Beyond demand-control : Emotional labour and symptoms of burnout in teachers

Teaching is a profession that involves a high level of emotional labour. This includes such behaviours as surface acting (displaying an emotion that is not actually felt), deep acting (the activity undertaken to actually feel a required emotion), and suppression of emotion. In many professions, this emotional labour is thought to be related to high levels of burnout. The aim of our study was to show that emotional labour has a unique relationship with burnout that is separate from its relationship with the variables of the Demand Control Support (DCS) model. Emotional labour was studied, together with the variables of the Karasek Job Demand Control Support model, in a random sample of 365 mathematics teachers in the Netherlands. We used the Dutch Questionnaire on Emotional Labor (D-QEL) that measures: (1) surface acting, (2) deep acting, (3) suppression, and (4) emotional consonance. In line with other studies, job characteristics were found to be specifically related to emotional exhaustion. Surface acting was significantly related to depersonalization, and emotional consonance (the absence of emotional labour) was related to personal accomplishment. We conclude that whereas the DCS model has been valuable for understanding emotional exhaustion, emotional labour provides an additional perspective for understanding work stress.

Keywords: burnout; teaching; work-related stress; job demands; job control; support; emotive effort; Emotional labour

Teaching is a profession that requires almost constant interaction with students. A teacher has to be enthusiastic and lively in order to catch and hold students' attention. Furthermore, teachers may adopt extreme ebullience when praising a student for a good answer or give an impression of calm confidence when confronted with a disruptive student. Furthermore, teachers are expected to ensure the orderly conduct of classes throughout the day, every working day. In order to perform these tasks adequately, teachers have to show or exaggerate some emotions (Ogbonna & Harris, [31]) and minimize or suppress the expression of other emotions (Ybema & Smulders, [41]). Teachers consider the faking of emotions to be stressful (Ogbonna & Harris, [31]), but the exact consequences for their job-related well-being following strict regulation of their emotions are not fully known.

Teaching is a profession with a high level of stress (Johnson, Cooper, Cartwright, Donald, Taylor, & Millet, [24]) and a high level of burnout (Brouwers & Tomic, [11]). Because emotional exhaustion is a key characteristic of the burnout syndrome (Maslach, Schaufeli, & Leiter, 2001), an emotional labour perspective may be useful in trying to understand the development of emotional exhaustion. Hochschild was the first to note that, especially in service jobs, employees are often required to show certain emotions in order to please the customer. Having to show such emotions while one is not actually feeling them, or having to suppress one's own emotions when their expression does not seem appropriate, were taken together in devising the concept of emotional labour (Hochschild, [23]). Hochschild introduced the term surface acting to refer to the display of the characteristics of an emotion that are regarded as appropriate, but are not actually felt. Deep acting describes the activity that is undertaken to actually feel an emotion that is thought to be required. Finally, emotional consonance captures the situation where somebody effortlessly feels the emotion that is required in a certain situation.

Hochschild already noted that the requirement to express emotions that are incompatible with experienced emotions would cause repeated stress (Hochschild, [23]). As repeated stress is thought to lead to reactions of psychological strain and physical illness (van der Doef & Maes, [39]), an increasing number of studies have explored the relationship between emotional labour and job stress or burnout. Several studies have reported significant relationships between emotional labour and the emotional exhaustion dimension of burnout (Brotheridge & Grandey, [7]; Brotheridge & Lee, [10]; Pugliesi, [35]; Zammuner & Lotto, [43]) and thus made the theoretical importance of this relationship clear. These studies did not, however, assess the relevance of this relationship, taking into consideration the importance of several well-known stressful work characteristics as job demands, job control, and social support (Karasek, [25]). The aim of the present study is to assess the relevance of an emotional labour perspective in studying burnout when the variables of Karasek's job characteristics model are taken into account.

The relationship between job characteristics and well-being is usually studied with the Demand Control Support (DCS) model, and this model has also been successfully applied to the study of job-related well-being such as emotional or work-related exhaustion and burnout (van der Doef & Maes, 1999). In general, there is now accumulated evidence that the three now-classical variables from this model (high job demands, low possibilities to regulate one's work, little support) each cause strain, and that their effect is cumulative (de Lange, Taris, Kompier, Houtman, & Bongers, [16]). For burnout, there is also substantial evidence of an interactive effect of demands and control (the strain hypothesis of the DCS model) and much less evidence for the iso-strain hypothesis, which specifies a negative effect of the combination of high demands, low control, and little social support (van der Doef & Maes, 1999).

