated that service quality results from overall assessments in an integral dimension of expectations. Grönroos (1984) addressed service quality can be measured in the two dimensions of process quality and technical quality, and these two dimensions are intricately connected, indicating service quality is related to overall confirmation. Confirming this argument, Homburg and Stock (2004) argued that quality perception is related to customers' confirmation of cognitive and affective factors of services. In a hotel setting, satisfying interaction experiences with employees confirmed

customer expectation of emotional services and so, enhanced overall service quality perception (Ladhari, 2009). Further, Brady and Cronin (2001) demonstrated quality perception is determined by overall impression of a given service. Since the nature of service involves intangibility and inseparability, service quality seems to be influenced by overall confirmation of expectations. In summary, the literature suggested that perceived quality is influenced by customer expectation and overall confirmation, so we hypothesized as follows:

**H2**: Emotional service expectations have a positive effect on perceived quality.

**H3:** Overall confirmation has a positive effect on perceived quality.

Gracia et al. (2011) found that five service quality components, including reliability, assurance, responsiveness, empathy, and tangibles, have positive effects on customer emotions and, as a result, increase satisfaction in the hotel and restaurant service context. Han and Jeong (2013) also empirically demonstrated that customers' service quality perception significantly influences service experiences and emotional satisfaction. Due to their close relationship, quality perception is often used as an indicator of measures of customer satisfaction (Homburg and Stock, 2004; Wong, 2004).

Ladhari (2009) claimed that satisfaction consists of cognitive (i.e., customer's judgment) and affective components (i.e., happy, pleasant, and joyful), both of which are evaluated to confirm service expectation. In addition, customer satisfaction seems to be influenced by various external attributes, such as instore environment and feelings (Han, 2013). Therefore, this study hypothesized that both perceived quality and overall confirmation influence customer satisfaction.

**H4**: Perceived quality has a positive effect on customer satisfaction.

**H5**: Overall confirmation has a positive effect on customer satisfaction.

# Two Emotional Strategies: Deep and Surface Acting

Emotional services, such as displays of enthusiasm, friendless, and warmth, are identified as the important services of worth in service delivery (Diefendorff et al., 2006; Sutton et al., 2013). The literature has confirmed that emotional services have positive effects on customer perceptions and satisfaction (Groth et al., 2009; Grandey and Gabriel, 2015). Lin and Liang (2011) study also confirmed that displays of positive emotions increase customers' positive responses, such as customer delight, repurchase intent, and positive word-ofmouth. Similarly, Johanson and Woods (2008) argued that emotional services help achieve organizational goals, such as customer satisfaction and firms' long-term profitability.

Service firms, in general, require employees to regulate their emotions to express only positive emotions to customers. To comply with the job requirement, employees tend to adopt two main strategies of emotional displays: deep acting and surface acting. Deep acting refers to "good faith" by "putting one's self in another's shoes" (Diefendorff et al., 2006) and it is a genuine feeling created within themselves (Hennig-Thurau et al., 2006). Deep acting is regarded as sincerely showing the emotions that match their genuine feelings and organization desires by feeling customers' feelings and having empathy (Grandey, 2003). In turn, employees who use deep acting are likely to be customeroriented and so, provide their customers sincere service (Allen et al., 2010). Deep acting has been reported to increase job performance and, as a result, lead to positive feedback from customers (Hatfield et al., 1994). In contrast, surface acting involves "simulating emotions that are not actually felt" (Ashforth and Humphrey, 1993, p. 92). Surface acting occurs when service employees modify only their visible emotions and deceive customers by "putting on a mask," without actually changing how they feel. Surface acting can bring out an emotional discrepancy between true feelings and expressions because employees just pretend to feel positive to meet the organization's requirements (Johnson and Spector, 2007).

# The Moderating Role of Employee Deep/Surface Acting

Mattila and Enz (2002) stated that customers' evaluation of service consumption experience depends on how effectively employees display positive emotions during customer interactions. Deep acting strategy allows employees to express their genuine feelings in line with the desired emotions that service firms require, and customers expect (Diefendorff and Greguras, 2009). Deep acting is related to trustworthiness and authenticity that have been known as the main components of service performance to enhance quality perceptions and satisfaction (Krumhuber et al., 2007). Indeed, employees who use deep acting are likely to understand customers and respond to their needs well (Sandström et al., 2008).

When employees cannot modify their inner feelings, employees may use the surface acting strategy (Diefendorff and Greguras, 2009), but this may result in negative outcomes (Hatfield et al., 1994). Grandey et al. (2005) stated "when service providers do not seem sincere in their expressions. . . it is less likely to create a positive impression in the customer; instead, a false smile may seem manipulative and the employee's impression management attempt fails" (p. 52). Since people tend to prefer honesty and authenticity in social interactions, employees' fake emotional displays are unlikely to meet customers' emotional expectation (Grandey, 2003; Hennig-Thurau et al., 2006). From the employees' perspective, surface acting may also impair their service performance in that expressing a fake emotion involves additional waste of cognitive resources, affecting job satisfaction (Johnson and Spector, 2007). According to fit theory, when there is wide discrepancy between outward and inward feelings at work, employees are likely to be frustrated and dissatisfied with their jobs, decreasing work performance (Gabriel et al., 2015) which is a primary reason of negative feedback from customers (Heskett et al., 1994).

All things considered, it was anticipated detecting deep acting is likely to strengthen the link of expectation-confirmation toward satisfaction, while detections of surface acting may

counteract it. In this study, Hypotheses 6 through 10 were proposed with regard to the positive moderating effects of customer detection of deep acting while Hypotheses 11 through 15 were about the negative moderating effects of customer detection of surface acting. All hypothesized paths in the study model are shown in **Figure 1**.

**H6–10:** Customer detection of deep acting positively moderates the relationship among the study constructs (i.e., emotional service expectation, overall confirmation, perceived quality, and satisfaction).

**H11–15**: Customer detection of surface acting negatively moderates the relationship among the study constructs (i.e., emotional service expectation, overall confirmation, perceived quality, and satisfaction).

# MATERIALS AND METHODS

# Participants and Procedure

A web-based online survey using email invitation was conducted using a nationally recognized consumer research panel service. A sample of 220 individuals from across the United States over 20 years old was used with screening questions of their hotel service experience with scenario explanation. All participants who satisfied the sampling conditions, opted in to the survey in exchange for a credit of \$1. Participants answered the questions to examine the role of emotional service expectation toward customer satisfaction and the moderating roles of detections of deep acting and surface acting on their relationships in the service industry. The measurement items were adopted based on the literature review. **Table 1** shows demographic results, including gender, age, income, marital status, education level, and annual income. Males made up 51.4% and 48.6% was female. Participants were categorized by age into young adults (ages 20– 39 years; n = 142) and older adults (ages 40–50 and over+ years, n = 78). More than half of the sample had less than \$50,000 in annual income.

# Measures

The questionnaires contained question items focusing on the emotional service expectation, overall confirmation, perceived quality, customer satisfaction, and customer detection of deep and surface acting. Latent variables are not directly observable, so compound multi-item scale measures, including at least two items for each variable, were used (Kenny, 2014). Based on the existing literature, well-validated measurement items for study constructs were adopted and included in the questionnaire. Specifically, emotional service expectation measures were adopted from Davis et al. (1999) three-item scale using sympathy, enjoyment, and compassion dimensions. Overall confirmation was assessed using a two-item scale developed by Oliver (1980). For perceived service quality, three items were adopted from Fornell and Larcker (1981) study. A 5-point Likert-type scale was used, with response options ranging from 1 to 5 in three different dimension standards (e.g., 1 = "poor," 5 = "excellent"; 1 = "inferior," 5 = "superior"; and 1 = "low," 5 = "high"). For customer satisfaction, two items were borrowed from Westbrook and Oliver (1991) study, designed to measure customer satisfaction and emotions with services.

TABLE 1 | Respondents' socio-demographic characteristics.

fpsyg-10-00321 March 8, 2019 Time: 17:21 # 5


To assess customer detections of employees' deep and surface acting, three-item measures were derived from previous study (Groth et al., 2009). These measures were developed "from Grandey (2003), originally developed by Brotheridge and Lee (2003)" (Groth et al., 2009, p. 964). Given the fact that most recent studies still adopt the measures of Brotheridge and Lee (2003) or Grandey (2003), Groth et al. (2009) measures are relatively updated (see Baranik et al., 2017; Uy et al., 2017; Lam et al., 2018; Moin, 2018). Except for quality dimensions, all other variables consistently used the same semantic measurement tool (e.g., 1 = "I would not feel this way at all," 5 = "I would feel this way very much"). All items included in each variable are reported in **Table 2**.

# RESULTS

## Measurement Model Tests

Before analyzing the structural model, the measurement model was assessed with Confirmatory Factor Analyses (CFA) using AMOS 22 and SPSS. CFA evaluated goodness of fit with the six key variables, including emotional expectations, conformation, quality perception, satisfaction, detections of deep acting and surface acting. The results confirmed good fit indices (χ 2 <sup>86</sup> = 175.408, p < 0.000; RMSEA = 0.063; CFI = 0.978; NFI = 0.959; GFI = 0.925; RFI = 0.942; IFI = 0.978; TLI = 0.970) TABLE 2 | Items, standardized factor loadings and Cronbach's α.


