On the Costs and Benefits of Emotional Labor: A Meta-Analysis of Three Decades of Research Ute R. Hu ��lsheger Maastricht University Anna F. Schewe Bielefeld University This article provides a quantitative review of the link of emotional labor (emotion���rule disso- nance, surface acting, and deep acting) with well-being and performance outcomes. The meta- analysis is based on 494 individual correlations drawn from a final sample of 95 independent studies. Results revealed substantial relationships of emotion���rule dissonance and surface acting with indicators of impaired well-being ( s between .39 and .48) and job attitudes ( s between .24 and .40) and a small negative relationship with performance outcomes ( s between .20 and .05). Overall, deep acting displayed weak relationships with indicators of impaired well-being and job attitudes but positive relationships with emotional performance and customer satisfaction ( s .18 and .37). A meta-analytic regression analysis provides information on the unique contribution of emotion���rule dissonance, surface acting, and deep acting in statistically predicting well-being and performance outcomes. Furthermore, a mediation analysis confirms theoretical models of emotional labor which suggest that surface acting partially mediates the relationship of emotion���rule dissonance with well-being. Implications for future research as well as pragmatic ramifications for organizational practices are discussed in conclusion. Keywords: meta-analysis, emotional labor, emotion regulation, well-being, job attitudes, performance Supplemental materials: http://dx.doi.org/10.1037/a0022876.supp Service plays a vital role in today���s economies. Indeed, service activities now account for about 70% of the gross domestic product in the United States as well as in European countries (Central Intelligence Agency, 2009). Accordingly, more than two thirds of the labor force in the United States and Europe is working in the service sector (Central Intelligence Agency, 2009), and this proportion is expected to grow even further in the years to come (Paoli & Merllie ��, 2008). As part of their daily work these employees have to interact with others, be it custom- ers, patients, students, or children. During these in- teractions they have to perform emotional labor, pub- licly displaying certain emotions while hiding others (Co ��te ��, 2005 Hochschild, 1983). The management of emotions has become part of organizational rules and occupational norms because organizational decision makers as well as employees believe that the expres- sion as well as suppression of certain emotions helps influence customers and clients to meet higher-order performance goals (Holman, Martinez-In ��igo, Totter- dell, 2008a Rafaeli & Sutton, 1987). Emotional la- bor has consequently become part of many individ- uals��� daily work despite the potential detrimental effects for employees��� psychological health (Hoch- schild, 1983). Starting with the seminal work by Arlie Russell Hochschild (1983) and fueled by the developments in the labor market, research into emotional labor has been burgeoning in the last three decades (Fisher & Ashkanasy, 2000 Zapf, 2002). However, despite the growth of scholarly work on emotional labor, a num- ber of important questions remain to be answered (e.g., Bono & Vey, 2005 Fisher & Ashkanasy, 2000 Rubin, Staebler Tardino, Daus, & Munz, 2005). Nu- merous researchers have investigated the link of emotional labor with well-being and different kinds of performance outcomes (e.g., task performance, affective delivery), both theoretically as well as em- pirically. Yet findings are inconsistent regarding the Ute R. Hu ��lsheger, Faculty of Psychology and Neuro- science, Maastricht University, Maastricht, the Netherlands Anna F. Schewe, Department of Psychology, Bielefeld Uni- versity, Bielefeld, Germany. Both authors contributed equally to this study. We thank all researchers providing information about unpublished studies or study details. Furthermore, we wish to express our sincere thanks to Gu ��nter W. Maier and Joyce E. Bono for helpful comments on an earlier version of this paper and Maria Hills for her stylistic advice. Correspondence concerning this article should be ad- dressed to Ute R. Hu ��lsheger, Faculty of Psychology and Neuroscience, Work- and Social Psychology, Maastricht University, PO Box 616, 6200 MD Maastricht, the Nether- lands. E-mail: ute.hulsheger@maastrichtuniversity.nl Journal of Occupational Health Psychology 2011, Vol. 16, No. 3, 361���389 �� 2011 American Psychological Association 1076-8998/11/$12.00 DOI: 10.1037/a0022876 361

