Emotional labor: A conceptualization and scale development

https://doi.org/10.1016/S0001-8791(03)00038-1Get rights and content

Abstract

Despite increased research attention, the emotional labor construct remains without a clear conceptualization and operationalization. This study designed a conceptually grounded, psychometrically sound instrument to measure emotional labor with an emphasis on the experience of discrete emotions—the Discrete Emotions Emotional Labor Scale (DEELS). This conceptualization and operationalization of emotional labor departs from existing efforts because it focuses on the behavior of emotional expression, encompassing genuine, faked, and suppressed positive and negative emotional displays. Results provide initial evidence for the convergent, discriminant, and criterion-related validity of the DEELS.

Introduction

Emotions are a pervasive, inseparable part of the human experience and of organizational life. Emotions shape perceptions, direct behavior, and influence interactions with others (Frijda, 1986). Despite their pervasiveness, research has only recently begun to examine the role of emotions and affect in the workplace (Brief & Weiss, 2002; Weiss & Cropanzano, 1996). One area within the emotional arena receiving increased research attention is emotional labor, a construct first defined by Hochschild (1983) as the “management of feeling to create a publicly observable facial and bodily display” (p. 7).

Following a number of qualitative articles describing the nature and outcomes of emotional labor in a variety of occupational groups (e.g., James, 1989; Rafaeli & Sutton, 1991; Stenross & Kleinman, 1989; Sutton, 1991; Tolich, 1993), the 1990s witnessed a shift to more quantitative conceptualizations and operationalizations (e.g., Brotheridge & Lee, 1998; Mann, 1999a; Morris & Feldman, 1997; Wharton, 1993). However, consensus among researchers has yet to be reached on either theoretical or methodological grounds. Conceptual and operational clarity is needed to advance emotional labor research. Accordingly, this study presents a conceptualization of emotional labor, development of a corresponding operationalization—the Discrete Emotions Emotional Labor Scale (DEELS)—and preliminary validation evidence.

Following Hochschild’s (1983) original conceptualization of emotional labor, several others have been advanced (see Grandey (2000) and Zapf (2002) for reviews). Conceptual ambiguity persists, but each conceptualization has in common the underlying assumption that emotional labor involves managing emotions and emotional expression to be consistent with organizational or occupational “display rules,” defined as expectations about appropriate emotional expression (Goffman, 1959). Emotional labor is “the act of displaying appropriate emotion (i.e., conforming with a display rule)” (Ashforth & Humphrey, 1993, p. 90), regardless of whether the emotion is discrepant with internal feelings. This commonality in conceptualization is accompanied by differences in theoretical approaches. Generally, theoretical treatments converge around three themes: internal states, internal processes, and external behavioral displays.

The first theoretical perspective emphasizes the internal state of emotional dissonance, or “the state that exists when there is a discrepancy between the emotional demeanor that an individual displays because it is considered appropriate, and the emotions that are genuinely felt but that would be inappropriate to display” (Mann, 1999a, p. 353). Emotional dissonance, like cognitive dissonance, creates an unstable state within the individual and may lead to negative outcomes such as estrangement between self and true feelings (Hochschild, 1983), job-related stress (Adelmann, 1995; Pugliesi, 1999; Wharton, 1993), and emotional exhaustion (Morris & Feldman, 1997).

Researchers agree that dissonance is a component of emotional labor, but there is disagreement over whether it is a necessary condition. Mann argued that emotional labor is present only when an individual fakes or suppresses an emotion; she excluded genuinely felt displays in her conceptualization. Ashforth and Humphrey argued that emotional labor is performing in accordance with display rules; an employee who genuinely feels enthusiastic and appropriately expresses this is still performing work, although he or she is not experiencing dissonance. In this case there is arguably a good fit between the employee and the requirements of the position (Arvey, Renz, & Watson, 1998).