<sup>1</sup>Reverse coded item.

of the measurement model. Factor loadings for the indicators for each variable were all significant and sufficiently higher than the recommended value of 0.50 (**Table 2**), indicating convergent validity (Grayson and Marsh, 1994). All Cronbach's alpha values for each variable were above the minimum threshold of 0.70 (Hair et al., 1998), suggesting internal consistency in measurement items. As shown in **Table 3**, composite reliability values for each construct ranged from 0.815 to 0.975, greater than 0.70, indicating good reliability (Hair et al., 2011). Average value extracted (AVE) values ranged from 0.609 to 0.929, all above 0.50, confirming convergent validity (Hair et al., 2011). Also, these AVE values were all greater than the square of correlation between pairs of constructs, achieving discriminant validity (Fornell and Larcker, 1981). All constructs' correlations were less than the threshold of 0.85, confirming further discriminant validity (Kline, 2005). The means, standard deviations, composite reliability, average variance extracted, and Pearson's correlations of variables are reported in **Table 3**.

TABLE 3 | Means, standard deviations, reliabilities, AVE, and correlations.


DA, deep acting; SA, surface acting; ESE, emotional service expectation; OC, overall confirmation; PQ, perceived quality; CS, customer satisfaction; SD, standard deviation; AVE, average variance extracted. Model measurement fit: χ 2 <sup>86</sup> = 175.408, p < 0.000; RMSEA = 0.063; CFI = 0.978; NFI = 0.959; GFI = 0.925; RFI = 0.942; IFI = 0.978; TLI = 0.970. <sup>a</sup>Composite reliabilities highlighted in shade are along the diagonal. <sup>b</sup>Correlations between constructs are below the diagonal. <sup>c</sup>Squared correlations between constructs are above the diagonal.

TABLE 4 | Results of the tests of path coefficients.


∗∗∗p < 0.001, ∗∗p < 0.01, <sup>∗</sup>p < 0.05. Goodness-of-fit: χ 2 <sup>25</sup> = 53.801, p < 0.000; GFI = 0.961; NFI = 0.980; CFI = 0.989; RFI = 0.964; IFI = 0.989; TLI = 0.981; RMSEA = 0.066. ESE, emotional service expectation; OC, overall confirmation; PQ, perceived quality; CS, customer satisfaction.

# Structural Model Tests

The results of the structural equation modeling (SEM) with maximum likelihood estimation procedure showed a good model fit (χ 2 <sup>25</sup> = 53.801, p < 0.000; χ 2 /df = 2.151; GFI = 0.961; NFI = 0.980; CFI = 0.989; RFI = 0.964; IFI = 0.989; TLI = 0.981; RMSEA = 0.066). **Table 4** presents empirical findings of the hypothesized relationships within the original model. Emotional service expectations were found to significantly and positively influence the overall confirmation (β = 0.582, p < 0.001, H1), but not significantly influence perceived quality (β = −0.054, p = 0.401, H2). Overall confirmation significantly and positively linked to perceived quality (β = 0.899, p < 0.001, H3) and customer satisfaction (β = 0.930, p < 0.001, H4). The hypothesized path from perceived quality to customer satisfaction was found to be insignificant (β = 0.004, p = 0.973, H5). Overall, the findings supported Hypotheses 1, 3, and 4, but provided no support for Hypotheses 2 and 4.

# Tests of Measurement Invariance

Prior to testing moderating effects, a measurement invariance test across two groups of moderators was conducted (Steenkamp and Baumgartner, 1998). In this study, respondents were divided into two groups of each of moderators of deep and surface acting by a median-split method: high (n = 130) and low (n = 90) groups of deep acting; and high (n = 113) and low (n = 107) groups of surface acting. For each moderator, an unconstrained model (baseline model) and a constrained model (invariance model) were generated and tested. In the invariance model, factor loading, factor variances, and covariances were constrained to be equivalent across the two groups from the baseline model. As indicated in **Table 5**, Fit indices of Root Mean Square Error of Approximation (RMSEA) and Comparative Fit Index (CFI) were checked because they are relatively less sensitive to sample size

to sample size (Byrne, 2001). With regard to deep acting, the overall model fits were excellent both for the baseline model (χ 2 <sup>52</sup> = 125.213, CFI = 0.966; RMSEA = 0.073) and the invariance model (χ 2 <sup>58</sup> = 134.694, CFI = 0.965; RMSEA = 0.071) (**Table 5**). The difference in χ 2 between those two models was insignificant (1χ 2 <sup>6</sup> = 1.580, p > 0.05), supporting the invariance model. The results confirmed that the measurement model was equivalent across high and low groups, so the invariance model was employed for the subsequent analyses of moderating effects of deep acting.

(Fan et al., 1999), while the Chi-square index is relatively sensitive

The structural invariance test for surface acting were conducted in the same way; the baseline (unconstrained) model (χ 2 <sup>52</sup> = 118.909, CFI = 0.976; RMSEA = 0.070) and the invariance (constrained) model (χ 2 <sup>58</sup> = 123.271, CFI = 0.976; RMSEA = 0.066) showed excellent fits to the data. Again, there was no significant difference in χ <sup>2</sup> between the baseline and invariance models (1χ 2 <sup>6</sup> = 0.702, p > 0.05), supporting the invariance model. Therefore, the invariance model was adopted for subsequent analysis of moderating effects of surface acting.

# Tests of Moderating Effect of Deep and Surface Acting

For the moderating effect tests of deep acting, a baseline model was generated by adding the hypothesized links among the variables based on the invariance model in both high and low deep acting groups. As shown in **Table 6**, the baseline model satisfactorily fit to the data (χ 2 <sup>50</sup> = 94.997, p < 0.001, CFI = 0.979; RMSEA = 0.059). In the nested models, the hypothesized paths were constrained to be equal for high and low groups. The baseline model and a series of the nested models (equal path model) were compared in pairs to analyze moderating effects



### TABLE 6 | Results of path and model comparison of deep acting.

∗∗∗p < 0.001, ∗∗p < 0.01, <sup>∗</sup>p < 0.05. <sup>1</sup>Each of the nested models was compared with a baseline model (unconstrained): χ 2 <sup>50</sup> = 94.997, CFI = 0.979; RMSEA = 0.059.

of deep acting. Contrary to our expectations, the differences in χ <sup>2</sup> between high and low deep acting groups on each of the hypothesized paths were all found to be insignificant, rejecting Hypotheses 6, 7, 8, 9, and 10 (**Figure 2** and **Table 6**). In other words, the original relationships among the variables were not moderated by deep acting. The findings related to deep acting were reported in **Figure 2** and **Table 6**.

In the same way, a baseline model and a series of nested models (equal path model) for surface acting were compared in pairs (**Table 7**), and the baseline model provided an excellent fit to the data (χ 2 <sup>50</sup> = 96.225, CFI = 0.983; RMSEA = 0.059). On the path from emotional service expectation to overall confirmation, there was a significant difference in Chi-square between the baseline model and the nested model (χ 2 <sup>51</sup> = 112.931, CFI = 0.978; RMSEA = 0.068, 1χ 2 <sup>1</sup> = 16.706, p < 0.001), in support of Hypothesis 11. The positive effect of emotional service expectation on overall confirmation was significant when surface acting was less detected (β = 0.909, t-value = 5.677, p < 0.001), while such a positive effect turned to be insignificant when surface acting was strongly detected (β = 0.197, t-value = 1.258, p = 0.208). That means that surface acting offsets the positive influence of emotional service expectation on overall confirmation. The other nested models were not significantly different from the baseline model, rejecting Hypotheses 12, 13, 14, and 15. The findings of path model comparisons for surface acting were reported in **Figure 3** and **Table 7**.

# DISCUSSION

This study aimed to develop an alternative conceptual model articulating the role of emotional service expectation toward customer satisfaction and the moderating roles of customer detections of deep acting and surface acting on their relationships; the five paths were proposed among the variables, including emotional service expectation, overall confirmation, perceived quality, and customer satisfaction, and


### TABLE 7 | Results of path and model comparison of surface acting.

∗∗∗p < 0.001, ∗∗p < 0.01. <sup>1</sup> Each of the nested models was compared with a baseline model (unconstrained): χ 2 <sup>50</sup> = 96.225, CFI = 0.983; RMSEA = 0.059.

then detections of deep and surface acting were imposed on each of the paths to verify the moderating effects. We hypothesized that the detection of deep acting would cement each of the relationships while the detection of surface acting would weaken those relationships.

As expected, the role of customer's emotional service expectation was significant in overall service confirmation (Hypothesis 1), which further influenced perceived quality (Hypothesis 3) and satisfaction (Hypothesis 4). The results are consistent with expectation violation theory (Burgoon, 1993) positing that an individual interacts with others with expectancy in which interaction outcomes are determined. Also, the study's results supported Farrell and Oczkowski (2009) arguments that service outcomes are likely to be confirmed by employee–customer interactions that take place for service delivery. However, this study did not find the significant relationship between perceived quality and customer satisfaction (Hypothesis 5); this is not congruent with the dominant belief in existing studies. We postulate, overall confirmation sufficiently strongly influences customer satisfaction (Hypothesis 4), so the influence of perceived quality on customer satisfaction becomes relatively weak in this study construct.