size and direction of effects. This makes it difficult to draw reliable conclusions about the nature of the relationships and impedes the development of clear suggestions for management practices. A statistical integration of extant empirical research by means of a meta-analysis would consequently benefit this field of research in two important regards. First, a meta- analysis will help clarify whether the relationships of emotional labor facets with well-being and perfor- mance are generalizable across samples and settings or whether they are situation-specific. Second, it will allow the estimation of mean effect sizes of the relationships between emotional labor facets with well-being and performance outcomes and convey important information on the direction and strength of relationships. Overall, these findings will help evaluate propositions that have been made in models of emotional labor (Grandey, 2000 Holman, Mart�� ��nez-In ��igo, & Totterdell, 2008b Rubin et al., 2005). Furthermore, they will provide an overview of the benefits and costs of emotional labor and reveal pragmatic ramifications for organizational practices. In 2005, Bono and Vey provided a first quantita- tive summary of antecedents and consequences of emotional labor. They conducted a bare-bones meta- analysis based on 11 studies and 16 independent samples. The goal of the present study is to build upon and extend their research in five crucially im- portant regards: First, as Bono and Vey (2005) pointed out, the research field was not mature enough to compute robust estimates of associations between emotional labor and its consequences when they con- ducted their first meta-analysis. Yet research on emo- tional labor has been blossoming in recent years, and many studies have been published since 2005. We will therefore base our meta-analysis on a consider- ably enlarged sample of primary studies, which will benefit the reliability and stability of meta-analytic estimates (Hunter & Schmidt, 2004). Second, in their bare-bones meta-analysis Bono and Vey corrected studies for sampling error only that is, they estimated sample size weighted mean correlations between the variables in question. In addition to sampling error we will correct for predictor and criterion unreliabil- ity. Third, we will investigate not only well-being but also performance outcomes and their relationship with the three central emotional labor facets of emo- tion���rule dissonance, surface acting, and deep acting. Fourth, we will conduct a meta-analytic regression analysis, testing the unique contribution of each of the three emotional labor facets in statistically pre- dicting well-being and performance outcomes. Fifth and finally, we will test whether surface acting (par- tially) mediates the relation between emotion���rule dissonance and outcome variables. Deep Acting, Surface Acting, and Emotion���Rule Dissonance Extant models of emotional labor conceptualize emotion regulation���the process of managing ex- pressions and feelings by the two emotional labor strategies of surface and deep acting���as the core of emotional labor (Grandey, 2000 Holman, Mart����nez- In ��igo, & Totterdell, 2008b Rubin et al., 2005). Deep acting is an antecedent-focused form of emo- tion regulation that affects the perception and pro- cessing of emotional cues at the onset of an emotion that is, before they elicit behavioral, experiential, or physiological response tendencies (Gross, 1998). An- tecedent-focused emotion regulation occurs before an emotion develops, and it aims at changing the situa- tion or the perception of a situation (Grandey, 2000 Gross, 1998). When engaging in deep acting, indi- viduals try to align required and true feelings. To reach this goal, they can direct attention toward plea- surable things or thoughts to stir up the required emotion (attentional deployment), or reappraise the situation to induce the required emotion (cognitive change Grandey, 2000). Consequently, deep acting results in genuine emotional displays of the required emotions. Surface acting, on the other hand, is a response- focused form of emotion regulation that is applied when the emotion has already developed. It does not involve an adjustment of one���s actual feelings, but refers to the management of the emotional expres- sion. Individuals engaging in surface acting put on a mask. They adjust the emotional response by sup- pressing, amplifying, or faking emotions. In conse- quence, the emotional experience and the emotion expression remain discordant when individuals en- gage in surface acting (Grandey, 2000 Gross, 1998 Totterdell & Holman, 2003). Apart from deep and surface acting, emotional dissonance is considered in models of emotional la- bor (Holman, Mart�� ��nez-In ��igo, & Totterdell, 2008b Rubin et al., 2005). Indeed, many researchers ascribe emotional dissonance a central role in the emotional labor process (e.g., Co ��te ��, 2005 Morris & Feldman, 1996 Van Dijk & Kirk, 2006). However, researchers have used different and sometimes ambiguous con- ceptualizations of the concept (cf. Van Dijk & Kirk, 2006). Early work on emotional labor described emo- tional dissonance as the discrepancy between felt 362 HULSHEGER �� AND SCHEWE