Given that experiencing emotions increases one’s level of physiological and psychological arousal, expressing genuinely felt emotions may lead to emotional exhaustion (Ashforth & Humphrey, 1995). Following Morris and Feldman’s (1996) argument, frequent expression of a variety of intense emotions for a long duration may constitute labor. Thus, dissonance may not be required for emotion work to be laborious.

The second theoretical perspective focuses on the internal processes involved in creating an emotional display, typically self-regulation processes (Brotheridge & Lee, 1998; Grandey, 2000). Gross (1998) defined emotional regulation as “the process by which individuals influence which emotions they have, when they have them, and how they experience and express these emotions” (p. 275). Emotional labor research has translated these regulatory processes into notions of deep acting (attempts to modify internal feelings to be consistent with display rules) and surface acting (modifying outward displays to be consistent with display rules). Both surface acting and deep acting produce behavioral emotional display, albeit through different means.

Notions of surface and deep acting are linked with those of emotional dissonance. Brotheridge and Lee (1998) acknowledge that surface acting may be a manifestation of dissonance. These processes appear to be the bridge between the internal state and behavioral display; specifically, emotional dissonance drives self-regulation processes that are in turn manifested in behavioral emotional displays.

The third theoretical approach focuses on the external behavioral displays of employees. Ashforth and Humphrey’s (1993) conceptualization of emotional labor, “the act of displaying appropriate emotion (i.e., conforming with a display rule),” emphasizes the act or behavior rather than the internal state or process driving such behavior. Ashforth and Humphrey “prefer to focus on behavior rather than on the presumed emotions underlying behavior” (p. 90). They argue that compliance with display rules is ultimately manifested in behavior that is observable and influences interaction (e.g., service transactions). Similarly, self-regulatory processes are ultimately manifested in behavioral display. A focus on the behavioral display of emotion may be beneficial given the difficulty in tapping the unobservable dissonant states and internal processes of individuals.

External behavioral displays are given attention in all of the approaches and we believe the behavior of emotional expression (or the lack of it) is the most proximal component of emotional labor. In addition, we believe it is necessary to examine the interplay of felt emotion in conceptualizing the construct. Researchers should be aware of the internal emotional states and processes in addition to behavioral displays, and attempt to integrate these components for complete understanding of emotional labor (Rubin, Tardino, Daus, & Munz, in press).

It could appear that the emotional labor domain is in a theoretical quandary, flooded with a multitude of conceptualizations. However, recognizing the complexity of emotional expression on the job, emotional labor may best be conceptualized as a network of related constructs. The varying theoretical perspectives are not in opposition and may be viewed as complementary. For example, display rules may motivate an employee to experience an internal state of dissonance, requiring the employee to employ self-regulation strategies, resulting in an observable emotional display. The theoretical orientation a researcher adopts will depend on the research question. Recognizing the complexity of emotional labor as a network of distinct but related constructs can advance emotional labor research.

Our conceptualization of emotional labor focuses primarily on behavioral expression and non-expression of felt or unfelt emotions in accordance with display rules. We propose that emotional labor is the (1) expression of emotions and (2) non-expression of emotions, which may or may not be felt, in accordance with display rules. Emotional displays may be characterized broadly as either positive or negative. This framework may be illustrated along two dimensions as shown in Fig. 1. The first dimension classifies a display as either an appropriate expression of an emotion or an appropriate non-expression of an emotion. The second dimension is a felt continuum, indicating whether the expression or non-expression is consistent with an internal feeling.

All cells represent compliance with display rules. To illustrate, suppose that negative displays are prohibited and positive displays are required. Cells 2 and 3 represent cases where compliance is inconsistent with felt emotion. In Cell 2, negative emotions are felt, but not expressed (an appropriate suppressed display). In Cell 3, positive emotions are expressed, but not felt (an appropriate faked display). These cells have been the primary focus of emotional labor research. Cells 1 and 4 represent cases where compliance is consistent with felt emotion. In Cell 4, positive emotions are both felt and expressed (an appropriate genuine display). In Cell 1, emotion is neither felt nor expressed. Cell 1 will be excluded from the proposed measure as it represents the absence of both felt and expressed emotion, even though not expressing an unfelt emotion in accordance with display rules is conceptually labor according to the proposed definition.

