Emotion work as a source of employee well- and ill-being: the moderating role of service interaction type

ABSTRACT

This study presents a framework for classifying service occupations based on the type of interaction between employees and customers to clarify the mixed relationships between emotion work and well-being. Drawing on the challenge-hindrance stressor literature, we propose that positive emotion requirements are a challenge stressor for almost all service providers, while negative emotion requirements and emotion-rule dissonance are only challenge stressors in certain service contexts. Data from 33 independent samples comprising 7,075 employees showed that all emotion work variables were positively related to emotional exhaustion. In line with expectations, only positive emotion requirements were consistently and positively linked to personal accomplishment. Service interaction type measured by expert ratings was used to test under what circumstances negative emotion requirements and emotion-rule dissonance are challenge stressors with positive consequences. Results revealed that negative emotion requirements were positively related to personal accomplishment when interactions were highly complex and employees were highly required to identify with customers. When interactions were standardized and of low complexity, when service providers were substitutable, and when there was no need to identify, there were negative relationships. For emotion-rule dissonance, positive relationships emerged when employees engaged in standardized interactions, were substitutable, but needed to identify with customers.

KEYWORDS:

• Emotion work
• emotional labour
• challenge and hindrance stressors
• occupational stress
• service interaction type
• well-being

Service work has become the most important labour sector in developed countries and accounts for nearly three fourths of the labour force (Central Intelligence Agency, 2017). A central aspect of service work is the interaction of employees with customers, clients, patients, or others (henceforth customers). These interactions require a service provider to perform emotional labour, or emotion work (Zapf, 2002), which means displaying and suppressing emotions while concealing others. Numerous researchers have delved into emotion work, and have especially focused on the link to positive and negative consequences for employees. Various theoretical models distinguish emotion work requirements from the emotion work process (Bhave & Glomb, 2016; Brotheridge & Grandey, 2002; Grandey & Gabriel, 2015). The majority of scholarly work so far has investigated the process of emotion work, that is the regulation of inner feelings and emotional expressions, and has linked different emotion regulation strategies such as surface and deep acting to employee ill- and well-being (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013). There is less research that investigates emotion work at the level of job characteristics (e.g., Bhave & Glomb, 2016; Brotheridge & Grandey, 2002; Zapf et al., 2001).

These fewer studies, however, have revealed inconsistent findings on the relationships with employee well-being. On the one hand, Brotheridge and Grandey (2002), for example, found positive relationships of emotional labour frequency and requirements to hide negative feelings with psychological strain. This is consistent with Zapf et al. (2001), who identified similar relationships for positive emotion requirements, negative emotion requirements, sensitivity requirements and emotion-rule dissonance. On the other hand, the study by Bhave and Glomb (2016), for instance, revealed a positive link between emotional labour requirements in general and job satisfaction, and the study of Diefendorff and Richard (2003) found such a positive relation for positive emotion requirements. These equivocal results make it difficult to draw reliable conclusions for health-related job design and to provide clear suggestions for management practices, which is why scholars have repeatedly called for studies that can uncover the reasons for the positive and negative consequences (Bhave & Glomb, 2016; Grandey et al., 2013; Pugh et al., 2011).

To fill this gap, we propose to apply the concept of challenge and hindrance stressors (Cavanaugh et al., 2000; LePine et al., 2005) to emotion work variables because it provides a well-supported theoretical explanation why some stressors are positively associated with employee well-being beyond their positive link to strain. Building on studies that have defined the characteristics of challenge stressors (e.g., Kern et al., 2020; Van den Broeck et al., 2010), we examine which emotion work variable may be labelled a challenge or a hindrance stressor and which boundary conditions may affect a classification. We posit that the motivational potential attributed to challenge stressors (LePine et al., 2004) may help to better understand when and why employees enjoy to perform emotion work.

Requirements to show and feel organizationally desired emotions typically vary across occupations (Humphrey et al., 2008; Morris & Feldman, 1996). However, such occupational differences have often been neglected in research (Grandey et al., 2013), although occupations serve as a meaningful context variable that may influence how stressful emotion work is for job incumbents (Brotheridge & Grandey, 2002; Rafaeli & Sutton, 1987). In this study, we deal with an occupational portrait of emotion work in two ways: first, we distinguish between the requirement to show positive and negative emotions, as the necessity to display certain emotions is not per se high or low (e.g., Zapf & Holz, 2006; Zapf et al., 1999), and we include emotion-rule dissonance as one of the main stressors discussed in research on emotion work (Holman et al., 2008; Hülsheger & Schewe, 2011). Second, we examine the type of interaction that characterizes service occupations as a boundary condition. Research has emphasized the need for an integrative approach that clarifies occupation-related differences in the consequences of emotion work and goes beyond Hochschild’s (1983) occupational index (Grandey et al., 2013; Humphrey et al., 2008; Schaubroeck & Jones, 2000).

In doing so, the present study makes the following contributions to the literature: first, by merging individual-level survey data with occupation-level information, we unveil systematic variations in the relationships between emotion work variables and employee well-being and thus methodologically expand our knowledge in this field. This approach is less error-prone because it is less dependent on individual appraisal processes. Second, by applying the challenge-hindrance stressor framework, we bring together current stress perspectives with emotion work research to clarify why some emotion work variables may have favourable effects and psychological costs at the same time, whereas others are exclusively detrimental. In that way, we also complement the findings of Bhave and Glomb (2016) or Singh and Glavin (2017), who investigated emotion work requirements as a higher-order factor and could therefore only draw conclusions for generally high or low requirements. Third, we present and test a theoretical framework for the classification of service occupations based on the typology of Mills and Margulies (1980) to illuminate the different consequences of emotion work for service providers. This knowledge can provide researchers with a profound understanding of how occupations shape employees’ responses (cf. Johns, 2006), and practitioners with advice on which employees need special support to prevent potential harm.

Positive and negative consequences of emotion work

There is ample evidence suggesting that emotion work contributes to psychological strain, such as emotional exhaustion and psychosomatic complaints (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013), what can be attributed to the resource investment and regulatory effort. Long-term or excessive emotion work may rapidly deplete or even exceed resources (e.g., Xanthopoulou et al., 2013). This overtaxing of regulatory processes is linked to impaired well-being (Wong et al., 2017). Moreover, high emotion work requirements increase the likelihood of situations in which a service provider has to express emotions that do not match with the emotions one typically feels in such situations (Zapf, 2002). This so-called emotion-rule dissonance as a job characteristic (Holman et al., 2008) has been associated with employee burnout (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013). By contrast, research has also found positive associations of emotion work with job satisfaction (Bhave & Glomb, 2016), work motivation (De Jonge et al., 2008) and personal accomplishment (Zapf & Holz, 2006). To explain the ambivalent effects on individual well-being, we contend that emotion work may match with the concepts of challenge and hindrance stressors under given boundary conditions.

The challenge stressor – hindrance stressor framework

The challenge-hindrance stressor framework has been developed to explain why some stressors, called challenge stressors, have positive and negative effects simultaneously (Cavanaugh et al., 2000; LePine et al., 2005). Both challenge and hindrance stressors are similar in that they are considered stressful, require increased effort and exertion, and thus drain personal resources (LePine et al., 2007; Webster et al., 2011). As a result, they may lead to fatigue and emotional exhaustion (Boswell et al., 2004). In addition to their positive effects on strain, however, challenge stressors such as time pressure or job responsibility have been found to be positively linked to work engagement (Crawford et al., 2010), job satisfaction (Podsakoff et al., 2007) and personal accomplishment (Kern & Zapf, 2021). According to recent literature, stressors are labelled challenge stressors, if they a) relate to one’s core work tasks and foster the attainment of meaningful personal and/or work goals (Van den Broeck et al., 2010), b) are appraised as a legitimate and central aspect of the job (Schmitt et al., 2015; Semmer et al., 2019), c) lead to valued outcomes and offer opportunities for mastery, learning, and personal growth (e.g., Kern et al., 2020; Ohly & Fritz, 2010), and d) have a fair chance to be overcome successfully (Wallace et al., 2009).

Hindrance stressors are seen as obstacles or roadblocks that keep employees away from goal attainment and threaten personal growth and learning, thus eliciting negative emotions (Rodell & Judge, 2009). They refer to stressful aspects of one’s job that are either potentially avoidable (e.g., frequent computer breakdowns) or unreasonable (e.g., a high amount of annoying paperwork that should be done by someone else; cf. Semmer et al., 2007, 2019). Classic examples are personal conflicts, role ambiguity, or organizational constraints, all of which have exclusively been found to positively relate to health impairment and negatively to job satisfaction (LePine et al., 2004; Webster et al., 2011).

Applying the challenge – hindrance stressor framework to emotion work

Emotion work requirements may be labelled a challenge if the above-mentioned criteria for challenge stressors apply. In the next section, we discuss the possible classification for each of the following job characteristics: requirements to show positive emotions, requirements to show negative emotions, and emotion-rule dissonance.

Requirements to show positive emotions

The requirement to express positive emotions is closely linked to service work in general, as in most service occupations the expression of positive emotions, such as friendliness, is expected by explicit or implicit display rules. Dealing with customers in a positive manner enables employees to achieve organizational and/or personal goals. As displaying positive emotions conforms to normative societal expectations (Von Gilsa & Zapf, 2013), it is likely that positive emotion requirements are seen as a legitimate aspect of service work. As long as they are matched by the individual prerequisites of a service provider, there should be a fair chance of meeting these requirements successfully. This should be linked to feelings of self-efficacy and personal accomplishment (Zapf & Holz, 2006). However, being required to express positive emotions on a frequent basis should also tax individual resources and thus be linked to strain. Taken together, we assume that positive emotion requirements are more likely to match with the concept of challenge stressors, as they a) have a strong relationship to core work tasks, since they help in accomplishing service interactions, b) are usually task-inherent aspects in most service jobs and are therefore normally appraised legitimate and unavoidable, c) satisfy affiliation needs and are thus related to valued outcomes, and d) are likely to be coped with successfully. Thus, we expect that:

Hypothesis 1: Positive emotion requirements are positively related to (a) emotional exhaustion and to (b) personal accomplishment.