A growing body of research has demonstrated the roles of emotional strategies in customer judgments and perceptions (Groth et al., 2009; Hülsheger et al., 2010; Grandey and Gabriel, 2015). Hülsheger et al. (2010) suggested that influences of emotional strategies on the customer's decisions are significant and universal. However, in this study, the moderating effects of customer detection of emotional strategies were limited to the influence of emotional service expectation on overall confirmation; Customer detection of employees' surface acting negatively influences service confirmation in the beginning stage of interaction (Hypothesis 11) and, as a result, indirectly decreases perceived quality and

satisfaction. Whereas all the other hypothesized moderating effects were found to be insignificant. Especially, detections of deep acting hardly influence customer perception and evaluation of services. These results are somewhat consistent with Hur et al. (2015), indicating there were no direct impacts of employees' deep acting strategies on service assessments.

# Theoretical and Practical Implications

Despite a growing awareness of the critical roles of employees' emotional services, little has been known about how customers' emotional service expectations and employees' emotional displays influence service assessments in the hospitality service context. The major theoretical contribution of this study demonstrates the role of employees' emotional display strategies in the relationship between customer emotional service expectation and overall confirmation that further influenced satisfaction and perceived quality. This study extends the wellknown body of literature articulating the roles of expectation in customer satisfaction (Oliver, 1980; Teas, 1993) by integrating emotional factors in a manner in which customer emotional expectation-confirmation influences perceived quality and satisfaction that are moderated by emotional display strategies. This study provides confirmative evidence that emotional service expectation and emotional display strategies play a critical role in service confirmation, suggesting that comprehensive understating of customer behaviors in the hotel service context is critical.

Also, this study successfully extends the traditional model framework consisting of direct and mediating effects based on expectancy-disconfirmation theory by incorporating emotional components as moderators. There are two contrasting theories explaining the effects of expectations. Direct effect hypothesis claims that expectations have a direct, positive impact on customer satisfaction (Park et al., 2018) while Affective expectation model emphasizes roles of discrepancy between actual and expected service outcomes in satisfaction (Wilson and Klaaren, 1992). The results of this study confirmed both of the expectation theories. Emotional service expectation has direct, positive effects on service outcomes (H1) while its influences on overall confirmation differ by detections of high and low surface acting (H11).

Customers have typical expectancies about employees' emotional services, such as "service with a smile" (Parasuraman et al., 1985; Schminke et al., 2014). Burgoon and Walther (1990) claimed that expectancy is associated with predictions rather than desires. Consistent with their arguments, customers seem to predict authentic emotional services from employees; surface acting such as faking smiles appears not enough to enhance positive outcomes and even worse, has negative impacts on overall confirmation. Service managers are "wise to want workers to be sincere, to go well beyond the smile that's just painted on" (Hochschild, 1983, p. 33). Even though some researchers argued that there is no difference in customer reactions and service outcomes between displays of authentic and fake emotions (Beal et al., 2006; Chi et al., 2011; Wang and Groth, 2014), the study's results suggest that customer decisions can differ by customers' detections of employees' emotional displays, whether they are real or fake. Thus, service organizations need to intervene to help their employees adopt deep acting strategies to meet customer emotional service expectations. Possible training may include perspective taking and empathy training to help employees understand customer's perspectives and display genuine emotions. In summary, managerial emphasis should be on hiring those who can genuinely express their emotions and training them to understand customer's needs sincerely (Parker and Axtell, 2001).

# Limitation and Future Research

As is the case with any research, there are limitations that can be suggested for future research. First, this study explored the influence of customers' predictive expectations as an independent variable, so the other dimensions of expectations need to be investigated. Expectations can be classified into normative (i.e., what should be), predictive (i.e., what customers really expect), and equitable expectations (i.e., what customers should receive, taking into account the expenses borne) (Medrano et al., 2016). For the future research, it seems worthwhile comparing influences of different dimensions of customers' expectations. Second, this study used an online panel in which respondents were asked questions recalling their most recent service experiences. While the use of an online panel is suitable to collect a large sample and the use of experience and memory based measures of emotions allows researchers to investigate real service users, these methods are subject to several response biases, such as selection bias, non-response bias, recall bias, and memory bias as questionnaires are not distributed perfectly randomly and/or participants have difficulty in retrieving the detailed information of the past experience. Another limit of this study is that the questionnaires were collected in the United States, and there may be different findings across cultures and nations. Norms and interpretations about emotion-communication may be specific to cultures, so customer responses to employees' emotional displays are likely to vary across cultures (Triandis, 1989). For example, compared with individualist cultures, collectivist cultures tend to emphasize connectedness to other people, conformity, and social contribution (Ariely, 2008), so more severe emotional duties and obligations are likely to be implicitly imposed on employees in these cultures. In turn, customers in collectivist cultures may strongly predict employees' deep acting strategies, while being more sensitive to surface acting. Fourth, factors unrelated services, such as weather, personality and customer emotions per se, may influence customers' assessment of employees' emotion displays (Penz and Hogg, 2011; Pelegrín-Borondo et al., 2015). Therefore, more properly controlled experiments seem to be necessary for future research to clearly determine causality among the variables. Lastly, it is necessary to explore which categorical emotional displays are responsible for the degree of customers' detections of deep and surface action (Pelegrín-Borondo et al., 2015). Then, future research may give even greater insights into human resource management practically and academically in the hospitality industry.

# AUTHOR CONTRIBUTIONS

fpsyg-10-00321 March 8, 2019 Time: 17:21 # 10

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# REFERENCES

Allen, J. A., Pugh, S. D., Grandey, A. A., and Groth, M. (2010). Following display rules in good or bad faith? Customer orientation as a moderator of the display rule-emotional labor relationship. Hum. Perform. 23, 101–115. doi: 10.1021/ ja800916d

Ariely, D. (2008). Predictably Irrational. New York, NY: HarperCollins.


# FUNDING

This work was supported by the research fund of Hanyang University (HY-2018).


summary and theoretical extension of the work design literature. J. Appl. Psychol. 92:1332.


approach to the emotions produced to tourists by the hotel. J. Vacat. Mark. 21, 351–365. doi: 10.1177/1356766715580869


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Jeong, Park and Hyun. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Emotional Labor in Knowledge-Based Service Relationships: The Roles of Self-Monitoring and Display Rule Perceptions

### Shenghua Huang<sup>1</sup> , Hongbiao Yin<sup>1</sup> and Lifang Tang<sup>2</sup> \*

<sup>1</sup> Faculty of Education, The Chinese University of Hong Kong, Shatin, Hong Kong, <sup>2</sup> Faculty of Education, Northeast Normal University, Changchun, China

Focusing on knowledge-based service relationships, this study examined knowledgebased service workers' (i.e., school teachers) emotional labor process and the consequential outcomes for their well-being. The study also examined the roles of two antecedents, namely, teachers' perceptions of display rules and self-monitoring tendencies. A sample of 1,656 school teachers participated in the study. The results showed that self-monitoring generally had stronger, though maladaptive, effects than display rule perceptions on individuals' use of emotional labor strategies (ELS) (i.e., surface acting and deep acting) and well-being (i.e., anxiety, depression, contentment, and enthusiasm). Both self-monitoring and display rule perceptions were positively related to two ELS. There were relatively stronger relationships between self-monitoring and surface acting, and between display rule perceptions and deep acting. Surface acting was positively related to anxiety and depression and negatively related to contentment and enthusiasm. Deep acting was positively related to anxiety, contentment, and enthusiasm. The examination of indirect effects showed that selfmonitoring was positively related to anxiety and depression and negatively related to enthusiasm and contentment. Display rule perceptions were weakly, but positively, related to anxiety and depression. These results suggest that self-monitoring may be less beneficial than previously thought. Knowledge-based service workers' display rule perceptions and deep acting may not necessarily be harmful to their well-being, but reflect their role identification and commitment. Theoretical contributions and practical suggestions of this study were discussed.