emotions and emotions that are expressed to meet organizational display rules (Hochschild, 1983 Rafaeli & Sutton, 1987). Thus, emotional dissonance does involve three different aspects: emotions re- quired by display rules, expressed emotions, and felt emotions (Zerbe, 2000). Researchers have used dif- ferent combinations of these three aspects to concep- tualize and measure emotional dissonance. Some view emotional dissonance as the discrepancy be- tween required and felt emotions (e.g., Morris & Feldman, 1996 Zapf & Holz, 2006), which has also been referred to as ���emotion���rule dissonance��� (Hol- man, Mart�� ��nez-In ��igo, & Totterdell, 2008b), others conceptualize it as the discrepancy between ex- pressed and felt emotions (Co ��te ��, 2005 Van Dijk & Kirk, 2006), which has been circumscribed as ���fake emotion display��� by Holman and colleagues. These differences in conceptualizations of emotional disso- nance have important implications for the role as- cribed to emotional dissonance in the emotional labor process. While emotion���rule dissonance is an ante- cedent to emotion regulation in terms of deep and surface acting, fake emotion display is a consequence of emotion regulation (Holman, Mart�� ��nez-In ��igo, & Totterdell, 2008 Co ��te ��, 2005). Although different points of view exist regarding the conceptualization of emotional dissonance, the majority of research assesses emotional dissonance as emotion���rule dis- sonance (Dormann & Kaiser, 2002 Holman, Chiss- ick, & Totterdell, 2002 Zapf & Holz, 2006). This is also in line with theoretical models of emotional labor defining emotional dissonance as an antecedent to surface and deep acting (Holman, Mart�� ��nez-In ��igo, & Totterdell, 2008b Rubin et al., 2005). In the fol- lowing we will therefore focus on emotional disso- nance in terms of emotion���rule dissonance. Emotion���rule dissonance is a form of person-role conflict (Abraham, 1999 Rafaeli & Sutton, 1987) stemming from the incongruence between emotions that are actually felt and emotions that are required by display rules (Abraham, 1999 Brotheridge & Grandey, 2002 Morris & Feldman, 1996 Morris & Feldman, 1997 Van Dijk & Kirk, 2006) and result- ing in an unpleasant state of tension. Although there are conceptual and empirical relations between emo- tion���rule dissonance and surface acting, the concepts are to be carefully differentiated. While emotion���rule dissonance describes a ���state of being��� (p. 97 Grandey, 2000), surface acting describes the effortful process of managing one���s emotions. Thus, emotion��� rule dissonance is an emotional state, while surface acting is the active process of managing emotions. Consequences of Emotional Labor Emotional labor is a multifaceted construct which has been argued to have positive as well as negative consequences for individuals and organizations (Co ��te ��, 2005 Zapf & Holz, 2006). Emotion���rule dis- sonance, surface acting, and deep acting are expected to relate differentially to well-being and performance outcomes. These differential relationships can be ex- plained by different mechanisms underlying emo- tion���rule dissonance, surface, and deep acting and by the extent to which these threaten or conserve inter- nal resources (Hobfoll, 1989 Holman, Mart����nez- In ��igo, & Totterdell, 2008b). In deriving our hypoth- eses about the link of these three central aspects of emotional labor with well-being and performance outcomes we draw on established theoretical models of emotional labor (Grandey, 2000 Holman, Mart�� ��nez-In ��igo, & Totterdell, 2008b Rubin et al., 2005). An overview of our theoretical framework is depicted in Figure 1. The Relation of Surface Acting and Deep Acting With Well-Being and Performance Various mechanisms may be put forth explaining the relationships of surface and deep acting with well-being and performance outcomes. Ego-depletion. According to Baumeister and colleagues, purposeful self control and regulatory processes are effortful and deplete mental resources (Baumeister, Bratslavsky, Muraven, & Tice, 1998 Muraven, Tice, & Baumeister, 1998). Surface acting involves the constant monitoring of actual and de- sired emotions. Consequently, regulating emotions by