Requirements to show negative emotions

The requirement to display negative emotions is less common in typical service jobs. Rather, the absence of negative emotions is more common and their expression is mainly required in situations that are per se perceived as aversive and stressful. These situations include, for example, customer incivility and require an employee to reprimand the customer in order to protect one’s self-esteem (Zapf, 2002). Such negative emotion requirements are strongly related to burnout (Zapf et al., 2001, 1999). Another reason for the negative effects relates to the social meaning of negative emotion requirements (cf. Semmer et al., 2007). Bank tellers, call centre agents, and other service employees are more likely to interpret situations requiring the display of negative emotions as an unnecessary and unreasonable burden that should not occur in one’s job. They might be seen as an obstacle that complicates providing the service and hinders one from accomplishing the primary tasks. For these reasons, negative emotion requirements seem to be more in line with the concept of hindrance stressors in many service jobs.

Some service occupations, however, require the expression of negative emotions such as anger or disappointment on a regular basis. Teachers (Reyna & Weiner, 2001), bill collectors (Sutton, 1991), and police officers (Daus & Brown, 2012) are often cited in this regard. In these so-called “social control” occupations, negative emotions are usually required when customers who do not comply with expectations or regulations have to be sanctioned, when a service employee wants to motivate the customer to a certain behaviour or when a service provider wants to underscore the seriousness of an issue (Humphrey et al., 2008). A high school teacher, e.g., may feel required to express anger and frustration to rebuke a student for laziness or lack of attention with the aim of increasing the student’s academic performance. Zapf, Isic et al. (2003) argued that the requirement to show negative emotions is related to deliberate behaviour in order to influence the situation and to reach higher-level goals. In this case, negative emotion requirements are more likely to be appraised as central for one’s job and foster the attainment of intended goals, as long as they are successfully managed. This is in line with Stenross and Kleinman (1989), who found that criminal detectives enjoy dealing with criminals, although they were frequently required to express negative emotions. One reason for this might be that interactions with criminals are considered as being a primary task, reflecting “real detective work”. These assumptions match with the concept of challenge stressors, implying that there might also be a positive relationship between negative emotion requirements and personal accomplishment. Initial evidence comes from Zapf et al. (1999), who found a positive relationship for human service workers in a children’s home. We posit that negative emotion requirements cannot be unambiguously classified as a challenge or a hindrance stressor unless the type of service interaction is considered as a moderating condition. Since the relationships with personal accomplishment should be positive or negative depending on the type of interaction that characterizes a service job, we do not formulate a hypothesis here. However, as negative emotion requirements always require increased effort and are linked to regulatory costs, they will be positively related to strain. Thus, we expect:

Hypothesis 2: Negative emotion requirements are positively related to emotional exhaustion.

Emotion-rule dissonance

Various models on emotion work include the concept of emotion-rule dissonance, which is considered a job characteristic and refers to the frequency of situations in which organizational display rules require the expression of emotions that are normally not felt in the situation (Holman et al., 2008; Zapf, 2002). It relates to any emotion work requirement (positive and negative emotion requirements or sensitivity requirements etc.) and the respective discrepancy to one’s actual felt emotions and precedes emotion regulation in terms of surface and deep acting. It should be noted that emotion-rule dissonance can also refer to specific situations in which one is required to express a particular emotion that is not actually felt. This case rather describes a momentary specific state that should be named state emotion-rule dissonance (cf. Zapf et al., 2021). We understand emotion-rule dissonance as a job stressor stemming from a stable characteristic of recurring requirement situations of a specific workplace. Classic examples are employees in complaint management offices, where positive emotions often need to be shown in situations in which customers behave negatively.

Hülsheger and Schewe (2011) argued that emotion-rule dissonance matches with the concept of hindrance stressors for two reasons: first, it poses a double bind situation on employees because they can either adhere to organizationally required display rules that lead to feelings of inauthenticity or they behave according to their inner feelings and take the risk of dissatisfying a customer by ignoring organizational display rules. As a result, employees may interpret such situations as a failure to adequately handle difficult customer interactions, which induces an unpleasant state of tension. Second, emotion-rule dissonance shares the characteristics of a person-role conflict (Rafaeli & Sutton, 1987). Role conflicts have been identified as hindrance stressors (Dawson et al., 2016), as they are usually avoidable obstacles or roadblocks to goal attainment. In support of these notions, numerous studies have found positive relationships with emotional exhaustion (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013) and negative ones with personal accomplishment (e.g., Andela et al., 2016; Giardini & Frese, 2006).

There is broad evidence that emotion-rule dissonance is a hindrance stressor. We propose, however, that this is not always true. In some cases, emotion-rule dissonance can also be a challenge stressor for several reasons: first, the relationship with personal accomplishment was rather weak in the meta-analysis of Hülsheger and Schewe (2011), and some studies found positive relationships with personal accomplishment (Zapf et al., 1999) and job satisfaction instead (Stenross & Kleinman, 1989). Second, there could be service employees who interpret emotion-rule dissonance as a task-inherent, challenging and unavoidable part of their job (cf. Xanthopoulou et al., 2013). Zapf (2002) argued that employees may believe that they are able to attain valued goals and accomplish important work tasks by effectively managing such a situation despite the unpleasant aspects that emotion rule-dissonance encompasses. A service agent at baggage tracing, for example, may be very annoyed by the aggressive behaviour of a passenger who misses his or her baggage and gets upset, but is well aware that expressing that will not help to solve the situation. The employee might see the need to express more neutral emotions in order to detach from the situation and fulfil one’s professional role as a service agent, regardless of how the required emotional display is achieved. This perspective dovetails with suggestions from Rafaeli and Sutton (1987), who argued that the consequences of emotion-rule dissonance depend on the internalization of display rules. It seems therefore less possible to match emotion-rule dissonance with the concept of challenge or hindrance stressors unless third variables are considered (cf. Xanthopoulou et al., 2013). Accordingly, no hypothesis is formulated for the relationship with personal accomplishment. However, as emotion-rule dissonance should be linked to negative emotional states, regulatory costs and resource consumption (Hülsheger & Schewe, 2011), we expect that:

Hypothesis 3: Emotion-rule dissonance is positively related to emotional exhaustion.

Service interaction type as a boundary condition

We posit that negative emotion requirements and emotion-rule dissonance may be challenge or hindrance stressors depending on whether they represent a legitimate aspect of one’s job or violate one’s expectations and contravene one’s professional role (cf. Semmer et al., 2019). A self-pitying patient, for example, might be more of a challenging case for a nurse, whereas the same person might be a pesky customer for a hairdresser. Both variables might thus be meaningful and enjoyable for certain groups of service providers, while not for others. Accounting for these differences, Grandey and Diamond (2010) suggested considering interactional characteristics to enlighten differences in the relationships with employee well-being. Hülsheger and Schewe (2011) tested service interaction type as a moderator but did not find any effects in their meta-analysis that distinguished between service encounters and service relationships. We test a broader variety of service interaction type using the typology of service organizations developed by Mills and Margulies (1980). They distinguished between seven dimensions to describe the interaction between a focal organization and its customers. For our study, we adapted the proposed model to the service employee’s perspective regarding customer interactions (see Figure 1). In total, we distinguished four dimensions that include interaction complexity, interaction customization, non-substitutability, and identification with customers.

Figure 1. Measurement of service interaction type based on the typology by Mills and Margulies (1980) and scoring examples from the expert ratings.

To avoid redundancies, we merged three dimensions of the model by Mills and Margulies (1980) into one single factor named “interaction complexity” (Figure 1). The rationale behind the merging is that the dimensions “information” and “decision” denote a type of service that is rather complex and dependent on a higher degree of qualification, such as medical, therapeutic and counselling services. With regard to the “power” dimension, Mills and Margulies (1980) argued that employees are typically perceived in a superior position when services require abundant information and complex decision making. Therefore, our higher-order concept of “interaction complexity” should also cover status differences between employees and customers to a certain extent.

The other dimensions were only redefined regarding the service employees’ perspective. The dimension “problem awareness” is described as a customer’s knowledge about the problem. From a service provider’s perspective, this leads to a higher or lower level of standardization, which is why we named it “interaction customization”. “Transferability of the employee” indicates to what extent a service encounter is bound to a certain employee. We use the term “non-substitutability” for this dimension. In the original model, identification with customers, as well as the potential for conflicts, have been subsumed under “attachment”. We do not consider conflict potential, as it is rather related to customer stressors (cf. Dormann & Zapf, 2004) than to occupational differences, and therefore focus on the requirement to identify with customers (“identification with customers”).

One dimension of the model by Mills and Margulies (1980) was not used at all. The dimension “time”, referring to the duration of a customer-employee interaction, was dropped, as it is difficult to interpret. Grandey and Diamond (2010) argued that an evaluation of the temporal relationship between service providers and customers is hardly possible by considering the occupation alone. Rather, the time aspects may also depend on the location (e.g., barber shop in a big city vs. in a small town) or on the number of regular customers.

In the next section, we describe how the four operationalized dimensions influence the meaning of emotion work for employees in different service jobs, except for positive emotion requirements. The rationale for excluding positive emotion requirements is that there were no theoretical arguments for differential effects based on the characteristics of challenge and hindrance stressors. In order to be fully transparent, however, we will provide the results of the moderation analyses on an exploratory basis. An overview of our examination model is shown in Figure 2.

Figure 2. Examination model.

Complexity of service interactions

Interaction complexity determines the type of information exchange (Von Cranach et al., 1986). Service occupations characterized by complex interactions require high professional skills, since defining the nature and structure of a customer’s problem, as well as finding a strategy to solve it, is part of the interaction (to an extreme extent, this applies, e.g., to psychotherapists or physicians). In such service contexts, negative emotions are more likely to be required when employees deliberately intend to modify the interaction in order to achieve a specific objective (e.g., Stenross & Kleinman, 1989; Sutton, 1991). One may not express negative emotions just for fun, but only if absolutely necessary. Due to their strong link with higher-order goals, negative emotion requirements are likely to be a legitimate and central aspect of one’s job that increases employees’ coping motivation (cf. Holman et al., 2008; Rafaeli & Sutton, 1987). Additionally, employees typically enjoy greater autonomy and are also better equipped with the necessary instrumental means to manage customer interactions (Grandey & Diamond, 2010). Thus, they should be better able to deal effectively with negative emotion requirements and, as a result, perceive customer interactions as more rewarding and satisfying, although negative emotions are involved (Lilius, 2012). In this case, we propose that negative emotion requirements fit with the challenge stressor definition.