Keywords: knowledge-based service, emotional labor, well-being, self-monitoring, display rule perceptions

# INTRODUCTION

Employees' emotional labor has emerged as an essential way to cultivate customer satisfaction and loyalty and to forge and maintain sound customer relationships (Totterdell and Holman, 2003; Côté, 2005; Grandey et al., 2005; Hülsheger et al., 2015). Emotional labor is the process by which employees display appropriate emotional expressions as required implicitly or explicitly

### Edited by:

Jungkun Park, Hanyang University, South Korea

### Reviewed by:

Sang-Lin Han, Hanyang University, South Korea Catherine S. Daus, Southern Illinois University Edwardsville, United States

> \*Correspondence: Lifang Tang tanglf571@nenu.edu.cn

### Specialty section:

This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology

Received: 11 September 2018 Accepted: 25 March 2019 Published: 09 April 2019

### Citation:

Huang S, Yin H and Tang L (2019) Emotional Labor in Knowledge-Based Service Relationships: The Roles of Self-Monitoring and Display Rule Perceptions. Front. Psychol. 10:801. doi: 10.3389/fpsyg.2019.00801

by their organizations (Hochschild, 1983; Grandey, 2000; Brotheridge and Grandey, 2002). Unlike emotional management in a general sense, emotional labor is performed in exchange for wages (Grandey, 2000). The importance of emotional labor in service companies is self-evident. A successful service delivery depends on the positive interactions or cooperation between customers and employees (Gummesson, 1991; Gremler et al., 2001; Groth et al., 2009). Customer satisfaction and loyalty are also largely influenced by the emotional aspects of these interactions (Gremler et al., 2001; Groth et al., 2009; Hülsheger et al., 2015). Generally, previous studies have shown that friendly and authentic emotional expressions are highly predictive of customers' positive reactions and satisfaction (Côté, 2005; Grandey et al., 2005; Hennig-Thurau et al., 2006; Groth et al., 2009). Nevertheless, the improvement of service quality through emotional labor is thought to be at the expense of employees' well-being (Hochschild, 1983; Holman et al., 2008; Hülsheger and Schewe, 2011; Scott et al., 2012). A large number of studies have investigated the effects of the emotional labor process on employees' mental health and well-being (Holman et al., 2008). These studies generally revealed some negative effects of emotional labor on service employees. For example, emotional display rules are believed to be an extra burden on employees (Hochschild, 1983; Brotheridge and Grandey, 2002). Emotional labor process depletes limited mental resources and leads to employees' emotional exhaustion and reduced job satisfaction (Grandey, 2000; Hülsheger and Schewe, 2011; Humphrey et al., 2015). Positive emotional expressions achieved through faking and hiding result in employees' sense of inauthenticity (Grandey, 2000; Humphrey et al., 2015).

Fortunately, the interests of the organization and the employees are not irreconcilable. Some researchers have suggested that emotional labor may not necessarily be harmful for employees. Emotional labor can be less stressful when employees recognize the importance of emotional labor and adopt appropriate strategies (Brotheridge and Grandey, 2002; Hülsheger and Schewe, 2011). The desired emotional expressions can be better achieved when organizations grant employees sufficient autonomy and provide them with proper instruction (Grandey, 2000). Emotional labor can even enhance employees' sense of accomplishment and job satisfaction, especially when there is a fit between the emotional requirements of the environment and employees' characteristics or personality (Hülsheger et al., 2013; Humphrey et al., 2015; Diefendorff et al., 2016). Despite the fact that more and more researchers tend to examine the complexity of emotional labor and its outcomes in light of different individual and organizational factors (Humphrey et al., 2015; Diefendorff et al., 2016), most studies were conducted among traditional service employees, such as bus captains, call center agents, and waiters/waitresses (e.g., Totterdell and Holman, 2003; Grandey et al., 2005; Goldberg and Grandey, 2007; Scott et al., 2012; Hülsheger et al., 2015). Few studies have investigated the emotional labor performed by highcommitment, knowledge-based service (KBS) workers, such as teachers, lawyers, and accountants.

There are three important aspects distinguishing KBS workers from traditional service employees. First, interactions between traditional service employees and customers exist in frequent, one-time, and short-term encounters, but those between KBS workers and customers involve some customized, stable, and long-term relationships (Gutek et al., 1999; Holman, 2003). Second, the emotional display rules of KBS workers are different from those of traditional service employees. As Mills and Moshavi (1999) suggested, for the high-commitment KBS workers, it is critical to exhibit both social affiliation and professional authority during their interactions with clients. These KBS workers generally need to maintain positive emotions and show friendliness in order to build relationships and elicit client cooperation. However, they still have some autonomy and can show strictness and authority when necessary (Mills and Moshavi, 1999; Holman, 2003). Third, KBS workers' motivation to perform emotional labor could be quite different from those of traditional service employees. The primary goal of their work is to provide information and suggestions using their expertise and knowledge, rather than to ensure their clients' comfort and relaxation. The emotional display rules for KBS workers thus are often implicit. KBS workers are less forced to perform emotional labor. The manner in which KBS workers regulate their emotions at work is seldom monitored by an immediate supervisor or colleagues. In sum, KBS workers have some but "limited" autonomy in expressing their true selves in the workplace. With different job requirements and work environments, different KBS workers (teachers, lawyers, and accountants) may interact with their clients (or students and parents) differently. Nevertheless, they still have to manage emotions and perform emotional labor at work.

The unique characteristics of KBS workers' emotional labor make some fundamental assumptions of emotional labor questionable. For example, do KBS workers perform emotional labor for the sake of organizations and mainly in exchange for wages (Grandey, 2000)? Are employees in "people work" faced with higher emotional demands with limited personal control over emotions (Brotheridge and Grandey, 2002)? If KBS workers have more autonomy over their emotions and are not forced to display certain emotions, why are they still motivated to manage emotions at work and keep sound relationships with their clients? Thus, it is important to explore whether it is the perceived, external display rules or the internal self-monitoring tendencies that better explain KBS workers' motivation in performing emotional labor. Display rule perceptions, on the one hand, denotes KBS workers' awareness of the often implicit emotional rules for their interactions with clients (Brotheridge and Grandey, 2002; Diefendorff et al., 2005; Diefendorff and Greguras, 2009). Self-monitoring, on the other hand, refers to their general tendency or willingness to monitor their own emotional expressions during personal interactions (Snyder, 1974; Snyder and Gangestad, 1986; Diefendorff et al., 2005; Bono and Vey, 2007). In other words, display rule perceptions reflect the external emotional demands and rules as perceived by KBS workers, while self-monitoring represents KBS workers' own internal monitoring tendency (Reeve et al., 2007). Previous studies on traditional service employees have examined the relationships between emotional labor and dispositional variables

(e.g., self-monitoring) or situational variables (e.g., display rule perceptions) (Diefendorff et al., 2005; Bono and Vey, 2007; Holman et al., 2008). However, mixed findings are reported. These inconsistencies may be due to factors such as role identification, demand-ability fit, and contextual difference (Hülsheger et al., 2013; Humphrey et al., 2015; Diefendorff et al., 2016). Considering the uniqueness of KBS workers' emotional labor, a closer look at the emotional labor of KBS workers can contribute to the literature. Through a comprehensive comparison between display rule perceptions and self-monitoring and their effects, we can obtain a deeper understanding of KBS workers' emotional labor and the consequential well-being outcomes.

Thus, this study aims to examine how KBS workers' display rule perceptions and self-monitoring are related to their different emotional labor strategies (ELS) and the consequential well/ill-being outcomes. Specifically, this study attempts to address three questions as follows: (a) whether and how are different ELS related to KBS workers' well-being outcomes? (b) whether and how are display rule perceptions related to KBS workers' ELS and the consequential well-being outcomes? (c) whether and how is self-monitoring related to KBS workers' ELS and the consequential well-being outcomes? In the following sections, we review relevant theories and studies, propose and test our hypotheses, and discuss the theoretical contributions and practical implications of the results. We end by outlining research limitations and proposing directions for future studies.

# LITERATURE REVIEW AND HYPOTHESES

# Knowledge-Based Service and Customer Relationships

Typically, KBS workers use their knowledge, ideas, and other intellectual capital to fulfill their work; they diagnose client priorities and recommend a course of action to their clients (Mills and Moshavi, 1999). According to Mills and Moshavi (1999), KBS workers must balance "the paradox of getting close to the client while remaining somewhat detached (p. 53)." In other words, the relationship between KBS workers and clients is both a vertical and a peer relationship. The vertical relationship secures client compliance. KBS workers need to establish their own social status and authority in the vertical relationship. They may thus carefully maintain a social and professional distance from their clients, for example, through holding objective or sometimes critical attitudes toward them (Mills and Moshavi, 1999; Gremler et al., 2001). The peer relationship, on the other hand, contributes to in-depth understanding and cooperation. That is, KBS workers need to foster psychological attachment and social affiliation with their clients (Mills and Moshavi, 1999; Gremler et al., 2001).

In a similar vein, Holman (2003) identified two types of service interaction: mass service and high commitment service. Mass service is characterized by standardized service; one-time, short-term encounters between customers and employees; and low added value (Gutek et al., 1999). High commitment service is, in contrast, characterized by customized service; long-term, relatively stable relationships between customers and employees; and high added value (Gutek et al., 1999; Holman, 2003). Workers in high commitment service industries (e.g., KBS workers) are generally better paid and have greater decisionmaking latitude at work. They also have more opportunities for professional development and promotion (Gutek et al., 1999; Holman, 2003).

In addition, there are no explicit emotional display rules in their organizations. As compared to traditional service workers, KBS workers are usually granted great autonomy and flexibility in maintaining both their social affiliation and professional authority (Mills and Moshavi, 1999; Holman, 2003; Diefendorff and Greguras, 2009). Professional affairs are of more relevance to their jobs. Service workers and clients constitute small and comparatively independent economic units. Their interactions are relatively private and exist both inside and outside regular organizational contexts (Gummesson, 1991; Mills and Moshavi, 1999). A direct external supervision over workers' emotion regulation and expressions becomes less necessary or possible.

The unique characteristics of KBS workers' emotional labor thus deserve more attention. The co-existence of formal and personal relationships between KBS workers and clients makes their emotional labor process more flexible (Gutek et al., 1999; Holman, 2003). KBS workers perform emotional labor less frequently but more intensely than employees in mass service industries (Brotheridge and Grandey, 2002). As for the emotional display rules, although positive interactions are generally desired, KBS workers sometimes are able to display neutral or even negative emotions to maintain social and professional distance from their clients (Mills and Moshavi, 1999). More importantly, KBS workers' motivation to engage in different ELS may be quite different. Rather than being instructed or forced to perform emotional labor, KBS workers may voluntarily choose to regulate their emotions (Gosserand and Diefendorff, 2005). Specifically, KBS workers may perform emotional labor because they tend to monitor their emotional expressions during social interactions (self-monitoring), or because they identify and choose to follow the implicit emotional display rules at work (display rule perceptions) (Diefendorff et al., 2005).