Although in general agreement with Ashforth and Humphrey’s behaviorally focused conceptualization, our conceptualization extends and is differentiated from previous work in several ways. First, our framework explicitly accounts for underlying felt emotion that co-occurs with conformance to display rules. Ashforth and Humphrey’s definition does not account for the underlying emotional state, though they do note that individuals may conform to display rules with or without feeling the corresponding emotions (e.g., a service worker may either genuinely display enthusiasm when interacting with customers, or this enthusiasm may be fake).

Second, the proposed conceptualization acknowledges that conforming to display rules may involve expressing an appropriate emotion or not expressing an inappropriate emotion (e.g., conformance may involve expressing enthusiasm or suppressing frustration).

Third, we recognize the role of genuinely felt displays. Genuinely felt displays as well as faked displays and suppression may constitute emotional labor. Faking and suppressing may be more taxing than genuinely expressing emotions, but inclusion of genuine displays in an emotional labor framework and corresponding operationalization allows for greater comprehensiveness. Existing operationalizations do not explicitly distinguish among genuine expression, faked expression, and suppression dimensions (e.g., sample items include “When I work with customers/clients, the way I act and speak often doesn’t match what I really feel” and “Pretend to have emotions that you don’t really feel” and “I was not really being me” (Brotheridge & Lee, 1998; Mann, 1999a, Mann, 1999b; Morris & Feldman, 1997)).

Fourth, our conceptualization distinguishes between positive and negative emotions. Display rules may often require expressing positive emotions and not expressing negative emotions, but such norms may not always be the case and should not be assumed. Further, research suggests that positive and negative affective states have distinct behavioral antecedents and correlates (Carver, 2001; Tellegen, Watson, & Clark, 1999; Watson, Weise, Vaidya, & Tellegen, 1999); this may extend to emotional labor as well.

Finally, our framework encompasses other forms affect in general, not exclusively emotions, but also moods. Frijda (1993) suggests that moods are typically less intense, of shorter duration, or less likely to be attributable to a specific object and to have a specific response, as compared to emotions. Affect is a term used to encompass a variety of feeling states including moods and emotions (George, 1996). Thus the emotional labor construct might be better termed “affective management.” However, we will retain the term emotional labor by convention, recognizing that the construct broadly encompasses emotions as well as moods.

The proposed operationalization is comprised of three subscales representing three cells in Fig. 1—genuine expression (Cell 4), faked expression (Cell 3), and suppression (Cell 2). Each subscale includes 14 discrete positive and negative emotions (e.g., enthusiasm, happiness, frustration, and sadness) representative of six emotion families (i.e., love, joy, anger, sadness, fear, and hate). Because the proposed operationalization includes specific emotions beyond a broad positive–negative distinction, we have named the measure the Discrete Emotions Emotional Labor Scale—the DEELS. Measurement of discrete emotions may be the greatest departure of the DEELS from alternative assessments that often elect not to specify the emotional state or combine potentially different emotional experiences (e.g., “Pretend to have emotions that you don’t really feel,” Morris & Feldman, 1997; “I laughed or frowned…” Mann, 1999a, Mann, 1999b). By assessing specific emotions representative of different emotion families, the DEELS removes the guesswork regarding what respondents consider positive and negative emotions. This focus on discrete emotions is consistent with a recent call for increased attention to discrete emotions rather than more general moods and affective states (Brief & Weiss, 2002).