If complexity is low, the exchanged information is simple and interactions are mostly instrumental, which means that interpersonal relations with customers are rather superficial. (Gutek, 1999; Gutek et al., 2000). Thus, the service usually focuses on information transfer and task execution rather than on compassionate social behaviours. In such contexts, negative emotion requirements are more likely to impede the attainment of service goals and unnecessarily complicate the interaction (cf. Dormann & Zapf, 2004). Employees may feel that they lose control of the interaction because dealing with unwanted but necessary negative emotions requires a high amount of emotion regulation that is usually not expected from them (Grandey et al., 2004). This fits with the concept of hindrance stressors. Thus, we expect:

Hypothesis 4a: Interaction complexity moderates the relationship between the occupation-level part of negative emotion requirements and personal accomplishment, such that the relationship is positive when interaction complexity is high and negative when interaction complexity is low.

Concerning emotion-rule dissonance, interaction complexity should have a different impact. In occupations with complex interactions, employees’ resources are heavily taxed by information processing and decision making during an interaction (Lilius, 2012), and any interruption can be seen as an obstacle to successful service delivery. Emotion-rule dissonance disturbs the interaction and confronts the employee with an unexpected additional task: the regulation of emotions (Zapf, 2002). Thus, employees may have fewer resources to cope with it and are likely to perceive emotion-rule dissonance as a threat to the attainment of their primary task, both of which corresponds to the characteristics of hindrance stressors (for health professionals, for example, see Xanthopoulou et al., 2013).

In little complex interactions, emotion rule-dissonance tends to be associated with positive emotion requirements that are not felt (cf. Humphrey et al., 2008; Zapf, 2002). Here, customer expectations are rather low and service goals can be easily accomplished by expressing the desired – but probably faked – emotion. In addition, employees should have more capacity for emotion regulation and therefore are more likely to see a challenge in emotion-rule dissonance that must be overcome to satisfy customer needs. As a result, feelings of success may arise after completing the service task (Lilius, 2012; Wong et al., 2017). Following this argumentation, emotion-rule dissonance should match with the definition of a challenge stressor.

Hypothesis 4b: Interaction complexity moderates the relationship between the occupation-level part of emotion-rule dissonance and personal accomplishment, such that the relationship is positive when interaction complexity is low and negative when interaction complexity is high.

Customization of service interactions

Service jobs in areas such as physical therapy, nursing, or teaching require that employees frequently adapt customer interactions in order to meet customer expectations or to achieve organizational goals. The service relation is individual and personal rather than impersonal and anonymous. Employees are usually empowered to express a wide range of emotions (cf. Morris & Feldman, 1996), depending on the situation and the customer, to accomplish their assigned work tasks. It is therefore likely that negative emotion requirements are perceived as a reasonable part of their professional role (cf. Humphrey et al., 2008). Whenever employees believe that negative emotions are necessary in a situation to attain a goal, they will meet the requirement without losing their authentic self-expression (e.g., Sutton, 1991). This is illustrated by the above-mentioned example of a reprimanding teacher, who will feel required to express negative emotions for good reason. Here, negative emotion requirements should correspond to the concept of challenge stressors.

By contrast, employees in service occupations, such as servers in a fast food restaurant, do not have to situationally personalize their behaviour, but rather follow a highly standardized script when dealing with customers (Bitner et al., 1994). We assume that employees in such occupations experience less individual reward, such as appreciation or personal recognition by the customer, because of their emotion work, as the interaction hardly allows for a variety of behaviour. Moreover, standardized interactions are more likely to occur in service jobs that appreciate the display of positive emotions (Humphrey et al., 2008). When a service provider, however, has to express negative emotions, this oftentimes indicates disproportionate customer expectations or customer misbehaviour (cf. Dormann & Zapf, 2004). For example, a flight attendant who has to shout at an unruly passenger is likely to judge the situation as something that goes beyond what should be expected from him or her. In this case, the requirement to express negative emotions is an unnecessary burden that complicates the work routine, suggesting that it is a hindrance stressor.

Hypothesis 5a: Customization of service interactions moderates the relationship between the occupation-level part of negative emotion requirements and personal accomplishment, such that the relationship is positive when interactions are highly customized and negative when interactions are highly standardized.

For emotion-rule dissonance, we expect a different moderation pattern. In occupations with individualized interactions, service outcomes are highly and simultaneously co-produced by the customer and the employee (Bettencourt et al., 2002; Mills & Margulies, 1980). The high degree of interdependence requires to establish and maintain trustful relationships with customers, and employees will reveal a lot about themselves to inspire confidence (Parasuraman et al., 1985; Surprenant & Solomon, 1987). Emotion-rule dissonance, however, jeopardizes trust, as service providers are no longer able to spontaneously respond to the highly variable customer needs and customers may easily detect inauthentic behaviour (Grandey et al., 2005). Thus, successful service delivery will be threatened by emotion rule dissonance, which is in line with the concept of hindrance stressors.

However, if interactions are highly standardized, the costs of being inauthentic should be lower. Employees can activate appropriate and well-practiced behavioural scripts to deal with dissonant emotion requirements (Bitner et al., 1994), and customers may reward authenticity less (cf. Grandey et al., 2005; Humphrey et al., 2008). By adhering to the display rule, employees avoid dissatisfying the customer and being disloyal to the organization, which should compensate for the disadvantages associated with the expression of genuine but deviant feelings (Yagil & Medler-Liraz, 2013). Thus, emotion rule-dissonance should be a manageable stressor associated with professionalism and the fulfilment of primary service tasks, which is in accordance with challenge stressor criteria. We therefore expect:

Hypothesis 5b: Customization of service interactions moderates the relationship between the occupation-level part of emotion-rule dissonance and personal accomplishment, such that the relationship is negative when interactions are highly customized and positive when interactions are highly standardized.

Non-substitutability of service providers

Non-substitutable types of service imply that a service provider knows about relevant background information and the history of a customer problem to accomplish the service successfully (e.g., lawyers, criminal detectives, and physicians). These types of service relationships are often more stable, durable, and the problem-solving process consists of several interactions (Grandey & Diamond, 2010; Gutek et al., 2000). The reason why negative emotion requirements in these occupations may be a challenge stressor is that customer interactions are more likely to provide opportunities to experience meaningful work and achieve personally rewarding goals (Grandey & Diamond, 2010). Additionally, rules for negative emotional displays should refer to socially accepted or role conforming norms (Sutton, 1991) rather than purely organizational policies and are therefore more likely to be accepted by a service provider. As in the case of criminal detectives in the study of Stenross and Kleinman (1989), negative emotion requirements are closely related to the service role and worth the effort of coping, suggesting that they are challenge stressors. Thus, feelings of success and proudness may result from successfully meeting the requirement.

In highly substitutable service occupations, however, customers are not dependent on a certain employee and may even not realize when being served by another person (e.g., a cashier in a supermarket or a call centre agent, cf. Gutek, 1999; Gutek et al., 2000). Cases in which negative emotions are required usually refer to an interruption of service delivery (e.g., customer incivility; Grandey et al., 2004) and threaten the achievement of employees’ primary task. A highly replaceable employee is likely to wonder why he or she, of all people, is confronted with the negative situation that prevents him or her from performing the actual task. Thus, overcoming difficult customer situations that require the expression of negative emotions may be less intrinsically motivating and rewarding, all of which are characteristics of hindrance stressors.

Hypothesis 6a: Substitutability of service providers moderates the relationship between the occupation-level part of negative emotion requirements and personal accomplishment, such that the relationship is positive when service providers are non-substitutable and negative when service providers are highly substitutable.

When a certain employee is critical for successful service delivery, he or she is required to build up and maintain a long-term relationship with customers (Bitner et al., 1994). This personal rapport might be threatened when employees’ inner feelings are conflicting with required ones resulting in a state of inauthenticity (Hochschild, 1983). As argued by Yagil and Medler-Liraz (2013), providing an authentic service, however, represents an important personal endeavour in such service contexts. Following this line of argumentation, emotion-rule dissonance might not comply with the concept of challenge stressors, as long as it dissents from an employee’s wish to provide an authentic service.

However, if a particular service provider is less important, service interactions are fleeting encounters, and emotional displays tend to be more superficial (Grandey & Diamond, 2010; Gutek et al., 2000). Chances are high that the same customer will not need to be served again in the future, so one’s authenticity should be less threatened by emotion-rule dissonance. For example, responding to customer complaints with a professional and probably faked smile is considered part of the daily routine, and dealing with it shows that one is doing a good job. By resolving emotion-rule dissonance with any type of emotion regulation, employees adhere to the display rules (Ashforth & Humphrey, 1993) and fulfil their organizational role (Humphrey et al., 2008), both of which should be motivating aspects in line with the concept of challenge stressors.

Hypothesis 6b: Substitutability of service providers moderates the relationship between the occupation-level part of emotion-rule dissonance and personal accomplishment, such that the relationship is negative when service providers are non-substitutable and positive when service providers are highly substitutable.

Identification with customers

The requirement to identify with customers is characterized by taking the customer’s perspective to better comprehend the situation and to be better able to solve the problem (Mills & Margulies, 1980). Whereas emergency physicians do not necessarily have to take the patients’ perspective, as they make decisions largely independently, hairdressers, in contrast, need to understand a customer’s target notions in order to act in the interest of the customer. Herein is acknowledged that, for example, complexity and identification are distinct variables, as a hairdressers’ occupation is low in complexity but high in identification requirements; the opposite pattern applies to a dentist’s occupation.

Ashforth and Humphrey (1993) suggested that identification plays an important role in the emotion work process and argued that it leads to a higher self-congruence when performing emotion work. Employees who identify with their customers, do their utmost to create an end result that meets customer needs, regardless of which emotional displays are required and whether they need to be faked or regulated. Even in difficult situations, when employees are required to express negative emotions or such emotions that are not automatically felt, they can uphold an authentic self-concept (i.e., consistency between behaviour and professional role; Humphrey et al., 2015) by faking or intensifying a desired emotion in “good faith” (Rafaeli & Sutton, 1987). For example, nurses sometimes need to express negative emotions to prevent a patient from behaving in a way that is harmful to the recovery process. This is resource-intensive, but is consistent with personal goals, as nurses usually consider the patient’s successful recovery as their mission. Due to the link to valued goals, negative emotion requirements and emotion-rule dissonance should therefore meet the concept of challenge stressors and show positive relationships with employee well-being.

Identification with customers is often less important in commercial or customer services (Humphrey et al., 2008). In these occupations, employees will only regard positive emotion requirements as legitimate. Any demand for emotion regulation resulting either from negative emotion requirements or from emotion-rule dissonance is likely to be perceived as an unnecessary obstacle preventing successful “fact-based” service delivery. Thus, negative emotion requirements and emotion-rule dissonance should correspond to the concept of hindrance stressors and show negative associations with personal accomplishment.