surface acting is an effortful process that can be expected to drain mental resources. Indeed, funda- mental research on emotion regulation has revealed that regulating emotions by faking, suppression, or exaggeration impairs subsequent performance on di- verse tasks, such as hand-grip or anagram tasks (Baumeister et al., 1998 Muraven, Tice, & Baumeis- ter, 1998 Schmeichel, Demaree, Robinson, & Pu, 2006 Schmeichel, Vohs, & Baumeister, 2003). Re- searchers have also contrasted the depleting effects of response-focused and antecedent-focused emotion regulation. Their studies revealed that in contrast to antecedent-focused emotion regulation, response- focused emotion regulation was effortful and led to impaired mental performance, for instance on mem- ory and complex decision-making tasks (Richards & Gross, 1999 Richards & Gross, 2000 Zyphur, War- ren, Landis, & Thoresen, 2007). It has therefore been 363 COSTS AND BENEFITS OF EMOTIONAL LABOR

argued that surface acting, a response-focused form of emotion regulation, requires considerable mental effort. When employees engage in surface acting, actual and desired emotions need to be constantly monitored and the individual needs to invest contin- uous effort to change the emotional expression. This continuous effort drains mental resources and thereby enhances strain and diminishes well-being (Co ��te ��, 2005 Grandey, 2003 Mart�� ��nez-In ��igo, Totterdell, Alcover, & Holman, 2007). Moreover, it has been argued that, to the extent that surface acting draws on a limited reservoir of mental resources (Sideman Goldberg & Grandey, 2007 Zyphur et al., 2007), these resources are lacking for the performance of other job-related tasks that involve executive func- tioning. Surface acting can therefore be expected to impair not only employee well-being but also perfor- mance. In contrast, building on Gross and colleagues��� re- search on the cognitive costs of suppression and reappraisal (Gross, 1998 Richards & Gross, 1999, 2000) researchers have argued that deep acting re- quires less cognitive resources than surface acting (Totterdell & Holman, 2003). This contention rests on two assumptions, namely that deep acting is sim- ilar to reappraisal and that the reappraisal processes involved in deep acting diminish mental resources only at the onset of an emotion. It has consequently been argued that the amount of cognitive (e.g., atten- tion) and motivational (e.g., drive, resilience) re- sources invested is considerably lower for deep act- ing than for surface acting (Sideman Goldberg & Grandey, 2007 Totterdell & Holman, 2003). How- ever, recently, this assumption has been challenged by Liu and colleagues (Liu, Prati, Perrewe, & Ferris, 2008) who argued that the suppression and reap- praisal mechanisms investigated in laboratory set- tings cannot be compared with the workplace where employees need to regulate their emotions. They suggest that in contrast to the reappraisal manipula- tions in typical laboratory studies, actual deep acting might require ���a great deal of mental energy in the form of motivation, engagement, and role internal- ization��� (p. 2416) and might therefore be even more psychologically demanding than surface acting. Be- cause no study has examined the actual cognitive and motivational energy demand of deep and surface acting directly, the question whether deep acting con- sumes more or less mental resources than surface acting cannot be answered yet. Nevertheless, it can be concluded that deep acting is an effortful regula- tory process that drains mental resources to a cer- tain extent. Felt inauthenticity. People strive toward au- thenticity and self-expressive behavior, but display rules might impede an employee���s genuine experi- ence and expression (Hochschild, 1983). Especially surface acting may constrain personal authenticity, Figure 1. Model based on Grandey (2000), Holman et al. (2008b), and Rubin et al. (2005). 364 HULSHEGER �� AND SCHEWE

because employees��� emotional expressions and ac- tual feelings are at odds (Brotheridge & Lee, 2002). Empirical studies illustrated that the suppression of negative feelings and the simulation of positive emo- tions lead to lower self-authenticity (Brotheridge & Lee, 2002 Erickson & Ritter, 2001 Simpson & Stroh, 2004). Inauthenticity, in turn, is associated with depressed mood and stress (Erickson & Wharton, 1997 Sheldon, Ryan, Rawsthorne, & Ilardi, 1997). In contrast to surface acting, there is no discrepancy between felt and displayed emotions when employees engage in deep acting. When em- ployees use deep acting strategies their sense of au- thenticity is consequently not compromised. These theoretical arguments and empirical