The discrete emotions represented in the scale are derived from a semantic classification proposed by Shaver, Schwartz, Kirson, and O’Connor (1987) which reduced a variety of emotions into six families: love, joy, surprise, anger, sadness, and fear. Two additional categories were considered: (1) the hate category, thought to reflect the negative valence of the love family, was included due to its potential relevance in a work context; and (2) the shame family, which concerns the self and cognitions about the self (Lewis, 1993), was included to tap possible self-referent emotions. Shaver et al. provide exemplars of each of the basic emotion categories. For example, the basic emotion category anger includes subordinate emotions such as irritation, agitation, annoyance, grouchiness, grumpiness, rage, fury, wrath, spite, etc. Initial item selection was done by the authors to select items that (1) were identified as the six basic emotions in Shaver et al., (2) were thought to be in common language usage (e.g., irritation rather than fury), and (3) were likely to be experienced in a work setting (e.g., aggravation rather than revulsion).

Forty emotion words were selected from this initial procedure. The 40 items were further reduced to 18 items based on (1) perceived potential ambiguity (e.g., affection may connote several forms of liking); (2) redundancy with other items (e.g., surprise was selected rather than astonishment); and (3) representation of differing levels of emotional intensity (e.g., inclusion of both disliking and hate).

A pilot study in which the DEELS was administered in paper-and-pencil format was conducted to investigate item characteristics and item functioning in a sample of 112 Masters students (61% female; average age was 26.5 years; 64% White, 13% Asian, 3% Black, 3%, Hispanic, and 17% other or missing). Following the pilot study, four items were removed from the scale. Emotions from the shame family (i.e., shame and humiliation) were removed in this phase. Shame is not one of the basic emotion families identified by Shaver et al. and its self-referent nature presents difficulties in conceptualizing how it may be expressed; these difficulties in expression may have been responsible for their infrequent endorsement. Love and surprise were removed due to participants’ confusion about what these items represent in a work setting as well as having low endorsement. The 14 remaining items had adequate item means and distributions. The final scale of 14 items represents six categories—the original Shaver et al. categories of basic emotions (with the exception of the surprise category) and the additional hate category. Each emotion category contains two or three emotions as follows: love category (liking and concern), joy category (enthusiasm, happiness, and contentment), anger category (anger, aggravation, and irritation), sadness category (distress and sadness), fear category (fear and anxiety), and the hate category (hate and disliking). The scale does not capture the full range of emotions, however the items are representative of the basic emotion families and these families are designed to be representative of the construct space.

As noted above, the DEELS is comprised of three subscales—genuine expression, faking, and suppression. For each subscale, respondents are asked to consider each discrete emotion in relation to their interactions with customers/clients, supervisors, and coworkers over a six-month period. For the genuine subscale, respondents are asked, “How often do you genuinely express (enthusiasm) when you feel that way?” For the faked subscale, they are asked, “How often do you express feelings of (enthusiasm) on the job when you really don’t feel that way?” For the suppression subscale, they are asked, “How often do you keep (enthusiasm) to yourself when you really feel that way?” For the faked and suppression subscales, the words “faking” and “suppressing” are purposely omitted due to their socially undesirable connotations. For the genuine subscale, response choices range from (1) “I never genuinely express this” to (5) “I genuinely express this many times a day.” For the faked subscale, response choices range from (1) “I never express this when I do not feel like it” to (5) “I express this many times a day when I do not feel like it.” The response choices for the suppression subscale include an additional item: “I never feel this.” This response option is analyzed as a missing value, as never feeling an emotion does not allow for suppression of that emotion. The scale is presented in Appendix A.

As is evident from the response options, the DEELS assesses frequency of emotional expression. In addition to frequency, Morris and Feldman include variety, intensity, and duration as dimensions of external emotional displays. While only frequency is directly measured, the DEELS does provide an indirect means to assess variety and intensity; variety may be captured through our inclusion of emotions representative of different emotion families and intensity may be captured though our inclusion of emotions with varying intensities. A focus on frequency may not necessarily be a severe limitation of the proposed measure, following Morris and Feldman’s argument (1996) that frequency of interaction is the central component of emotional labor. Further, Zapf (2002) notes that “all studies that somehow measured emotion work measured the frequency, and it was the basic idea of Hochschild (1983) that too frequent emotional displays would overtax the employee and lead to alienation and exhaustion” (p. 242). Thus, the DEELS’ emphasis on frequency is consistent with emotional labor research.