Hypothesis 7: Identification with customers moderates the relationship of the occupation-level part of negative emotion requirements (a) and emotion-rule dissonance (b) with personal accomplishment, such that the relationship is positive when service providers have to highly identify and negative when they hardly identify.

Method

The present study contains two components: a) individual data of service job holders on emotion work and well-being and b) expert ratings on the service interaction type of various service occupations. By employing a two-level approach, both components were used to test the hypotheses. The reason why we did not conduct a meta-analysis is that the use of individual-level raw data rather than aggregated data is considered superior. Errors and measures can be checked and the analyses are standardized, thus maximizing the statistical power (Cooper & Patall, 2009).

Sample and procedure

To gather raw data on emotion work variables, we looked for studies using the Frankfurt Emotion Work Scales (FEWS; Zapf, Mertini et al., 2003; Zapf et al., 1999) to measure them. In addition, only studies that included burnout measured by the Maslach Burnout Inventory (MBI; Maslach et al., 1996; Schaufeli et al., 1996) were considered. Doing so, the most common databases (Google Scholar and PsychINFO) were scanned for potential studies, using keywords such as emotion work, emotional labour, emotional (job/labour) demands/requirements, and Frankfurt Emotion Work Scales. In order to avoid publication bias, we also personally contacted researchers who have published studies using the FEWS and asked for unpublished data. Due to confidentiality obligations, however, not all of them were able to share their raw data with us.

The preliminary database was restricted according to the following criteria: first, the studies had to contain at least two items of at least one emotion work variable as well as both outcome variables. Thus, 10 studies were excluded from further consideration. Second, the items used to measure emotion work must have been unmodified, leading to a reduction by 3 studies. Third, the data had to include information on the occupation of the participants, which is why 2 convenience studies were dropped. Finally, the results of 33 studies involving 7,759 persons were considered, all of which were conducted in German-speaking countries. Results from 9 studies have already been (partially) published (see Appendix for the references).

Before integrating the data into one multi-level dataset, we checked each sample for implausible values and coding mistakes. Cases with missing job titles were excluded and reduced the total sample size to 7,106. Then, each sample was split up into occupation-homogenous and, wherever possible, into organization-homogenous subsamples. Each of these subsamples was treated as a separate between-level cluster, implying that some of the clusters entailed the same occupation, but were treated as separate clusters because they came from different studies. The reason why employees who had the same profession but worked in different organizations were assigned to different clusters is that we assumed a shared meaning of emotion work among employees who belong to a particular organization (see the “roles as perspective theory” by Meier & Semmer, 2018). In sum, we identified 78 occupation-level clusters. As unbalanced cluster sizes negatively affect the accuracy of parameter estimation especially when some clusters are too small (Cools et al., 2009), we decided to drop 6 clusters that comprised less than 10 people.

Next, we compiled the complete dataset, consisting of 7,075 persons belonging to 72 between-level clusters. Note that there are more clusters than occupations (n = 32), as we treated data from different studies as separate clusters to reduce unmodeled heterogeneity. The average cluster size was 96. Participants worked in a broad range of typical service jobs that involve a high amount of interactions with customers. Some samples included also persons (n = 1,040) who do not hold typical service jobs but have contact with external or internal customers on a frequent basis, such as management, commerce, or technicians. On average, participants were 39.06 years old (SD = 12.06). Fifty-nine percent were female. An overview of sample characteristics and measurements per study is depicted in Table 1.

Table 1. Summary of demographic sample characteristics and variations of measurements per study

No.	Author	Study name	Publication status	Occupations	No. of clusters	N	Mean age	SD age	Gender (% female)	FEWS version	MBI version	MBI response format
1	Banat (2000)	Travel agencies	U	Travel agents	1	94	34.73	11.01	75.5	1	1	7-point frequency
2	Beitler, Scherer, Trumpold, and Zapf (2015)	Convenience	U	Office clerks, flight attendants, customer care employees	3	37	33.16	10.58	75.7	2	2	7-point frequency
3	Benka (2017)	Railway	U	Customer service representatives and train drivers	2	244	37.90	10.63	18.0	2	2	6-point frequency
4	Bolm-Audorff and Zapf (2000)	Representative sample	P	Craftsmen, office clerks, salespersons, nurses, teachers, waiters and waitresses, technical engineers, administrative workers, commercial representatives	8	169	41.80	12.13	47.9	2	1	6-point frequency
5	Bolm-Audorff and Zapf (2005)	Railway	U	Dispatchers and train drivers	2	139	37.53	8.50	0.7	2	1	6-point frequency
6	Bolm-Audorff and Zapf (2012)	Physicians	U	Physicians	1	49	34.18	5.05	40.8	2	1	6-point frequency
7	Bolm-Audorff and Zapf (2012)	Geriatric care	U	Elderly care workers	1	16	36.31	12.02	81.3	2	1	7-point frequency
8	Bolm-Audorff and Zapf (2012)	Nursing care	U	Nurses	1	69	37.67	11.01	82.6	2	1	7-point frequency
9	Dormann, Dormann, Zapf, and Irmer (2004)	School administrations	P	School secretaries	1	157	45.84	9.16	100.0	2	1	7-point frequency
10	Egold (2000)	Shoe sales	U	Shoe sales employees	1	107	35.58	13.10	89.7	1	1	7-point frequency
11	Holz (2000)	Kindergartens	P	Kindergarten teachers, management employees, and maids	3	612	38.81	9.97	93.8	1	1	7-point frequency
12	Holz (2003)	Insurance company	U	Insurance brokers and agents	1	399	37.70	9.92	56.4	2	1	7-point frequency
13	Holz (2003)	Commerce company	U	Cashiers, warehouse keepers, sales assistants	1	81	42.65	11.52	67.9	2	1	7-point frequency
14	Holz (2003)	Youth and social welfare offices	U	Office clerks	1	556	41.62	9.80	66.4	2	1	7-point frequency
15	Holz (2003)	Hospital	U	Physicians, nurses, nursing and medical assistants	3	128	40.52	10.34	85.9	2	1	7-point frequency
16	Holz (2003)	Youth welfare	U	Social and educational workers	1	66	40.40	9.64	50.0	2	1	7-point frequency
17	Holz (2003)	Security company	U	Security guards and safety specialists	1	310	43.55	12.13	16.8	2	1	7-point frequency
18	Holz and Zapf (2003)	Outpatient psychiatry	U	Psychotherapists, nurses, caretakers	3	94	39.08	9.67	71.3	2	1	7-point frequency
19	Isic (1999)	Call centre	P	Call centre employees in banking, insurance, and airline industry	1	250	31.32	7.08	74.8	1	1	7-point frequency
20	Kern and colleagues (2019)	Hospitals	P	Surgeons	7	642	47.10	9.24	34.9	2	2	6-point frequency
21	Listmann (2001)	Call centres	U	Call centre agents	1	101	26.37	7.60	70.3	1	1	7-point frequency
22	Machowski and Zapf (2010)	Bank	U	Bank tellers, office clerks, commercial representatives	6	553	41.52	11.45	52.9	2	2	7-point frequency
23	Schmutte (1999)	Bank	P	Bank tellers	1	122	37.39	9.14	50.0	1	1	7-point frequency
24	Seifert (1999)	Hotels	P	Front desk clerks and waiters or waitresses	1	175	25.80	7.25	68.6	1	1	7-point frequency
25	Trumpold & Zapf (2012)	Airline (Germany)	U	Flight attendants	1	60	40.13	8.73	73.8	2	2	5-point agreement
26	Trumpold & Zapf (2012)	Airline (Switzerland)	U	Flight attendants	1	58	36.66	11.68	65.5	2	2	5-point agreement
27	Vogt (1999)	Handicapped children’s home	P	Educators, social care workers, nurses	1	83	37.88	10.13	79.5	1	1	7-point frequency
28	Von Gilsa and Zapf (2012)	Home loan bank	U	Bank tellers	1	177	45.32	10.07	53.1	2	2	7-point frequency
29	Von Gramatzik (2005)	Charter airline	U	Flight attendants	1	96	33.36	6.23	75.0	1	1	7-point frequency
30	Zapf, Elian, Weber, Keck, and Trumpold (2007)	Service work	U	Flight attendants and hairdressers	2	25	33.56	9.09	76.0	2	1	7-point frequency
31	Zapf, Isic, Bechtoldt, and Blau (2003)	Call centres	P	Inbound/ outbound call centre agents	8	360	31.87	10.45	66.9	2	1	6-point frequency
32	Zapf, Machowski, and Bernhardt (2012)	Bank	U	Bank tellers, commercial representatives, team leaders, and department leaders	4	992	42.79	10.25	55.2	2	2	7-point frequency
33	Zapf and Trumpold (2012)	Automotive supplier	U	Commercial representatives	1	54	45.00	8.74	68.5	2	2	7-point frequency

Note. N = 7,075. References for all samples are depicted in the Appendix. U = unpublished data; P = (partially) published data. FEWS version: 1 = Zapf et al. (1999); 2 = Zapf et al. (2003). MBI version: 1 = Büssing and Perrar (1992); 2 = Büssing and Glaser (1998). Data of the samples were collected between 1999 and 2019.

In the second part of our study, we conducted a survey in which experts were asked to rate the 32 service occupations that were in our total sample. The use of such ratings has been criticized for possible rater bias, which is a substantial source of error (Saal et al., 1980). In order to reduce systematic bias, it has been proposed to select people who are both trained in assessments and familiar with rating targets, and to combine the responses of multiple raters (Hoyt, 2000; Hoyt & Kerns, 1999). We therefore decided to invite only work and organizational psychologists who had expertise in assessing work characteristics and to use the aggregate results as between-level predictors. Out of 30 scientists and practitioners who were contacted personally or by email, two professors, four research associates, three human resource managers, and three consultants completed the questionnaire (N = 12). The mean age of the experts was 42.00 years (SD = 14.69) and 83% were female.

Within-level measures

Positive emotion requirements, negative emotion requirements, and emotion-rule dissonance were measured using the corresponding scales from the FEWS (Zapf, Mertini et al., 2003; Zapf et al., 1999). In the current form, the scales consist of five items. In some samples, however, an earlier FEWS-version was used (see Table 1). In these cases, we only used the items that correspond to the current version. Most items were scored on a 5-point scale that ranged from 1 = very rarely/never to 5 = very often (several times an hour). In both FEWS versions, an additional response format was used. One item of each scale used an “A versus B” format, asking participants to rate how similar one’s own workplace is compared to two opposite fictional work places. Reliability analyses revealed that the three scales had high internal consistencies. Omega coefficients (McDonald, 1999) are displayed in Table 2.