The DCS model is thus useful for our understanding of burnout, but there is clear evidence that emotional labour also has significant relationships with burnout level (Zammuner & Galli, [42]). The relative importance of both emotional labour and the DCS model in predicting burnout level is, however, not known exactly. Pugliesi ([35]) found negative effects of emotional labour on job satisfaction that were independent of the effect of job demands and control. In another study the need to suppress emotions predicted emotional exhaustion, and lack of support from colleagues or supervisor had both an independent as well as a buffering effect (Ybema & Smulders, [41]). As the latter study did not include quantitative demands and control, the DCS model was only partially taken into account.

The aim of this study is to show that emotional labour has a unique relationship with burnout that is separate from its relationship with the variables from the DCS model. This line of reasoning results in several hypotheses. We expect to replicate the positive relationship between the emotional labour concept of surface acting and the burnout dimensions of emotional exhaustion and depersonalization (Brotheridge & Grandey, [7]; Grandey, [21]; Zammuner & Lotto, [43]):

• • Hypothesis 1a: surface acting will have a significant positive relationship with emotional exhaustion that is separate from its relationship with the variables of the DCS model.
• • Hypothesis 1b: surface acting will have a significant positive relationship with depersonalization that is separate from its relationship with the variables of the DCS model.

A positive relationship between deep acting and emotional exhaustion has been previously reported (Grandey, [21]), but this is not a consistent finding (Brotheridge & Grandey, [7]; Zammuner & Lotto, [43]). The effects of deep acting may vary, as successful deep acting will by definition result in emotional consonance. This also explains why a positive relationship between deep acting and personal accomplishment has been reported (Brotheridge & Grandey, [7]). In our study, no specific hypotheses for deep acting will be formulated.

In a comparative study of various professions, one type of emotional labour, the need to suppress emotions, was studied (Ybema & Smulders, [41]). The jobs of the more than 4,000 participants were first grouped into 40 categories. Police, firemen, and security officers indicated most strongly that they had to suppress emotions, but teachers experienced this quite often too. The need to hide emotions had a strong relationship with emotional exhaustion. This results in our second hypothesis:

• • Hypothesis 2a: suppression of emotions will have a significant positive relationship with emotional exhaustion that is separate from its relationship with the variables of the DCS model.
• • Hypothesis 2b: suppression of emotions will have a significant positive relationship with depersonalization that is separate from its relationship with the variables of the DCS model.

Emotional consonance is a dimension indicating that felt emotions do not call for the activation of regulatory processes (Zammuner & Galli, [42]). A high level of emotional consonance will indicate that a person effortlessly expresses emotions that are felt and that these emotions are at the same time required for the job. This should result in a heightened feeling of personal accomplishment and the absence of emotional exhaustion:

• • Hypothesis 3a: emotional consonance will have a significant negative relationship with depersonalization that is separate from its relationship with the variables of the DCS model.
• • Hypothesis 3b: emotional consonance will have a significant positive relationship with personal accomplishment that is separate from its relationship with the variables of the DCS model.

We decided to test these hypotheses in a sample of teachers.

The Dutch Association of Mathematics Teachers supported this project and provided means and assistance to send questionnaires to a sample of 1,000 that was randomly selected by a computer from their 2,000 members, all teachers in secondary schools. The total number of teachers of mathematics in the Netherlands is about 8,000. The response rate was 36.5%, and the final sample consisted of a representative sample of 269 men and 96 women, with a mean age of 48.8 years, SD=8.35. The mean number of years of experience in the profession is 21.37, SD=9.78, ranging from 1 to 42 years. The mean number of teaching hours per week is 18.69, with a range from 0 (administration) to 29, SD=6.77.

We assessed the prevalence of burnout symptoms using cut-off scores for the diagnosis of burnout for the Dutch population (Schaufeli & van Dierendonck, [37]). According to these criteria, emotional exhaustion was present in 23% of the sample, scores ≥20; depersonalization in 23.3%, scores ≥8; and low personal accomplishment in 53.4%, scores ≤28. Thirty teachers, 8.2% of the sample, met all three criteria, 17.5% met two, and finally 38.9% met one criterion.

The teachers received our questionnaire at their home address together with a letter of recommendation from the board of the Association. The questionnaire could be returned anonymously to our university in a postage prepaid envelope. After 3 weeks, all of the teachers received a letter reminding them of the survey.