School teachers, who are the focus of this study, are typical KBS workers. The agent–client relationships between teachers and students or parents are implicit and indirect. There is no contract signed directly between teachers and students. Each teacher has to act for a number of students and parents, and each student has several teachers who help them make decisions about their learning. Nevertheless, there are "invisible" contracts between teachers and students. On the one hand, most teachers understand their responsibilities and try to act in the best interests of their students. On the other hand, most students and parents trust teachers and let teachers decide on the proper courses of their learning and school activities. Actually, the co-existence of peer and vertical relationships, which is the most important characteristic of Mills and Moshavi's (1999) KBS workers, is much more commonly observed among teacher-student relationships than other agent-client relationships.

As KBS workers, teachers use their content knowledge and teaching experience to teach students, help students to find best ways to learn, and cooperate with colleagues and parents to fulfill their work. There are relative stable and long-term relationships between the teacher and the students. A teacher not only need to care for and about the students, but also establish authority and maintain classroom rule and discipline. In addition, there are no slogans such as "the students/parents are always right" for a teacher. Different teachers may have totally different understanding of the emotional display rules at the school (Hargreaves, 2000; Newberry, 2010; Yin, 2016). For example, some teachers may believe that teaching should be professional and emotion-neutral; others may believe that positive emotional interactions between teachers and students contribute to student learning and performance; and still others may believe that caring and loving their students are the most important responsibilities on their part. These dynamics among teachers reflect the unique nature of KBS workers' emotional labor. A further investigation on the relationships among teachers' display rule perceptions, self-monitoring, emotional labor and the consequential well-being outcomes will contribute to an in-depth understanding of the complexity of KBS workers' emotional labor.

# Emotional Labor Strategies and Well/Ill-Being

Emotional labor means managing one's emotions to earn a wage (Hochschild, 1983). When performing emotional labor, employees and workers have to use appropriate strategies to regulate their inner feelings or external expressions. Two essential types of ELS are surface acting and deep acting. Surface acting involves hiding or faking emotions and leads to a consequent dissonance between expressed emotions and inner feelings. Deep acting refers to modifying or changing one's inner feelings and thus leads to the natural expressing of emotions required (Hochschild, 1983; Grandey, 2000; Hülsheger and Schewe, 2011). According to Grandey (2000). For example, when getting irritated by a student, a teacher can remain calm and positive through hiding his or her true emotions (i.e., surface acting). The teacher may also reappraise the situations and attribute students' misbehaviors to a lack of self-control or disadvantaged family backgrounds (i.e., deep acting), and thus feel more compassion and less anger. Additionally, rather than to regulate the emotions directly, a teacher may also choose to leave the classroom or suspend the on-going classroom activities to cope with these situations. The initiation of the emotion regulation process is closely related to situational cues, such as general interaction expectations, display rules, and real-time positive or negative affective events (Gosserand and Diefendorff, 2005; Goldberg and Grandey, 2007). The strategies adopted in the emotion regulation process and their effects vary across different individual characteristics (e.g., gender, affectivity, emotional self-monitoring, and emotional intelligence) and organizational factors (i.e., job autonomy, social supports) (Grandey, 2000; Totterdell and Holman, 2003; Hülsheger et al., 2013). The empirical results suggest that individuals with higher emotional capacities (e.g., emotional intelligence, emotional knowledge) are found to use more deep acting than surface acting (Brotheridge and Grandey, 2002; Yin et al., 2017). Perceived emotional job demands are found to increase the use of both types of ELS. Job autonomy and social support, on the other hand, can provide a buffer against the negative effects of emotional labor (Totterdell and Holman, 2003; Diefendorff et al., 2005; Holman et al., 2008).

In general, surface acting is more maladaptive than deep acting, though deep acting can also do harm to individuals' well-being (Côté, 2005; Holman et al., 2008; Hülsheger and Schewe, 2011). Previous theories and research generally suggest that surface acting is negatively related to individuals' well-being, because the process leads to mental resource depletion, a sense of inauthenticity, and negative feedback from customers (Grandey, 2000; Côté, 2005; Hülsheger and Schewe, 2011; Hülsheger et al., 2013; Humphrey et al., 2015). However, the relationship between deep acting and wellbeing indicators is less consistent: deep acting is thought to have a dual effect on individuals' well-being (Hülsheger and Schewe, 2011). Although deep acting depletes mental resources and leads to emotional exhaustion, it is also related to positive customer feedback and a sense of pride and accomplishment (Grandey, 2000; Hülsheger and Schewe, 2011; Humphrey et al., 2015). Consistently, empirical studies have found that surface acting is positively related to emotional exhaustion, depersonalization, and a lack of personal accomplishment. Surface acting is also found to be negatively related to job satisfaction and self-efficacy (Goldberg and Grandey, 2007; Hülsheger and Schewe, 2011; Yin, 2016). Deep acting is generally found to be positively related to emotional exhaustion. However, mixed findings were reported in terms of the relationships between deep acting and job satisfaction and personal accomplishment (Hülsheger and Schewe, 2011; Yin, 2016).

For teachers or KBS workers in general, the emotional labor process is more flexible. However, there should be similar underlying mechanisms, through which ELS influence their well/ill-being outcomes. That is, when surface acting is adopted, a teacher has to spend sustained effort and mental resources (Brotheridge and Grandey, 2002). The feelings of fatigue and inauthenticity can be reflected in their frequent experience of uneasiness and depression as well as a lack of enthusiasm or calmness during work (Grandey et al., 2005). When deep acting is adopted, the mental resources depletion process also leads to the unpleasant experience of anxiety or depression (Hülsheger and Schewe, 2011). However, the faithful effort of teachers may also contribute to more positive attitudes and feelings toward their work (e.g., contentment, enthusiasm) (Brotheridge and Grandey, 2002; Totterdell and Holman, 2003; Yin et al., 2017). Given this dual process, it is necessary to explore the effects of ELS on ill-being and well-being indicators, respectively. In other words, ill-being and well-being should be treated as two distinct criteria rather than as two opposite ends of the continuum of well-being. Thus, based on Warr's (1990) conception of the four dimensions of well-being, which comprise

two well-being indicators (contentment, enthusiasm) and two illbeing indicators (anxiety, depression), the following relationships are hypothesized:

H1a: Surface acting is positively related to anxiety and depression, and negatively related to contentment and enthusiasm.

H1b: Deep acting is positively related to anxiety and depression, and positively related to contentment and enthusiasm.

# Display Rule Perceptions

Emotional display rules specify the appropriate kinds of emotional expressions and the ways emotions should be displayed (Brotheridge and Grandey, 2002; Diefendorff et al., 2005; Goldberg and Grandey, 2007). A widely adopted classification of display rule perceptions distinguishes between positive display rule perceptions (i.e., rules for expressing positive emotions) and negative display rule perceptions (i.e., rules for suppressing negative emotions) (Brotheridge and Grandey, 2002; Diefendorff et al., 2005). Although it is theoretically plausible that being required to "express positive emotions" is more palatable than to "hide negative emotions," many researchers have failed to differentiate the two dimensions of positive and negative perceptions (Gosserand and Diefendorff, 2005; Goldberg and Grandey, 2007). In fact, displaying positive emotions inevitably means not displaying negative ones, and vice versa. Therefore, it is simplest to classify display rules into three types: to express positive emotions (and thus hide negative ones), to express negative emotions (and thus hide positive ones), and to remain neutral (Côté, 2005). In the present study, we thus treat display rule perceptions as a single-dimension variable, which integrates expressing positive emotions and hiding negative ones (Gosserand and Diefendorff, 2005).

Since emotional display rules may be set explicitly or implicitly by organizations, the subjective perceptions of display rules can vary across individuals, especially when it comes to implicit rules (Diefendorff et al., 2005; Diefendorff and Greguras, 2009). Allen et al. (2014) suggested that people in collectivistic cultures are more concerned with social harmony and group cooperation, and thus perceive a greater number of emotional display rules and report more frequent use of ELS. For KBS workers, individualized perceptions of display rules are more common. As discussed above, the emotional display rules for KBS workers are generally implicit. Teachers are generally required to express positive emotions toward students and parents, although some negative emotions of teachers are still acceptable at school (Hargreaves, 2000). However, teachers may have different understandings of their expected roles and the appropriate emotional expressions at work. The lack of direct external supervision also makes it difficult to ensure each teacher has the same perception of and responses toward these emotional display rules.