Table 2. Means, standard deviations, internal consistencies, and intercorrelations among study variables

		M	SD	1	2	3	4	5	6	7	8
1	Positive emotion requirements	3.44	0.98	(.86)	.26***	.39***	.12***	.12***
2	Negative emotion requirements	2.22	0.98	.08	(.86)	.35***	.19***	–.03
3	Emotion-rule dissonance	2.94	1.00	.70***	–.07	(.87)	.34***	–.09***
4	Emotional exhaustion	0.00	1.00	.06	.48***	–.03	—	–.24***
5	Personal accomplishment	0.00	0.99	.10	–.11	.07	–.22	—
6	Interaction complexity	2.78	0.97	.26**	.41***	–.19*	.32*	.05	—
7	Interaction customization	2.79	0.73	.23**	.45***	–.23*	.27*	–.06	.80***	—
8	Non-substitutability	2.18	0.95	.12	.52***	–.35***	.39***	.03	.84***	.85***	—
9	Identification with customers	2.87	0.74	.41***	.46***	.14	.34**	–.03	.62***	.75***	.71***

Note. Above the diagonal correlations at the individual-level are displayed. Below the diagonal correlations at the occupation-level are displayed. N_within = 7075; N_between = 72. Omega coefficients in parentheses.

* p < .05. ** p < .01. *** p < .001 (two-tailed).

Positive emotion requirements refer to the requirement of displaying pleasant emotions and were measured with five items. A sample item is: “How often do you have to display pleasant emotions towards customers in your job (e.g., friendliness or kindness)?”

Negative emotion requirements assess with five items the necessity of expressing and dealing with unpleasant emotions. A sample item is: “How often do you have to display unpleasant emotions to your customers?”

Emotion-rule dissonance refers to the frequency of situations in which emotions are required that do not correspond to the emotions typically felt in such situations. One example is: “How often in your job do you have to display emotions that do not agree with your true feelings?” Although the scale consists of five items, we used only four items and excluded the item using the A vs. B format. The reason for this is twofold: (1) the formulation of this item was different in the two FEWS versions; (2) a confirmatory factor analysis (CFA) indicated a factor loading of .37 for this item, which might be due to the reverse response format.

Emotional exhaustion and personal accomplishment were either measured by the MBI (Maslach et al., 1996) in the German validated version (Büssing & Perrar, 1992), or by the MBI – General Survey (Schaufeli et al., 1996) in the German version (Büssing & Glaser, 1998). An overview of which version was used in each sample is depicted in Table 1. In samples that used the human service questionnaire, emotional exhaustion was measured with nine items. An example is: “I feel emotionally drained from my work.” Personal accomplishment was measured with eight items. A sample item is: “I feel I’m positively influencing people’s lives through my work.” About two-thirds of the samples used a 7-point scale, however, some samples used a 6-point scale (see Table 1). In samples using the general survey, emotional exhaustion was measured with five items. One example is: “I feel fatigued when I get up in the morning and have to face another day on the job.” Personal accomplishment was measured with six items. An example is: “I have accomplished many worthwhile things in this job.” The response format varied across the samples. Out of nine samples using the general survey, five used a 7-point scale, two samples used a 6-point scale and two samples a 5-point scale, respectively. As the outcome variables were measured by different forms and to some extent with different response formats, we standardized the mean scale values across the samples using the same version and response format. In doing so, we ensured not to delete the between-level variation. We did not include depersonalization and cynicism in our study because they are distinct dimensions. While depersonalization is described as a “dysfunctional mode of coping with the emotional demands of service provision by distancing oneself emotionally from recipients”, cynicism reflects “indifference or a distant attitude towards work” in general and ignores social relationships with others (Leiter & Schaufeli, 1996, p. 231). This theoretical difference was empirically confirmed by Larsen et al. (2017), who found that depersonalization and cynicism are correlated, yet distinct, facets of burnout and that depersonalization had no unique relationships with work-related outcomes when statistically controlled for cynicism. As a corollary, considering both aspects in the same analysis could greatly bias our results. Thus, only emotional exhaustion was used as an indicator of strain, whereas well-being was operationalized by personal accomplishment.

Between-level measures

Adapted from Mills and Margulies (1980), we measured the type of interaction with four different dimensions. Experts were provided definitions and explanations for each dimension and were requested to rate each occupational group on each dimension with one item. In order to check the match of the expert’s ratings, we first analysed the maximum rating difference per occupation and dimension and found that more than two thirds of the ratings differed by a maximum of two scale points, which indicated a high inter-rater agreement. For the remaining third of cases with a lower agreement, we inspected which occupations caused the variation. Almost all of these cases concerned occupations that could only be specified in a broad sense (such as commercial representatives), so that there is likely to be a true variation in service interactions depending on the specific workplace. Second, inter-rater reliability for each dimension was measured by Kendall’s coefficient of concordance (Kendall’s W; Kendall, 1970; Kendall & Babington Smith, 1939), which is an extension of Cohen’s Kappa for multiple raters. Results indicated satisfactory consistency among the raters, ranging from W = .58 for interaction complexity to W = .64 for identification with customers. Examples of the highest and lowest rated occupations per dimension are presented in Figure 1.

Interaction complexity refers to both the complexity of information processes and the complexity of decision-making processes during service encounters. Experts rated the extent to which the solution of customer demands are linked to special technical or interdisciplinary competencies on a 5-point scale ranging from 1 – very little complex to 5 – highly complex. The following instruction was given: “If service tasks can be performed by unskilled or semi-skilled employees, service interactions are considered little complex. If special skills are required to provide the service, interactions with customers are highly complex”.

Interaction customization asked for the level of individualization of service encounters. This means the extent to which a service provider is required to adapt service interactions in order to meet customer expectations. We provided the following explanation for our experts: “If the process of service interaction is precisely specified and almost identical across job incumbents and situations, there is a high degree of standardization. If the service provider has to adjust working behaviour depending on the situation, the problems, or needs of the service recipient, service interactions are highly individualized. Both aspects form the poles of this dimension.” Responses were given on a 5-point scale that reached from 1 – standardized to 5 – individualized.

Non-substitutability assesses the extent of non-substitutability of a specific service employee. Experts were informed that low substitutable types of service imply that a service provider knows about relevant background information and that a long-term and intensive personal relationship is required to accomplish the service. The 5-point response format ranged from 1 – low substitutable to 5 – highly substitutable.

Identification with customers Experts rated the extent to which service providers are required to identify with a customer. They were instructed that identification means if service providers must redefine a customer’s aim as one’s own aim and must take the perspective of the service recipient. Responses were given on a 5-point scale ranging from 1 – low identification to 5 – very high identification.

Statistical analysis

To capture the hierarchical structure of our data (i.e., individual measurements nested in job groups and samples), we performed multilevel structural equation modelling (MSEM) using Mplus 8.4 (Muthén & Muthén, 1998–2017). Each emotion work variable was treated as a latent variable with raw items as indicators. We initially tested whether the measurement models can be built on both levels of analysis for each stressor and whether these multilevel confirmatory factor analyses (MCFA) fit the data well. While, as expected, the factor loadings were positive and significant at both levels, the fit indices did not yield an acceptable fit. The modification indices showed that for each construct there was one highly significant residual covariance that might indicate either another latent factor or item similarities unrelated to the latent factor. To exclude any model misspecification, we therefore performed both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) to determine a possible second factor or higher-order factor. However, neither the EFA nor the CFA results provided evidence for the existence of a different factor solution. Thus, we continued with modelling single-factor MCFAs for all three emotion work variables. Next, we inspected the items causing the significant residual covariances and found that they relate to items with similar wording, which is probably the reason why they are correlated beyond the latent construct (measurement error). We therefore set free one residual covariance for each stressor at both levels. This significantly improved model fit: the results now revealed an acceptable fit for each emotion work variable (see Table 3), given the large sample size at the within-level of analysis (Fan et al., 1999; Gerbing & Anderson, 1992).

Table 3. Results of MSEM models predicting emotional exhaustion and personal accomplishment

		Emotional exhaustion		Personal accomplishment
No.	Parameter	Estimate	SE	Estimate	SE	χ^2	df	χ^2/df	p	RMSEA	SRMRw	SRMRb	CFI
Individual-level
1	Positive emotion requirements	.15***	.02	.12***	.03	95.13	31	3.07	.000	.017	.022	.075	.981
2	Negative emotion requirements	.22***	.03	–.03	.02	103.97	31	3.35	.000	.018	.018	.101	.985
3	Emotion-rule dissonance	.36***	.02	–.10***	.02	112.86	20	5.64	.000	.026	.016	.088	.986
Occupation-level
1	Positive emotion requirements	.02	.03	.05	.06
2	Negative emotion requirements	.09**	.03	–.14*	.08
3	Emotion-rule dissonance	–.01	.03	.04	.06
	Positive emotion requirements x
4	interaction complexity			.00	.06	141.12	36	3.92	.000	.020	.022	.082	.972
5	interaction customization			–.03	.15	155.10	36	4.31	.000	.022	.022	.088	.970
6	non-substitutability			–.11	.10	155.98	36	4.33	.000	.022	.022	.093	.970
7	identification with customers			–.06	.13	115.27	36	3.20	.000	.018	.022	.074	.979
	Negative emotion requirements x
8	interaction complexity			.27***	.06	113.16	36	3.14	.000	.017	.018	.103	.985
9	interaction customization			.36***	.12	115.76	36	3.22	.000	.018	.018	.101	.979
10	non-substitutability			.26***	.07	117.21	36	3.26	.000	.018	.018	.104	.984
11	identification with customers			.45***	.13	114.40	36	3.18	.000	.018	.018	.105	.984
	Emotion-rule dissonance x
12	interaction complexity			–.16	.12	132.37	24	5.52	.000	.025	.016	.094	.985
13	interaction customization			–.45***	.10	135.13	24	5.63	.000	.026	.016	.093	.984
14	non-substitutability			–.28***	.08	129.99	24	5.42	.000	.025	.016	.097	.985
15	identification with customers			.29*	.14	133.08	24	5.55	.000	.025	.016	.092	.985

Note. No. = Number of the analysed model. Models 1–3 excluded occupation-level predictors, models 4–15 included an occupation-level predictor and an interaction effect. All models contained both emotional exhaustion and personal accomplishment as outcome variables. SRMR_w for the individual-level, SRMR_b for the occupation-level. Unstandardized coefficients are displayed.

* p < .05. ** p < .01. *** p < .001 (one-tailed).