Our three-fold validation strategy examined item and scale properties, compared scale scores across samples in a known-groups validation, and examined relationships between the DEELS and other constructs.

Confirmatory, rather than exploratory, factor analysis was chosen to test competing models of the DEELS given that we relied on an existing framework of emotions and could make specific a priori hypotheses about the factor structure. Further, the use of exploratory factor analysis when an a priori factor structure exists has received criticism (Armstrong, 1967). Confirmatory analyses reduce the likelihood that a specific structure would be supported by a covariance matrix by chance.

We assessed three alternative models: a 1-factor solution including all scale items; a 3-factor solution based on the genuine, faking, and suppression subscales; and a 6-factor solution dividing the genuine, faking, and suppression subscales into positive and negative emotions. Given our proposed conceptualization of emotional labor and the evidence on positive and negative emotions constructs, we expected the 6-factor solution to provide the best fit.

The second step of the DEELS validation was to perform a known-groups validation (DeVellis, 1991). Known-groups validation can establish construct validity by assessing the measure’s ability to differentiate groups that are known to differ a priori. Because research suggests that health service jobs require positive expression and police work allows for negative expression (Hochschild, 1983; Rafaeli & Sutton, 1991), samples of assisted living employees and police investigators were selected for comparison. These groups were hypothesized to differ in their frequency of genuine expression, faking, and suppression of positive and negative emotions. Specifically, the following hypotheses were tested:

  • H1.

    Compared to police investigators, the assisted living employees will report greater frequency of:

    • H1a.

      genuinely expressing positive emotions

    • H1b.

      faking positive emotions

    • H1c.

      suppressing negative emotions

  • H2.

    Compared to the assisted living employees, the police investigators will report greater frequency of:

    • H2a.

      genuinely expressing negative emotions

    • H2b.

      faking negative emotions

    • H2c.

      suppressing positive emotions

We gathered additional support for construct validity by examining relations between the DEELS and other scales in its “nomological network” (Cronbach & Meehl, 1955). First, we examined relations between the DEELS and existing dissonance and surface acting measures (Brotheridge & Lee, 1998; Morris & Feldman, 1997) to establish convergent validity. These measures reflect expressing emotions that are unfelt and were expected to be positively correlated with the DEELS faking and suppression subscales.

  • H3.

    The DEELS faking and suppression scales will be correlated with the dissonance and surface acting dimensions of emotional labor from alternative measures.

Second, we examined relations between the DEELS and Morris and Feldman’s (1997) duration measure to establish discriminant validity. The DEELS assesses frequency of emotional expression and is expected to be relatively independent of duration of interaction.

  • H4.

    The DEELS faking and suppression scales will be uncorrelated with Morris and Feldman’s duration dimension.

Finally, we examined the relationships between the faking and suppression subscales and emotional exhaustion to establish the criterion-related validity of the DEELS. A frequently reported outcome of emotional dissonance is emotional exhaustion or other components of burnout (Abraham, 1998; Brotheridge & Lee, 1998; Brotheridge & Lee, 2002; Grandey, 2000; Hochschild, 1983; Kruml and Geddes, 2000a, Kruml and Geddes, 2000b; Morris & Feldman, 1997). Thus, we expected frequent discrepancy between felt and expressed emotion, as measured by the faking and suppression scales of the DEELS, to lead to emotional exhaustion.

  • H5.

    The DEELS faking and suppression subscales will be positively correlated with emotional exhaustion.