findings suggest a negative relationship between surface acting and well-being but not between deep acting and well- being. Authenticity of the emotion display. Satisfy- ing organizational display rules is an important ele- ment of performance in jobs involving interactions with clients. A customer service representative who is cheerful and friendly, a judge who has a neutral demeanor, a debt collector who displays anger, and a doctor who shows sympathy, they all adapt their emotional displays to job- and organization-specific display rules to fulfill their job roles. Display rules exist because it is assumed that displaying these specific emotions will influence clients and custom- ers in a particular way: The customer service repre- sentative is friendly because he wants an angry cus- tomer to calm down and abide by the organization, the debt collector expresses anger because he wants the debtor to pay her bill, and the doctor wants to give her patients hope and confidence. This is in line with the Emotion as Social Information Model positing that emotional displays provide observers with im- portant information and influence their behavior (Van Kleef, 2009 see also Keltner & Haidt, 1999). Yet emotional displays differ in the extent to which they are authentic or faked, and individuals are able to differentiate between genuine and fake emotional displays (Ekman, Friesen, & O���Sullivan, 1988). Only authentic emotional expressions entail the relevant cues that serve important social functions and have the desired effects on other individuals. A debt col- lector can only achieve his goal of enforcing a debt if the expressed anger is perceived as authentic by the debtor. Similar lines of argument apply when orga- nizations prescribe the display of positive emotions. Positive emotional displays evoke positive reactions only to the extent that others perceive them as au- thentic (Grandey, Fisk, Mattila, Jansen, & Sideman, 2005a Hennig-Thurau, Groth, Paul, & Gremler, 2006). Fundamental as well as applied experimental research revealed that authentic smiles elicit favor- able reactions from interaction partners as opposed to inauthentic smiles (Frank, Ekman, & Friesen, 1993 Hennig-Thurau et al., 2006). Evoking positive emo- tions in customers helps building up a strong employee��� customer rapport which is central to performance in terms of customer satisfaction and future loyalty in- tentions (Grandey, 2003 Hennig-Thurau et al., 2006). Surface acting is associated with inauthentic emotional expressions, while deep acting involves the authentic expression of emotions. These lines of arguments suggest surface acting to be negatively and deep acting to be positively related to perfor- mance outcomes, especially emotional performance and customer satisfaction. Enhancement versus impairment of social in- teractions. With his social interaction model of emotional labor, Co ��te �� (2005) drew attention to inter- personal processes that may explain how emotional labor relates to well-being. Co ��te �����s model builds upon a transactions framework (Rafaeli & Sutton, 1987), taking both interaction partners, that is, the employee and the customer, into account. The model suggests that the employee���s emotional display is appraised by the customer who in turn responds accordingly and thereby reaffects the employee and his or her emotional and psychological state of health. As men- tioned earlier, interaction partners are able to differ- entiate between authentic and inauthentic emotional displays (Grandey et al., 2005a), and they react more unfavorably to inauthentic compared with authentic displays of emotions (Hennig-Thurau et al., 2006). Surface acting involves inauthentic emotional dis- plays and thereby hinders positive interactions and evokes negative reactions from interaction partners. These negative reactions, for instance anger, disap- pointment, or disrespect, are stressors that reaffect the employee and impair his or her well-being. In contrast to surface acting, amplifying positive emotions through deep acting should result in favor- able responses by the interaction partner (Cote��, 2005). As clients or customers perceive authentic emotional displays, they respond favorably and ex- press positive emotions toward the employee. The result is an overall positive, satisfying interaction between employees and clients that is experienced as rewarding and provides the employee with a feeling of efficacy and personal accomplishment (Brother- idge & Lee, 2002). According to the conservation of resources theory (Hobfoll, 1989), experiencing re- warding social relationships at work is a resource that 365 COSTS AND BENEFITS OF EMOTIONAL LABOR