Following Hofmann and Gavin (1998), between-level predictor variables (e.g., interaction customization) were centred to the grand-mean to gain unbiased estimates of the relationships. Within-level predictor variables (emotion work variables) were not centred. Before proceeding to test the hypotheses, we ran null models to examine intraclass coefficients Type I (ICC1; Hox, 2002). Analyses revealed an ICC1 of .05 for emotional exhaustion and an ICC1 of .20 for personal accomplishment, indicating that 5% of the variance in emotional exhaustion and 20% of the variance in personal accomplishment can be explained by the between-level clusters.

To test our hypotheses, we analysed a series of models using MSEM with robust maximum likelihood estimation. First, the main effects of each stressor on emotional exhaustion and personal accomplishment were tested (Models 1–3). Then, we applied the latent moderated structural equation method (LMS; Klein & Moosbrugger, 2000) to analyse the moderation hypotheses and followed the recommendations by Preacher et al. (2016). However, as this technique requires a computation-intensive numerical integration, each model contained only one emotion work variable, one interaction type dimension, and both outcome variables (Models 4–15). However, the number of integration points was still too high. Therefore, we used the proposed Monte Carlo integration with 3.000–4.000 integration points in the Mplus software (Asparouhov & Muthén, 2012). To interpret the moderations, simple slopes were plotted covering a value spectrum of +/ – 2SD.

Model fit was assessed using the chi-square goodness-of-fit test and the following four descriptive measures: a) the Root Mean Square Error of Approximation (RMSEA), the Standardized Root Mean Square Residual (SRMR) for the within- and between-level, and c) the Comparative Fit Index (CFI). Schermelleh-Engel et al. (2003) recommended values less than .05 for RMSEA, as well as for SRMR_within, and values of .97 for CFI to indicate a good fit of the data. However, fit statistics could not be obtained for the MSEM models analysing the interaction effects (i.e., non-linear terms). Thus, model fit was approximated as follows: (1) we analysed corresponding models without the latent interaction and report their fit statistics (Table 3); (2) we computed the models with the interaction and compared all parameter estimates for any significant changes. If there were no differences, we considered the full model to fit well. With regard to the between-level of analysis, an evaluation of models with descriptive fit indices is less accurate, as sample size (n = 72) is rather low. We followed the recommendations by Asparouhov and Muthén (2018), who suggested that the SRMR_between with values higher than .08 can still indicate a good fit. Means, standard deviations, internal consistencies, and intercorrelations among the study variables are presented in Table 2.

Results

Effects of emotion work requirements on well-being

To test Hypotheses 1–3, we performed MSEM, containing one emotion work requirement and both outcomes in one model. Fit indices, especially the descriptive fit indices, indicated acceptable fits to the data (Table 3). Hypothesis 1 predicted that positive emotion requirements are positively related with emotional exhaustion and personal accomplishment. As shown in Table 3, individual-level positive emotion requirements were positively linked to both emotional exhaustion and to personal accomplishment, thus supporting Hypothesis 1.

Hypothesis 2 assumed a positive relationship of negative emotion requirements with emotional exhaustion at the individual-level of analysis. Results revealed that negative emotion requirements were significantly related to emotional exhaustion (Table 3). Thus, Hypothesis 2 found support in the data. Moreover, the results indicated a null relationship between negative emotion requirements and personal accomplishment.

Regarding emotion-rule dissonance, Hypothesis 3 proposed a positive individual-level relationship with emotional exhaustion. This assumption was supported by the data (see Table 3). Furthermore, there was a small negative and significant relation to personal accomplishment.

Moderating effects of service interaction type

In a series of models, we examined the influence of service interaction type on the link between emotion work stressors and personal accomplishment. In each model, one aspect of service interaction type was added as an occupation-level predictor and a latent interaction term was built at the between-level. Model fit indices illustrated in Table 3 indicate overall acceptable fits to the data. Note that the chi-square values must be interpreted taking into account the large sample size at the within-level (Fan et al., 1999; Gerbing & Anderson, 1992). The results for all interaction effects are presented in Table 3.

Hypothesis 4 proposed that the relationships between negative emotion requirements, emotion-rule dissonance, and personal accomplishment are influenced by the complexity of service interactions. Consistent with Hypothesis 4a, the interaction between negative emotion requirements and complexity was significant. Simple slopes depicted in Figure 3 show that negative emotion requirements were positively related to personal accomplishment when employees had extremely complex interactions with customers. If interaction complexity was rather low, the relationship was strongly negative. Thus, Hypothesis 4a was supported by the data. Hypothesis 4b postulated a stronger positive relationship between emotion-rule dissonance and personal accomplishment when service interactions are little complex. Contrary to this assumption, the interaction was not significant. However, the coefficient direction was as expected (cf. Figure 4). Overall, Hypothesis 4b was not supported, but the results showed a confirming trend.

Figure 3. Illustration of simple slopes for negative emotion requirements and customer interaction characteristics with values +/ – 1 SD and +/ – 2 SD. ^a This coefficient is greater than 1.96 times its standard error. It would therefore be significant in a two-tailed test. However, we do not declare this coefficient as significant because the slope is contrary to our hypothesis, which was tested with a one-tailed test.

* p < .05. ** p < .01. *** p < .001 (one-tailed).

Figure 4. Illustration of simple slopes for emotion-rule dissonance and customer interaction characteristics with values +/ – 1 SD and +/ – 2 SD.

* p < .05. ** p < .01. *** p < .001 (one-tailed).

Hypothesis 5 proposed that the degree of customization of service interactions moderates the relationships of negative emotion requirements and emotion-rule dissonance with personal accomplishment. With respect to negative emotion requirements, the results revealed a significant interaction. The simple slopes plotted in Figure 3 illustrate that there was a negative significant relationship when service interactions were standardized. When service employees were very highly (+ 2SD) required to customize interactions, the relationship with personal accomplishment tended towards zero. This result partially supports Hypothesis 5. For emotion-rule dissonance, an interaction effect was found as well. In confirmation of Hypothesis 5b, the relationship between emotion-rule dissonance and personal accomplishment was negative when interactions were highly customized. However, when interactions were little customized, there was a positive relationship between emotion-rule dissonance and personal accomplishment (Figure 4).

Hypothesis 6 predicted moderation effects for the relationships of negative emotion requirements and emotion-rule dissonance with personal accomplishment by non-substitutability of service providers. Supporting Hypothesis 6a, we found a significant interaction for negative emotion requirements. The simple slopes showed no significant relationship when service employees were not easily substitutable (Figure 3). When service providers were highly substitutable, however, there was a strong negative relationship. This result partially supports our expectations, as we assumed a positive relationship under the high or very high moderator conditions. Regarding emotion-rule dissonance, a positive relationship with personal accomplishment was expected when service employees were easily substitutable. In line with Hypothesis 6b, the interaction term was significant and the simple slopes showed a positive relationship with personal accomplishment when employees were substitutable (Figure 4). By contrast, the relationship was negative and significant when employees were not substitutable.

In Hypothesis 7, it was expected that negative emotion requirements and emotion-rule dissonance are positively related to personal accomplishment when service providers are required to identify with customers. Regarding negative emotion requirements, the interaction term was significant. Simple slopes plotted in Figure 3 indicated a positive and significant relationship between negative emotion requirements and personal accomplishment when employees had to strongly identify with customers (+2SD). When service providers were not required to identify, however, there was a strong negative and significant relationship. This finding fully supports Hypothesis 7a. In support of Hypothesis 7b, the interaction term was also significant for emotion rule-dissonance. Simple slopes showed a high positive relationship when service providers were required to identify with customers, whereas the relationship was negative but not significant when they were not required to do so (Figure 4).

Discussion

Despite the abundance of emotion work research, there is still considerable uncertainty about when and why emotion work is harmful and/or beneficial. Recent literature indicates that occupational differences may play a role (e.g., Bhave & Glomb, 2016; Singh & Glavin, 2017). Therefore, the main objective of this study was to develop a systematic approach to categorizing service occupations in order to unravel the ambivalent relationships with well-being. Building upon the challenge-hindrance stressor framework (Crawford et al., 2010; LePine et al., 2005), we argued that positive emotion requirements generally refer to the concept of challenge stressors, while negative emotion requirements and emotion-rule dissonance are only challenge stressors for employees in specific service environments.

Our results on the relationships with emotional exhaustion and personal accomplishment at the individual level mirror established findings from both primary studies (e.g., Zapf & Holz, 2006) and meta-analyses (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013). On the one hand, requirements to show positive and negative emotions as well as emotion-rule dissonance were positively related to emotional exhaustion, upholding the notion that they are stressful. On the other hand, we found that positive emotion requirements were positively related to personal accomplishment, which is consistent with the definition of a challenge stressor. Negative emotion requirements, however, showed no relationship and emotion-rule dissonance showed a small negative relationship with personal accomplishment, thus backing a search for moderators. We used four dimensions that characterize service interactions in order to explain under which conditions negative emotion requirements and emotion-rule dissonance match with challenge and hindrance stressor definitions. The results mostly confirmed our hypotheses and are discussed in detail below.

On an exploratory basis, we tested the role of service interaction type also for positive emotion requirements, but the analyses yielded no significant moderation effects. Thus, positive emotion requirements seem to be a challenge stressor for almost all service providers. An explanation might be that the expression of positive emotions is commonly required in service jobs, conforms to normative expectations (cf. Semmer et al., 2019), and therefore represents a legitimate aspect. By displaying positive emotions, employees build up and maintain pleasant and trustful service interactions and thus facilitate successful service delivery. Customers who are treated friendly are also likely to react with positive emotions in return. Thus, employees may be rewarded for the display of positive emotions by customers, which fits seamlessly with the definition of challenge stressors. This finding confirms previous research (Kammeyer-Mueller et al., 2013) and highlights the notion that employees can enjoy performing emotion work despite the psychological costs (Humphrey et al., 2015).

With regard to negative emotion requirements, the analyses showed a hindrance stressor relationship pattern when employees are highly substitutable, work in service jobs that are characterized by low customized and little complex interactions with customers, and do not have to identify with them. However, when employees work in occupations that are characterized by highly complex interactions and are highly required to identify with their customers, there was a positive relationship with personal accomplishment. In contrast to our hypotheses, there were no significant positive relationships under the high (or very high) moderator conditions of customization and non-substitutability. Nonetheless, coefficients tended to go in the expected direction. Taken together, although mostly being a hindrance stressor, our results support the assumption that negative emotion requirements can also represent a challenge stressor in certain service contexts. There are several explanations as to why we did not find consistently positive relationships with personal accomplishment as expected. First, negative emotion requirements seem to fulfil challenge stressor criteria only in particular cases. We argued that it should apply to occupations like police officers, lawyers, team leaders, teachers or comparable ones in which negative emotions are expressed in a deliberate manner to accomplish intended goals (Zapf, Isic et al., 2003). Given the variety of service occupations existing in our societies, such occupations do not represent the majority. This is also reflected by the restricted spectrum of these occupations in our sample (cf. Table 1). Thus, challenge effects of negative emotion requirements might be limited to these few cases.