Section snippets

Procedure and participants

Data were collected from five samples: a graduate student sample and employee samples from four workplace settings—a hotel, a managed healthcare organization, an assisted living group home organization, and a metropolitan police force. Samples were selected based on convenience and on an attempt to collect data from diverse occupations likely to engage in differing levels and forms of emotional labor. After examining scale characteristics of the samples (see Appendix B), samples were combined

Confirmatory factor analyses

Confirmatory factor analyses conducted in LISREL VIII (Jöreskog & Sörbom, 1993) tested unidimensional, 3-, and 6-factor models using a combined sample (Samples 1–5). The unidimensional model had all items loading on a single factor. The 3-factor model separated the DEELS into genuine, faking, and suppression subscales. The 6-factor model further separated the genuine, faking, and suppression subscales into positive and negative emotions. To determine the extent to which the data conformed to

Discussion

The various analyses provided support for the conceptualization of emotional labor offered here and the validity of the DEELS. The confirmatory factor analyses showed reasonable support for the six subscales of the DEELS. Although the DEELS includes 14 discrete emotions, the ability to represent them with six factors is advantageous to researchers. Nonetheless, these discrete emotions more clearly specify the emotion space and may be of interest to researchers in certain contexts.

Support for

Acknowledgements

Portions of this research were supported by a Grant-In-Aid of Research, Artistry and Scholarship from the University of Minnesota. We are particularly grateful to the organizations that participated in this research.

References (42)

  • C.M Brotheridge et al.

    Emotional labor and burnout: Comparing two perspectives of “people work”

    Journal of Vocational Behavior

    (2002)
  • D Zapf

    Emotion work and psychological well being: A review of the literature and some conceptual considerations

    Human Resource Management Review

    (2002)
  • R Abraham

    Emotional dissonance in organizations: Antecedents, consequences, and moderators

    Genetic, Social, and General Psychology Monographs

    (1998)
  • P Adelmann

    Emotional labor as a potential source of job stress

  • J.S Armstrong

    Derivation of theory by means of factor analysis or Tom Swift and his electric factor analysis machine

    The American Statistician

    (1967)
  • R.D Arvey et al.

    Emotionality and job performance: Implications for personnel selection

  • B.E Ashforth et al.

    Emotional labor in service roles: The influence of identity

    Academy of Management Review

    (1993)
  • B.E Ashforth et al.

    Emotion in the workplace: A reappraisal

    Human Relations

    (1995)
  • A.P Brief et al.

    Organizational behavior: Affect in the workplace

    Annual Review of Psychology

    (2002)
  • Brotheridge, C. M., & Lee, R. T. (1998). On the dimensionality of emotional labor: Development and validation of an...
  • C.M Brotheridge et al.

    Testing a conservation of resources model of the dynamics of emotional labor

    Journal of Occupational Health Psychology

    (2002)
  • C.S Carver

    Affect and the functional bases of behavior: On the dimensional structure of affective experience

    Personality and Social Psychology Review

    (2001)
  • L Cronbach et al.

    Construct validity in psychological tests

    Psychological Bulletin

    (1955)
  • R.F DeVellis

    Scale development: Theory and application

    (1991)
  • N.H Frijda

    The emotions

    (1986)
  • N.H Frijda

    Moods, emotion episodes and emotions

  • E Goffman

    The presentation of self in everyday life

    (1959)
  • J.M George

    Trait and state affect

  • A.A Grandey

    Emotion regulation in the workplace: A new way to conceptualize emotional labor

    Journal of Occupational Health Psychology

    (2000)
  • J Gross

    The emerging field of emotion regulation: An integrative review

    Review of General Psychology

    (1998)
  • A.R Hochschild

    The managed heart

    (1983)
  • Cited by (317)

    • The role of volunteers in creating hospitality: Insights from museums

      2023, Journal of Hospitality and Tourism Management
    • The Emotional Toll of the COVID-19 Crisis on Local Government Workers

      2024, Review of Public Personnel Administration
    View all citing articles on Scopus
    1

    Fax: 1-607-254-2971.

    View full text