As stated earlier, situational cues, such as emotional job demands, display rules, and real-time positive or negative affective events, may make emotional labor necessary (Grandey, 2000; Totterdell and Holman, 2003). Empirically,

Brotheridge and Grandey (2002) found that both positive and negative display rule perceptions were positively related to emotional labor. Previous findings also indicated that the use of both types of emotional labor (i.e., surface acting and deep acting) increased in accordance with employees' awareness of emotional display rules/demands (Gosserand and Diefendorff, 2005; Goldberg and Grandey, 2007). For teachers, it is likely that highly perceived display rules will lead teachers to performing more emotional labor. Thus, consistent with previous studies, we suppose that display rule perceptions will be positively related to both surface acting and deep acting, depending on teachers' predispositions (Brotheridge and Grandey, 2002; Diefendorff et al., 2005). However, it should be noted that, considering the implicit display rules and limited external supervision, teachers are not forced to hide or fake their emotions even with high display rule perceptions. Thus, the high display rule perceptions of teachers may tend to encourage them to use more deep-level solutions, rather than force themselves to hide or fake emotions. In other words, the positive relationship between display rule perceptions and deep acting should be stronger than that between display rule perceptions and surface acting. We thus hypothesize as follows:

H2a: Display rule perceptions are positively related to both surface acting and deep acting; there is a stronger relationship between display rule perceptions and deep acting.

Additionally, since surface acting and deep acting are predictive of teacher well/ill-being, there should be indirect effects of display rule perceptions on teachers' well/ill-being outcomes via ELS. Researchers suggested that the emotional labor process (especially surface acting) initiated by display rules depletes attentional resources and energy (Goldberg and Grandey, 2007). Display rules and an inefficient emotional labor process may induce emotion-rule dissonance. That form of person-role conflict is further predictive of emotional exhaustion and job dissatisfaction (Holman et al., 2008; Hülsheger and Schewe, 2011). However, if display rule perceptions lead to the use of more adaptive ELS (i.e., deep acting in the case of teachers), these unfavorable effects should be less severe. Combining previous assumptions, we thus hypothesize that the perceived display rules and extra effort in emotional labor are stressful and may lead to anxiety and depression. However, teachers' faithful effort enhanced by a high display rule perceptions may balance the negative effect of surface acting, and thus generally contribute to more positive attitudes and feelings (e.g., contentment, enthusiasm). The following relationships are hypothesized:

H2b: Surface acting and deep acting mediate the relationship between display rule perceptions and well/ill-being indicators.

H2c: The total indirect effects of display rule perceptions on anxiety and depression are positive, while those on contentment and enthusiasm are also positive.

# Self-Monitoring

According to Snyder (1974), self-monitoring of expressive behavior refers to individuals' tendency to control and regulate

their expression according to situational cues. Low self-monitors are more likely to be themselves and exhibit consistency between inner feelings and external expressions (Day et al., 2002; Scott et al., 2012). By contrast, high self-monitors tend to hide their inner feelings and attune their facial expressions, voice, and posture to situational cues (Bono and Vey, 2007; Scott et al., 2012). Moreover, an individual's tendency to control and regulate his or her expressions may vary across day-to-day situations. For example, the need for self-monitoring decreases if individuals are interacting with close friends/relatives under familiar circumstances (Hülsheger et al., 2013; Huang et al., 2018).

Self-monitoring has long been treated as a desirable quality and associated with better personal performance, such as desirable public appearance, better impression management, harmonious relationships, and leadership (Snyder and Gangestad, 1986; Day et al., 2002).

However, the relationships between self-monitoring and ELS and the consequential well-being outcomes remain controversial. Some studies have suggested that self-monitoring is negatively related to individuals' well-being. In a meta-analytic review of self-monitoring, negative relationships were found between selfmonitoring and organizational commitment and job satisfaction, while positive relationships were found between self-monitoring and role conflict and role ambiguity (Day et al., 2002). In a similar vein, Baumeister et al.'s (1998) research on ego depletion also suggested negative effects of self-monitoring. According to Baumeister et al. (1998), self-regulation and other forms of volition lead to expenditure of limited resources, which in turn contributes to burnout and ill-health. In this respect, selfmonitoring is thought to be related to maladaptive coping and ill-being (Diefendorff et al., 2005).

Another group of studies, however, believed that high selfmonitors were more skilled and capable of performing emotional labor than low self-monitors. For example, high self-monitors are thought to have acute awareness of situational cues and show more willingness and efficiency in regulating emotional expressions (Bono and Vey, 2007; Scott et al., 2012). Some studies have also suggested a buffering effect of self-monitoring on the maladaptive relationship between surface acting and job satisfaction/work withdrawal (Scott et al., 2012). This second line of studies seems to conceptualize self-monitoring as inherently related to better emotion regulation skills and capacities, rather than as a willingness or tendency to monitor emotional expressions. Nevertheless, based on a social interaction model, Côté (2005) suggested that emotion regulation can increase work strain when receivers (i.e., customers) give negative responses, or decrease work strain if they give positive responses. Thus, assuming high self-monitors are more sensitive to situational cues (customer responses in this case), both the positive and the negative relationships between emotion regulation and work strain shall be stronger for high self-monitors (Côté, 2005).

Unlike display rule perceptions which reflect the external emotional demands and rules as perceived by KBS workers, KBS workers' self-monitoring reflects an individual tendency in terms of emotional management (Reeve et al., 2007). Thus, it is not expected that KBS workers' self-monitoring would have any context-specific characteristics different from those of traditional service employees. In addition, compared to display rule perceptions, KBS workers' internalized, self-determined monitoring of their emotional expressions is more predictive of their actions and behaviors (Reeve et al., 2007). The evidence regarding the negative effects of self-monitoring on individual well-being reveals that self-monitoring may not be as adaptive as some researchers suggest. Rather, as a tendency to regulate expressive behavior or as a dimension of self-regulation, selfmonitoring and the consequent emotional labor process deplete individuals' finite pool of resources, induce role conflict, and are related to work strain and reduced well-being (Baumeister et al., 1998; Brotheridge and Grandey, 2002; Day et al., 2002; Holman et al., 2008). In other words, high self-monitoring may be more predictive of emotional labor frequency but not effectiveness. While a high level of self-monitoring is predictive of both types of ELS, the relative strength of these relationships may depend on other factors, such as the availability of ego resources or individuals' emotional capacities (e.g., emotional intelligence, emotional knowledge) (Baumeister et al., 1998; Grandey, 2000). Thus, the following relationships are hypothesized:

H3a: Self-monitoring is positively related to both surface acting and deep acting.

H3b: Surface acting and deep acting mediate the relationship between self-monitoring and well/ill-being indicators.

H3c: The total indirect effects of self-monitoring on anxiety and depression are positive, while those on contentment and enthusiasm are negative.

The hypothesized model is presented in **Figure 1**.

# MATERIALS AND METHODS

# Procedures and Participants

Consistent with institutional review board procedures, this study was carried out in accordance with the recommendations of Survey and Behavioral Research Ethics Committee at the Chinese University of Hong Kong with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Survey and Behavioral Research Ethics Committee.

The participants in this study were a sample of school teachers in Hong Kong. We connected with school principals and invited teachers to participate in the survey. Teachers had the full autonomy to decide whether to participate in the survey or not. All completed questionnaires were mailed by the participants directly to the authors. Anonymity was guaranteed. The final sample contained a total of 1,656 teachers from 38 primary schools (n1 = 1,115) and 16 secondary schools (n2 = 541). The valid response rate was 41.4%. The sample comprised 465 male participants (28.1%), 1,167 female participants (70.5 %), and 24 participants who did not report their gender (1.4%). In addition, 68 participants (4.1%) reported that they were principals in their schools, 630 (38.1%) were subject panel heads,

889 (53.7%) were regular teachers, and 69 (4.2%) did not specify their positions. In terms of their educational backgrounds, most reported a bachelor's (680 participants, 41.1%) or master's degree (928 participants, 56.0%). **Table 1** summarizes these demographic factors in detail.

# Instruments

A total of 34 items were used to operationalize the eight latent variables. Participants rated each item on a five-point Likert scale from "1 = strongly disagree (or never)" to "5 = strongly agree (or


always)" according to the extent to which they agreed with the statements or the frequency of certain experiences. Translation and back-translation procedures were strictly followed to convert the English scales into Chinese versions.

# Display Rule Perceptions

Teachers' perceptions of display rules were measured using a five-item scale adapted from Diefendorff et al. (2005). While the original inventory measures two dimensions of display rule perceptions (i.e., positive and negative display rule perceptions), our data did not support a two-dimension structure. The final scale used in this study measures an integrated perception of the extent to which participants were required to display positive emotions while hiding negative ones (Gosserand and Diefendorff, 2005; Goldberg and Grandey, 2007). Sample items include "I am expected to suppress my bad moods or negative reactions to students" and "My school expects me to try to act excited and enthusiastic in my interactions with students." The Cronbach's α coefficient of the scale obtained in this study was 0.79.

# Self-Monitoring

A short version of Snyder and Gangestad's (1986) Selfmonitoring Scale was used to measure participants' selfmonitoring tendencies. Consistent with the practice of previous studies, the short version only includes seven non-reversely coded items from the original version (Scott et al., 2012). The participants were asked to report whether they were flexible and often attuned their public appearances to situational cues. Sample items include "In different situations and with different people, I often act like very different persons" and "I'm not always the person I appear to be." The Cronbach's α coefficient of self-monitoring obtained in this study was 0.72.