A second reason for the lacking positive effects might be a suppressor effect. It has been repeatedly stated that the positive impact of challenge stressors may be masked by the strain-related part common to all stressors (Boswell et al., 2004; Widmer et al., 2012). Thus, strain has to be statistically controlled in order to unveil (higher) positive relationships. Unfortunately, we could not apply this technique to our data, as emotional exhaustion had to be treated as a manifest variable in the multilevel models. When using a maximum likelihood estimation, observed variables measured at the individual-level cannot be analysed as predictors at the between-level of analysis. A latent measurement model would avoid this problem and should therefore be implemented in future research.

Finally, emotion-rule dissonance was positively related to personal accomplishment when non-substitutability and customization is low and when employees have to strongly identify with their customers. Although the results suggested a trend, no significant effect occurred for complexity. When service providers can easily be substituted and customer interactions follow standardized scripts, the expectations to be authentic are low. Employees in such service occupations typically have a larger number of customers, and interactions with customers have a shorter duration, so there is usually no personal relationship. Being positive and, if necessary, showing positive emotions that are not felt may not be as good as displaying true positive emotions, but can be sufficient and a sign of professionalism in managing the relations. As there are interactions with many customers, it is unrealistic that natural positive emotions come up all the time. “Stay calm and smile” can then represent a strategy that ensures a straightforward service delivery, even when an employee is exposed to an unpleasant state of tension during the interaction. This explanation matches with the results of Pugh et al. (2011) who found that surface acting is less harmful when it is negligible to express authentic emotions. However, the outlined strategy should only work if there is no personal rapport between service provider and customer that could be threatened by the display of superficial emotions. If service encounters are highly customized and the service provider is important for successful service delivery, there is probably a relationship where trust and honesty play a role and authentic emotions are expected. It is unlikely that emotions are deliberately faked, and if they were, this would not be seen as contributing to central goals of one’s own work, but rather as compromising service delivery. Thus, emotion-rule dissonance is a hindrance stressor in such occupations.

In occupations such as nursing, elderly care, or psychotherapy, employees usually identify with their customers. They are likely to redefine customer needs as a personal goal and are thus more poised to engage in emotion regulation (Gutek et al., 2000). Thus, emotion-rule dissonance showed a challenge stressor relationship pattern in these occupations. These results confirm the belief of Ashforth and Humphrey (1993) that identification plays a fundamental role in understanding the consequences of emotion work.

Some aspects of our results are noteworthy, even though they do not pertain to the study hypotheses. First, the analysed dimensions of service interactions were highly interrelated (see Table 2). This oftentimes suggests building a higher-order factor. However, research has found that different occupational groups are characterized by very specific patterns in service interaction type (e.g., Humphrey et al., 2008; Mills & Margulies, 1980), which casts doubt on a higher-order construct. Moreover, some dimensions showed completely opposite moderator effects and therefore provided a unique explanation to the ambivalent relationships with well-being. From a methodological point of view, the correlations may have been inflated due to the small sample size at the between-level of analysis (Grewal et al., 2004) and the fact that some occupations (and thus the predictor values) appeared sometimes more than once. For these reasons, we recommend that prospective studies examine service interaction type as specific dimensions.

Second, zero-order correlations of emotion work requirements with service interaction type revealed some interesting findings. Positive and negative emotion requirements showed positive relationships with interaction dimensions. However, the relationships were significantly higher for negative emotion requirements. This suggests that the expression of negative emotions is more often required in occupations in which services are highly complex and individualized, service providers are less substitutable, and in which employees are required to identify with customers. By contrast, emotion-rule dissonance is negatively related to three out of four interaction dimensions. Thus, when interactions represent a very central part of a service occupation, emotion-rule dissonance seems to less frequently occur.

Limitations and future research directions

As in most empirical studies, some limitations should be considered. First, the job clusters compiled may entail some inaccuracies. Some of the bigger samples used for our analysis provided no nuanced information on employees’ specific occupation. For example, we could only distinguish between leaders, bank clerks, and commercial representatives in one bank sample. Thus, it is not entirely guaranteed whether all of them actually do the same type of emotion work. We addressed this problem by employing a maximum likelihood estimation, which offers some protection against unobserved heterogeneity (Hox et al., 2010). However, the true effect sizes of the interaction effects might be larger than we found in our analysis. Future studies should investigate service interaction type by considering more specific groups of employees. This also matches with research that accounts for the shared appraisal of emotion work within teams (e.g., Becker & Cropanzano, 2015; Hu & Shi, 2015).

Second, as is often the case when using raw data from numerous studies, not all measures were ideal. In particular, the response formats of emotional exhaustion and personal accomplishment varied across studies. For this reason, they were treated as manifest variables in the analysis. Even though we only created standardized mean values for all studies using the same response format and not for each study separately, this procedure slightly reduced the intraclass correlations. However, as we nonetheless found significant interaction effects, this confirms our hypotheses even more.

Third, due to the cross-sectional nature of the samples used, it is not possible to draw conclusions on the causality of the link between emotion work and well-being. Up to now, most research results are still based on cross-sectional self-report data. As argued by Hülsheger et al. (2010), well-being or job satisfaction might also be an antecedent of emotion work. While they only provided evidence for causal pathways from emotion work on emotional exhaustion and performance, a study by Philipp and Schüpbach (2010) showed reciprocal relationships between emotional exhaustion and emotion work. Therefore, more longitudinal studies are needed to shed light upon the effect direction.

Based on the findings of this study, it is recommended that future studies should not use higher-order measures of emotion work requirements which disregard the quality of emotions (cf. Côté, 2005). In line with previous findings (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013; e.g., Zapf & Holz, 2006), we demonstrated that the analysis of qualitatively different emotion work variables provides deeper insights into service work and related consequences than the examination of higher-order emotion work requirements. For the future, it might also be useful to examine other emotion work requirements that characterize certain service jobs, such as sensitivity requirementsthat are, for instance, central for nurses or kindergarten teachers (Zapf & Holz, 2006). Furthermore, scholarship would benefit from the investigation of temporal relationships between service providers and customers, which could not be implemented in our study. As argued by Grandey and Diamond (2010), such differences between service providers cannot be assumed by considering the type of occupation alone. Thus, future research is needed that considers certain dyadic relations between service employees and customers.

Research implications

This study has implications not only for scholarship on emotion work but for stress research that seeks to understand the influence of the work context. Regarding emotion work, our study enriches research that follows a job design perspective to learn more about under what conditions interactions with customers are either beneficial or dysfunctional and draining to the employee. Our results complement Grandey and Diamond (2010) theoretical postulations that an occupational perspective should be integrated into research on emotion work. We add to findings on the link between emotion work and job satisfaction (Bhave & Glomb, 2016) or occupational health risks (Singh & Glavin, 2017) by focusing on specific stressors and testing a taxonomy of service job characteristics in a large data set. In line with the suggestions of Johns (2006), the results of our study indicate that occupational aspects matter when predicting positive or negative consequences of emotion work and that considering a taxonomy of service jobs is a fruitful direction. The investigation of emotion work at the occupational level may help in future research to learn more about contradicting findings in the literature and, for example, to broaden our knowledge on the circumstance under which deep acting or surface acting as strategies to solve emotion-rule dissonance are rather beneficial or harmful. The role of a challenge appraisal has already been examined by Huang et al. (2015); however, as they analysed data from call centre agents, it is unclear whether the results can be generalized to other occupations. Therefore, we encourage further studies to include the occupational perspective to the examination of emotion work.

Regarding literature on occupational stress, this is one of the first studies applying the challenge-hindrance stressor framework to a job characteristics approach of emotion work. We extend the knowledge on challenge stressors by suggesting a criteria-based classification and argue that, in certain cases, boundary conditions must be considered so that the challenge stressor definition applies. This is consistent with recent work that identified aspects like the specific nature and quality of the stressor (Chong et al., 2011), stressor exposure time (Baethge et al., 2018), or illegitimate tasks (Schmitt et al., 2015) as moderators. Our findings underpin the view that stressors can be functional for attaining goals and promote well-being if they meet the challenge stressor definition (Kern et al., 2020; Van den Broeck et al., 2010).

Implications for practice

Probably the most important takeaway for practitioners is that the type of service occupation plays a crucial role in predicting effects on employee well-being. We presented a framework for service jobs that puts the interaction between service providers and customers in the centre and showed that health-related effects are oftentimes dependent on interaction characteristics. The often-held proposition that service providers suffer from being exposed to high emotion work requirements in order to comply with organizational display rules needs to be carefully reconsidered. Our results indicated that emotion work may be associated with positive outcomes if they represent a central and legitimate part of one’s job (cf. Semmer et al., 2007) and are typically appraised as a challenge. We showed that positive emotion requirements have favourable effects for service employees, as they are automatically expressed in most interactions (Martínez-Iñigo et al., 2007), satisfy affiliation needs (Zapf, 2002), and are related to goal accomplishment. For negative emotion requirements or emotion-rule dissonance, the results demonstrate that the challenge effect is more likely to unfold when service providers are required to identify with their customers. Thus, managers are advised to provide a rationale for the engagement in more difficult customer interactions and should invest in training and communication that emphasizes the meaningfulness of displayed emotions on customer satisfaction and loyalty (e.g., Pugh, 2001). Employees who better understand the role of emotion work for the success of a service organization are more likely to appraise emotion work requirements as justified (see also Humphrey et al., 2015). We argued that legitimacy and role-conformity are important preconditions for a challenge stressor (see also Schmitt et al., 2015 for the challenge stressor time pressure). Even employees who have to endure a discrepancy between required and felt emotions in order to achieve goals, such as maintaining good social relationships, experience a feeling of pride after successfully overcoming the situation if this precondition applies.