# Emotional Labor Strategies

Surface acting and deep acting were measured using the ELS scale. The scale was adapted and validated by Diefendorff et al. (2005) basing on the previous works of Kruml and Geddes (2000) and Grandey (2003). Surface acting was measured by six items including "I fake the emotions I show when dealing with students or their parents," and deep acting was measured by four items including "I try to actually experience the emotions that I must show to students or their parents." The Cronbach's α coefficients of surface acting and deep acting obtained in this study were 0.90 and 0.74, respectively.

### Well/Ill-Being Indicators

fpsyg-10-00801 April 5, 2019 Time: 16:43 # 8

Warr's (1990) 12-item scale consisting of both ill-being and wellbeing indicators was used to measure the consequential criteria. The participants were asked to report the frequency with which they experienced the following four sets of feelings over the previous few weeks: tension, uneasiness, and worry (anxiety); depression, gloom, and misery (depression); calm, contentment, and relaxation (contentment); and cheerfulness, enthusiasm, and optimism (enthusiasm). The Cronbach's α coefficients of anxiety, depression, contentment, and enthusiasm obtained in this study were 0.91, 0.89, 0.81, and 0.88, respectively.

# Analyses

Two main statistical procedures, confirmatory factor analysis (CFA) and structural equation modeling (SEM), were conducted using Mplus 7. SEM was used to test the hypothesized model because it allows a set of relationships between multiple independent and dependent variables. Simultaneously, the latent construct also adjusts for any measurement error in both dependent and independent variables (Schreiber et al., 2006). Basic statistical analyses were conducted using SPSS 20. Chi-square values (χ 2 ), degree of freedom (df), and the corresponding significant values (p), in addition to other model fit information such as the root mean square error of approximation (RMSEA), the standardized root mean square residual (SRMR), the comparative fit index (CFI), and the Tucker-Lewis Index (TLI), were reported. Model fit is excellent when the CFI and TLI are greater than 0.95 and acceptable when the NNFI and CFI are no less than 0.90. In addition, RMSEA and SRMR must be less than 0.06 and 0.08 for an excellent model fit, and 0.08 and 0.10 for an acceptable model fit (Schreiber et al., 2006). To test the indirect effects, bootstrapping technique (bootstrap samples = 2,000) was adopted, and the 95% confidence intervals (95% CIs) were calculated (Preacher and Hayes, 2008).

# RESULTS

# Reliability, Descriptive Statistics, and Correlations

**Table 2** reports the latent correlations among variables investigated in the present study and their respective mean values (ranging from 2.59 to 3.53), standard deviations (ranging from 0.55 to 0.95), and Cronbach's α coefficients (ranging from 0.72 to 0.90). In average, the teachers reported highest values for display rule perceptions (M = 3.53) and the lowest values for their experiences of depression (M = 2.59). The Cronbach's α coefficients were all greater than the generally accepted criterion (α > 0.70). In addition, the factor loadings of each item were all significant, ranging from 0.32 to 0.91. These latent variables had small to medium correlations with each other, except for the high correlations between anxiety and depression (r = 0.86) and between contentment and enthusiasm (r = 0.91). No significant correlations were found between deep acting and either contentment or enthusiasm, nor between self-monitoring and enthusiasm.

An eight-factor model was tested using CFA. The eight factors were display rule perceptions, self-monitoring, surface acting, deep acting, enthusiasm, contentment, anxiety, and depression, respectively. Each item was loaded on its corresponding factors. Factors were correlated with each other. The results of CFA generally confirmed the construct validity of the measurement with an acceptable model fit (χ <sup>2</sup> = 2978.35, df = 499, p = 0.00, RMSEA = 0.055, SRMR = 0.057, CFI = 0.92, TLI = 0.90).

# SEM Results

The results of SEM are presented in **Figure 2**. The results indicate an acceptable model fit of the hypothesized model: χ <sup>2</sup> = 3079.59, df = 508, p = 0.00, RMSEA = 0.055, SRMR = 0.063, CFI = 0.91, TLI = 0.90. As shown in **Figure 2**, all four well/illbeing indicators had stronger relationships with surface acting than with deep acting. Surface acting was positively associated with anxiety (β = 0.38, p < 0.01) and depression (β = 0.40, p < 0.01), and negatively associated with contentment (β = −0.42, p < 0.01) and enthusiasm (β = −0.36, p < 0.01) (H1a supported). Deep acting was positively associated with anxiety (β = 0.11, p < 0.01), contentment (β = 0.10, p < 0.01), and enthusiasm (β = 0.10, p < 0.01), although no significant relationship was found between deep acting and depression (β = 0.06, n.s.). Thus, H1b was generally supported, except for the nonsignificant relationship between deep acting and depression. Furthermore, display rule perceptions and self-monitoring were both positively related to the two types of ELS (H2a and H3a supported), with self-monitoring having a stronger relationship with both types of strategy. In addition, the relationship between display rule perceptions and deep acting (β = 0.30, p < 0.01) was stronger than that between display rule perceptions and surface acting (β = 0.13, p < 0.01), while the relationship between self-monitoring and deep acting (β = 0.28, p < 0.01) was weaker than that between self-monitoring and surface acting (β = 0.56, p < 0.01).

# Indirect Effects

**Table 3** shows the indirect effects of display rule perceptions and self-monitoring on the four well/ill-being indicators via two types of ELS. A 95% CI including zero indicates a non-significant indirect effect, and a 95% CI not including zero indicates a significant indirect effect. In general, the results of the indirect effects via ELS indicated that deep and surface acting mediated the relationships between display rule perceptions or self-monitoring and well/ill-being indicators, generally supporting H2b and H3b, although there were no significant indirect effects on depression via deep acting.


∗∗p < 0.01; SD, standard deviation. Goodness-of-fit index: χ <sup>2</sup> = 2978.35, df = 499, p = 0.00, RMSEA = 0.055, SRMR = 0.057, CFI = 0.92, TLI = 0.90.

In further investigating the total indirect effects of display rule perceptions and self-monitoring, it was found that selfmonitoring had stronger total indirect effects compared to display rule perceptions on the four well/ill-being indicators. There were significant and positive, though weak, total indirect effects of display rule perceptions on anxiety (Estimate = 0.08, 95% CI = [0.05,0.12]) and depression (Estimate = 0.07, 95% CI = [0.03,0.10]). No significant indirect effects were found between display rule perceptions and contentment (Estimate = −0.02, 95% CI = [−0.06,0.01]) or enthusiasm (Estimate = −0.02, 95% CI = [−0.05,0.01]). In other words, the negative effects on contentment and enthusiasm via surface acting and positive effects via deep acting balanced each other out, and resulted in non-significant indirect effects between display rule perception and contentment or enthusiasm. Thus, a positive total indirect relationship between display rule perception and ill-being was supported, but a positive indirect relationship between display rule perception and well-being was rejected (H2c).

In contrast, both the total positive indirect effect of selfmonitoring on anxiety (Estimate = 0.25, 95% CI = [0.18,0.31]) and depression (Estimate = 0.24, 95% CI = [0.17,0.31]) and its total negative indirect effect on contentment (Estimate = −0.21, 95% CI = [−0.28,−0.14]) and enthusiasm (Estimate = −0.17, 95% CI = [−0.23,−0.11]) were significant, which supported H3c.

**Table 4** summarizes the results of the hypothesis tests.

# DISCUSSION

Focusing on the emotional labor of KBS workers, this study has summarized several unique characteristics of KBS relationships and identified two essential variables that impact KBS workers' strategy choices and consequential well/illbeing indicators. The results show that the two essential variables had similar impacting patterns, though with different strengths. In general, self-monitoring was more predictive than display rule perceptions for both surface acting and well/illbeing indicators.

TABLE 3 | The estimates of indirect effects and the 95% confidence intervals.


CIs, Confidence Intervals; DRP, Display rule perceptions; SM, Self-monitoring; SA, surface acting; DA, deep acting.

# Theoretical Contributions

This study contributes to the literature of emotional labor by its focus on KBS workers and the unique characteristics of their emotional labor. First, by using criteria for both ill-being (anxiety, depression) and well-being (enthusiasm, contentment), this study investigated the effects of both ELS in a more comprehensive manner. Consistent with previous findings (Hülsheger and Schewe, 2011; Humphrey et al., 2015), surface acting was maladaptive to KBS workers' well-being and was positively associated with anxiety and depression, and negatively associated with enthusiasm and contentment. The effects of deep acting were much more beneficial, in that significant and positive relationships were found between deep acting and both contentment and enthusiasm. Furthermore, the positive relationship between deep acting and anxiety and the nonsignificant one between deep acting and depression indicated that a moderate level of anxiety related to deep acting is evidence of dedication and engagement, rather than a proxy for psychological dysfunction (Humphrey et al., 2015).

Second, the finding that self-monitoring had stronger effects on participants' use of ELS than display rule perceptions is consistent with previous theories. It seems that internalized values and rules are typically more predictive of individuals' TABLE 4 | A summary of the hypotheses and results.


DRP, Display rule perceptions; SM, Self-monitoring; SA, surface acting; DA, deep acting; A, Anxiety; D, Depression; C, Contentment; E, Enthusiasm.

practices and behaviors than external demands or rules (Reeve et al., 2007). In other words, display rule perceptions are a more distal predictor and self-monitoring a more proximal predictor of individual behaviors regarding emotion regulation and management. In addition, there were stronger relationships between self-monitoring and surface acting and between display rule perceptions and deep acting. As mentioned earlier, previous studies have presented contradictory hypotheses and results concerning the effects of self-monitoring, and many have suggested that self-monitoring is inherently related to more efficient emotion regulation and adaptive strategies, such as deep acting (Bono and Vey, 2007; Scott et al., 2012). However, the present results indicate that for KBS workers, self-monitoring may not be as beneficial as previously thought. One plausible explanation is that self-monitoring only evokes or enhances the awareness of situational cues and the tendency to perform emotional labor, but does not guarantee coping efficiency or effectiveness. In other words, the effects of self-monitoring on the process of emotional labor may not be as self-evidently beneficial or detrimental as other relevant individual characteristics, such as emotional intelligence (the ability to perceive and regulate emotions efficiently) or positive and negative affect (the baseline or general affect/moods of individuals, which increases or

decreases the level of emotion-rule dissonance) (Grandey, 2000; Totterdell and Holman, 2003; Hülsheger et al., 2013). Researchers have also proposed that high self-monitors can comfortably use surface acting without damaging the self, that low self-monitors may choose deep acting to follow the display rules and at the same time be themselves, and that high self-monitors are more fit for short-term service encounters and low self-monitors for long-term service relationships (Day et al., 2002; Holman, 2003; Scott et al., 2012). Somewhat consistent with these proposals, the present results suggest a stronger relationship between KBS workers' self-monitoring and surface acting and a maladaptive effect of self-monitoring on their well-being.

Third, although display rule perceptions were not positively related to enthusiasm and contentment, the overall effects of display rule perceptions seem to be beneficial after controlling for the effects of self-monitoring. The results show that display rule perceptions have a stronger relationship with deep acting than with surface acting. Display rule perceptions were only weakly though positively related to anxiety and depression, while self-monitoring was positively related to anxiety and depression, and negatively to enthusiasm and contentment. These findings suggest that display rule perceptions have a more beneficial effect than self-monitoring. This suggestion is plausible, given that a moderate level of work strain is acceptable or even beneficial because it suggests not ill-being or dysfunction but rather engagement and dedication (Humphrey et al., 2015). Concepts such as person-job fit, trait-behavior congruence, and role identification have long been topics of importance in the field of emotional labor (Ashforth and Mael, 1998; Gosserand and Diefendorff, 2005; Bono and Vey, 2007; Humphrey et al., 2015). Researchers have also suggested that only when employees are able to challenge, but finally accept, their job roles and rules can they fully identify with and commit to their professions (Ashforth and Mael, 1998; Humphrey et al., 2015). A higher level of display rule perceptions, thus, may indicate that KBS workers identify with the often implicit emotional aspects of their work and faithfully choose the deep-acting strategy to fulfill these extra requirements. Nevertheless, no positive indirect relationship was found between display rule perception and contentment or enthusiasm. It is because the negative indirect effects on contentment and enthusiasm via surface acting and the positive indirect effects via deep acting balanced each other out. Thus, to further enhance the beneficial roles of display rule perceptions, it is desirable to strengthen the positive relationship between display rule perceptions and deep acting while weaken that between display rule perceptions and surface acting. This could be achieved by several ways. Specifically, it is necessary to promote KBS workers' understanding of and commitment to the emotional aspects of their works. Extra effort is also need to familiarize them with different ELS, especially the deeplevel solutions.

Finally, consistent with some previous studies, this study did not validate a two-dimension structure of positive and negative display rule perceptions, but adopted a single-dimension measure of display rule perceptions that combines displaying positive emotions and hiding negative ones (Gosserand and Diefendorff, 2005; Goldberg and Grandey, 2007). It is meaningful and feasible to distinguish between positive and negative emotions in research of emotional labor. For example, the Discrete Emotions Emotional Labor Scale (DEELS) developed by Glomb and Tews (2004) distinguishes between genuine, faked, and suppressed positive and negative emotional displays. The DEELS has been widely adopted and is proved to have good psychometrics qualities (Mahoney et al., 2011). Thus, while it is also theoretically important to distinguish between positive and negative rule perceptions, and to find their different implications for individuals' strain and well-being, more effort is needed to develop a valid and reliable scale that links the theoretical conceptualizations and hypotheses with the empirical results.

# Practical Implications

The results of this study bring some implications for school practitioners and KBS providers at large. First, while teachers and other KBS workers should always be granted the job autonomy and flexibility they deserve, it is also important to raise their awareness about the emotional requirements and display rules of their jobs. The combination of autonomy and requirements may facilitate the identifying process, and thus enhance KBS workers' commitment. That is, when fully identified with the emotional aspects of their professions, KBS workers may tend to perceive the emotional requirements as challenge rather than as threat or hindrance (Lazarus and Folkman, 1984; LePine et al., 2005). The faithful effort and the use of deep acting may still be resource-depleting, but, at the same time, rewarding and inspiring.

Second, for service employees in general, a major cause of their anxiety and work strain comes from limited job autonomy and strict display rules. The support received from managers and colleagues could be helpful. Managers and organizations have to balance between the interests of customers and employees. Sometimes, it is necessary for a manager to punish an employee to satisfy an angry customer. However, it is also important for him or her to show emotional understanding and to comfort and support the employee. In the educational contexts, school principals and other administrative staffs should assume their responsibilities to create a supportive and trustworthy school environment. Furthermore, to enhance parent involvement and promote the understanding and cooperation between teachers and parents may also help in reducing teachers' anxiety and work strain.

Third, the results showed that high self-monitors may still tend to use surface acting strategies such as faking and hiding. This is probably caused by a limited knowledge of emotional labor and relevant strategies. Traditionally, the selection and training of KBS workers may focus on their professional competences. Limited support is provided to help them understand the emotional aspects of their jobs and use more adaptive strategies such as deep acting. Thus, programs aimed to improve employee mindfulness and emotional intelligence will be helpful for both teachers and KBS workers in general. Additionally, emotionally supportive environments could be useful. Platforms that allow for emotional communication and understanding may help KBS workers to recover from both relationship-related and profession-related strain at work.

# Limitations and Directions for Future Research

fpsyg-10-00801 April 5, 2019 Time: 16:43 # 12

A major limitation of this study lies in its cross-sectional design and subjective data sources. While it is common to use selfreported data to measure personal perceptions and behavioral tendencies, sole reliance on self-reported data is vulnerable to social desirability bias and constrains the credibility of the results. Moreover, the use of cross-sectional design cannot exclude the possibility of reverse causality. Although it is theoretically convincing for display rule perceptions and self-monitoring to influence emotional labor and the consequential well-being of the participants (Grandey, 2000; Totterdell and Holman, 2003), it is possible that well-being may influence intrapersonal-level variations in self-monitoring tendencies or strategy adoption (Hülsheger et al., 2013). Further studies with longitudinal designs and multiple data sources can advance this line of research.

While this study has focused on KBS workers and used a sample of teachers, future studies are needed to confirm these relationships among other KBS occupations. With a particular focus on the effects of display rule perceptions, researchers may also compare the results of KBS workers and those of mass service employees. In addition, demographic factors (e.g., gender, age, tenure, education), characteristics of organizations (e.g., grade levels; private vs. public companies/schools; commercial vs. consumer-oriented law firms), and the cultural environments will definitely influence the ways in which KBS workers interact with their clients and perform their emotional labor. For example, professors at universities and teachers from primary and secondary schools play different roles in facilitating student learning. The interactions between professors and students are less frequent but any impropriate conducts of professors may lead to negative evaluations from students, which further influence their general evaluations and career developments. By contrast, school teachers, especially primary school teachers, may express their emotions and attitudes more frequently and directly (Hargreaves, 2000). The intensive emotional interactions between teachers and students also require more effort on the part of teachers. It is meaningful and important to test the potential moderating effects and compare group differences in future studies. It also would be worthwhile to test other criteria such as employee commitment or role identification and customer satisfaction or feedback (Hennig-Thurau et al., 2006; Groth et al., 2009).

# REFERENCES


# CONCLUSION

Based on the unique characteristics of KBS relationships, this study has investigated how self-monitoring and display rule perceptions are related to KBS workers' ELS and the consequential well-being indicators of anxiety, depression, enthusiasm, and contentment. For KBS workers, who are generally less aware of the emotional aspects of their jobs and proper coping strategies, self-monitoring seems to be more related to a reactive emotional labor strategy (surface acting) than a proactive one (deep acting). By contrast, a higher level of display rule perceptions may represent role identification and commitment, and thus is related to a dedicated use of deep acting. Despite some limitations, this study illustrates characteristic aspects of emotional labor in KBS relationships and differentiates the roles of self-monitoring and display rule perceptions.

# ETHICS STATEMENT

Consistent with institutional review board procedures, this study was carried out in accordance with the recommendations of Survey and Behavioral Research Ethics Committee at the Chinese University of Hong Kong with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Survey and Behavioral Research Ethics Committee.

# AUTHOR CONTRIBUTIONS

SH analyzed the data and wrote the first version of the manuscript. HY designed the research and finished the final version of the manuscript. LT collected the data and helped with the data analysis.

# FUNDING

This work was supported by the General Research Fund of Hong Kong SAR (Grant No. CUHK14413314).



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Huang, Yin and Tang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.