Yet emotion work also exhausts employee resources and is consistently associated with deleterious effects for individual health (Hülsheger & Schewe, 2011; Kammeyer-Mueller et al., 2013). Our results clearly showed that emotion work variables are generally linked to emotional exhaustion. In order to inhibit potential harm, managers should ensure that employees have sufficient time for recovery (Sonnentag et al., 2010). For some occupations requiring a very high amount of emotion work such as complaint management, managers should prevent the fatigue and performance erosion of their employees by providing opportunities to seclude oneself to recharge emotional resources.

Conclusion

The present paper is an answer to the call for research on emotion work that considers an occupational perspective, and presents a theory-driven framework for health-related predictions of qualitatively different emotion work stressors. While past research has predominantly framed emotion work in a negative light, inconsistent findings on the impact on well-being necessitate a more elaborate picture in the context of organizational and occupational boundary conditions. Accounting for service interaction type as one aspect that characterizes service jobs, the present study contributes to this more recent research trend and provides a first but not sufficient explanation that should be retested in prospective studies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Appendix

Primary studies included in the present article

Table

No.	Reference
1	Banat, J. (2000). Der Einfluss von interaktionsbezogenen Arbeitsmerkmalen und Kundenorientierung auf Burnout in indirekt personenbezogenen Dienstleistungstätigkeiten [The impact of interaction-related work characteristics and customer orientation on burnout in indirect person-related service jobs] (Unpublished diploma thesis). Goethe University, Frankfurt/Main.
2	Beitler, L. A., Scherer, S., Trumpold, K., & Zapf, D. (2015). Emotionen in der Dienstleistungsarbeit [Emotions in service work] (Technical Report). Goethe University, Frankfurt/Main.
3	Benka, J. (2017). Emotion-rule Dissonanz als Challenge Stressor – Eine Befragung von Kundenbetreuern im Nahverkehr und Triebfahrzeugführern [Emotion-rule dissonance as a challenge stressor – A survey of customer service representatives and train drivers] (Unpublished master thesis). Goethe University, Frankfurt/Main.
4	Bolm-Audorff, U. & Zapf, D. (2000). Psychische Arbeitsbelastungen und Gesundheit deutscher Arbeitnehmerinnen und Arbeitnehmer. [Job stressors and health of German employees] (Unpublished research report). Goethe University, Frankfurt/Main. Parts of the data published in Zapf, D., Isic, A., Bechtoldt, M., & Blau, P. (2003). What is typical for call center jobs? Job characteristics, and service interactions in different call centers. European Journal of Work and Organizational Psychology, 12(4), 311–340. https://doi.org/10.1080/13594320344000183
5	Bolm-Audorff, U., & Zapf, D. (2005). Psychische Arbeitsbelastungen und Gesundheit bei Bahnarbeitern. [Job stressors and health of railway employees] (Technical report). Goethe University, Frankfurt/Main.
6	Bolm-Audorff, U. & Zapf, D. (2012). Psychische Arbeitsbelastungen und Gesundheit von Ärzten [Job stressors and health of physicians] (Technical Report). Goethe University, Frankfurt/Main.
7–8	Bolm-Audorff, U. & Zapf, D. (2012). Psychische Arbeitsbelastungen und Gesundheit von Alten- und Krankenpflegern [Job stressors and health of elderly care workers and nurses] (Technical Report). Goethe University, Frankfurt/Main.
9	Dormann, T., Dormann, C., Zapf, D., & Irmer, C. (2004). Belastungen im Schulsekretariat. Aufgaben und Probleme an einer pädagogischen Schnittstelle [Stressors at school secretariats. Work tasks and problems at an educational interface]. Die Schule, 96, 345-358.
10	Egold, N. (2000). Zur Messung von Kundenorientierung bei Dienstleistern und der Zufriedenheit von Kunden - Validierung eines Instrumentes zur Erfassung der persönlichen Kundenorientierung von Dienstleistern [Measuring customer orientation of service providers and customer satisfaction - Validation of an instrument] (Unpublished diploma thesis). Goethe University, Frankfurt/Main.
11	Holz, M. (2000). Emotionsarbeit in Kindergartenstätten [Emotion work in kindergartens] (Unpublished diploma thesis). Goethe University, Frankfurt/Main. Data published in: Zapf, D., Seifert, C., Schmutte, B., Mertini, H., & Holz, M. (2001). Emotion work and job stressors and their effects on burnout. Psychology & Health, 16(5), 527–545. https://doi.org/10.1080/08870440108405525 And in: Zapf, D., & Holz, M. (2006). On the positive and negative effects of emotion work in organizations. European Journal of Work and Organizational Psychology, 15(1), 1–28. https://doi.org/10.1080/13594320500412199
12–17	Holz, M. (2003). Soziale Stressoren in Organisationen [Social stressors in organizations] (Unpublished dissertation). Goethe University, Frankfurt/Main.
18	Holz, M & Zapf, D. (2006). Emotionsarbeit in der ambulanten Psychiatrie [Emotion work in an outpatient psychiatry] (Technical report). Goethe University, Frankfurt/Main.
19	Isic, A. (1999). Psychische Anforderungen in Dienstleistungsberufen am Beispiel Call Center [Stressors and resources of call-center jobs] (Unpublished diploma thesis). Goethe University, Frankfurt/Main. Parts of the data published in: Isic, A., Dormann, C., & Zapf, D (1999). Belastungen und Ressourcen bei Call-Center-Arbeitsplätzen [Stressors and resources at the workplace of call-center agents]. Zeitschrift für Arbeitswissenschaft, 53(3), 202-208. Parts of the data published in: Zapf, D., Seifert, C., Schmutte, B., Mertini, H., & Holz, M. (2001). Emotion work and job stressors and their effects on burnout. Psychology & Health, 16(5), 527–545. https://doi.org/10.1080/08870440108405525 And parts of the data published in: Zapf, D., & Holz, M. (2006). On the positive and negative effects of emotion work in organizations. European Journal of Work and Organizational Psychology, 15(1), 1–28. https://doi.org/10.1080/13594320500412199
20	Parts of the data published in: Kern, M., Buia, A., Tonus, C., Weigel, T. F., Dittmar, R., Hanisch, E., & Zapf, D. (2019). Psychological stressors, resources, and well-being of surgeons in Germany: A cross-sectional study. Der Chirurg, 90(7), 576-584. https://doi.org/10.1007/s00104-018-0780-5
21	Listmann, S. (2001). Mitarbeiterstudie in Call Centern im Hinblick auf die derzeitige Arbeitsplatzgestaltung und Arbeitssituation [Survey-feedback study in call centers with regard to workplace design and work situation] (Unpublished diploma thesis). Goethe University, Frankfurt/Main.
22	Machowski, S. & Zapf, D. (2010). Gefährdungsbeurteilung: Mitarbeiter-Befragung zu psychischen Belastungen und Ressourcen am Arbeitsplatz. [Risk analysis: Employee survey of job stressors and resources at work] (Unpublished research report). Goethe University, Frankfurt/Main.
23	Schmutte, B. (1999). Emotionsarbeit und ihre Konsequenzen bei Bankarbeitern [Emotion work and its consequences for bank employees] (Unpublished diploma thesis). Goethe University, Frankfurt/Main. Data published in: Zapf, D., Seifert, C., Schmutte, B., Mertini, H., & Holz, M. (2001). Emotion work and job stressors and their effects on burnout. Psychology & Health, 16(5), 527–545. https://doi.org/10.1080/08870440108405525 And in: Zapf, D., & Holz, M. (2006). On the positive and negative effects of emotion work in organizations. European Journal of Work and Organizational Psychology, 15(1), 1–28. https://doi.org/10.1080/13594320500412199
24	Seifert, C. (1999). Die Bedeutsamkeit der emotionalen Arbeit als Stressfaktor im Servicebereich und ihre Interaktion mit Entscheidungsspielräumen. [The significance of emotion work as stress factor in the service sector and its interaction with job control] (Unpublished diploma thesis). Goethe University, Frankfurt/Main. Data published in: Zapf, D., Seifert, C., Schmutte, B., Mertini, H., & Holz, M. (2001). Emotion work and job stressors and their effects on burnout. Psychology & Health, 16(5), 527–545. https://doi.org/10.1080/08870440108405525 And in: Zapf, D., & Holz, M. (2006). On the positive and negative effects of emotion work in organizations. European Journal of Work and Organizational Psychology, 15(1), 1–28. https://doi.org/10.1080/13594320500412199
25–26	Trumpold, K. & Zapf, D. (2012). Emotional labor in organizations. (Unpublished research report). Goethe University, Frankfurt/Main.
27	Vogt, C. (1999). Emotionale Anforderungen am Arbeitsplatz [Emotional labor requirements] (Unpublished diploma thesis). Goethe University, Frankfurt/Main. Data published in: Zapf, D., Seifert, C., Schmutte, B., Mertini, H., & Holz, M. (2001). Emotion work and job stressors and their effects on burnout. Psychology & Health, 16(5), 527–545. https://doi.org/10.1080/08870440108405525
28	Von Gilsa & Zapf, D. (2012). Gefährdungsbeurteilung: Mitarbeiter-Befragung zu psychischen Belastungen und Ressourcen am Arbeitsplatz. [Risk analysis: Employee survey of job stressors and resources at work] (Unpublished research report). Goethe University, Frankfurt/Main.
29	Von Gramatzik, J. (2005). Stressbewältigung im Kundenkontakt [Stress management in customer interactions] (Unpublished diploma thesis). Goethe University, Frankfurt/Main.
30	Zapf, D., Elian, C., Weber, A., Keck, A., & Trumpold, K. (2007). Emotionsarbeit in Dienstleistungsberufen – Tagebuchstudie [Emotion work in service jobs – A diary study] (Unpublished research report). Goethe University, Frankfurt/Main.
31	Zapf, D., Isic, A., Bechtoldt, M., & Blau, P. (2003). What is typical for call center jobs? Job characteristics and service interactions in different call centers. European Journal of Work and Organizational Psychology, 12(4), 311–340. https://doi.org/10.1080/13594320344000183
32	Zapf, D., Machowski, S., & Bernhardt, L. (2012). Gefährdungsbeurteilung: Mitarbeiter-Befragung zu psychischen Belastungen und Ressourcen am Arbeitsplatz. [Risk analysis: Employee survey of job stressors and resources at work] (Unpublished research report). Goethe University, Frankfurt/Main.
33	Zapf, D., & Trumpold, K. (2012). Ergebnisbericht der Analyse arbeitsplatzbezogener Ressourcen und Belastungen bei einem deutschen Automobilzulieferer [Results report of the analysis of job resources and job stressors at a German automotive supplier] (Unpublished research report). Goethe University, Frankfurt/Main.