We used the penultimate version of the questionnaire to measure emotional labour, since it has good psychometric properties (Briët, Näring, Brouwers, & van Droffelaar, [42]). This instrument consists of four scales, of which three measure the following types of emotional labour: surface acting, deep acting, and suppression. The fourth scale measures emotional consonance. A high level of emotional consonance will indicate that a person effortlessly expresses emotions that are felt, and we regard this as the absence of emotional labour. We translated items for surface acting (items 1, 2, 3, 4, 5) and deep acting (items 7, 8) from the emotional labour questionnaire of Grandey ([21]), which employed items from the Emotional Labour Scale (Brotheridge & Lee, 1998, 2003). An example of a surface acting item is: "I pretend to have the emotions I need to display for my job." An example of a deep acting item is: "I make an effort to actually feel the emotions I need to display toward others." From the same study, we selected two items that measure emotional consonance (items 10, 11) that were taken from another study (Grandey, [20]). One of these emotional consonance items is: "I react to students' emotions naturally and easily." A third emotional consonance item was taken from a study in which it was a reversed item in a scale that measured dissonance: "The emotions I show the students match what I truly feel" (Kruml & Geddes, [26]). We added one deep acting item (emotive effort, item 4) from the study by Kruml and Geddes ([26]), item 9 in our questionnaire. Others developed an Emotion Requirement Scale to measure perceived display rules (Best, Downey, & Jones, [5]). We rephrased the items that were used to measure the: "Requirement to hide negative emotions" so that they ask about actually hiding anger and disgust and fear, and called the hypothesized separate dimension suppression (items 18, 19, 20). An example of a suppression item is: "I hide my anger about something someone has done." People apparently suppress emotions at work just as often as they fake them (Mann, [28]). All items were measured on a 5-point scale (1 = "never"; 5 = "always").

The structure of the resulting scale was tested on the scores of the present sample of teachers with confirmatory factor analysis (CFA). In order to test the proposed factorial structure of the measure, CFA with maximum likelihood estimation were used utilizing the AMOS 3.61 computer program (Arbuckle, [2]). As the aim of this study was to test a hypothesis, CFA was preferred above Principal Component Analysis, which is more suitable for exploratory purposes. The resulting comparative fit index (CFI), the Bentler-Bonnet non-normed fit index (NNFI), and the root mean square error of approximation (RMSEA) are reported. CFI and NNFI values>.90 indicate that the model is acceptable (Bentler & Bonnet, [3]). Values of RMSEA ≤.05 indicate a good fit of the model (Byrne, [12]).

The CFA of the 4-factor model resulted in indices that varied between.90 and.93 (χ2=210.20, df=71, p=.001, CFI=.93, NNFI=.90, RMSEA=.08). Item 20 was subsequently removed because this improved the internal reliability of suppression significantly (from α=.64 to α=.85). The Cronbach's alphas of the resulting scales were adequate to high: surface acting, five items, α=.83 (n = 362); deep acting, four items, α=.85 (n = 351); suppression, two items, α=.85 (n = 365); and emotional consonance, three items, α=.63 (n = 361). A slightly modified version of the scale was cross validated in a sample of nurses and the 4-factor structure of the scale was confirmed (Briët et al., [6]).

Burnout was measured with the Dutch version of the Maslach Burnout Inventory for Teachers (MBI-NL-Ed; Schaufeli & van Horn, [38]). The questionnaire consists of eight items that measure emotional exhaustion, five items that measure depersonalization, and seven items that measure personal accomplishment. Items can be answered on a 7-point scale ranging from "never" to "always/daily." The possible ranges for emotional exhaustion, depersonalisation, and personal efficacy are 0–48, 0–30, and 0–42, respectively. Indicative for burnout are high scores on emotional exhaustion and depersonalization and low scores on personal accomplishment. Cronbach's α was.90 for emotional exhaustion (n = 364), α=.64 for depersonalization (n = 360), and α=.85 for personal accomplishment (n = 358).

Social support was measured with two scales, each containing six items from the Emotional Support Subscale of the Social Support List–Discrepancies (SSL-D; van Sonderen, [40]). Answers to the items are given on a 4-point scale running from "I miss it" to "it happens too often." A high score indicates much social support. Items from the scales "emotional support from colleagues" and "emotional support from staff members" were both included and scores added to create a general measure of social support. This resulted in a Cronbach's α of.90, n=356.

An assessment of the workers' autonomy can be regarded as a refined measurement of the control dimension (de Jonge, Janssen, & van Breukelen, [14]). Autonomy was measured with six items from the Maastricht Autonomy List (MAL: de Jonge, Landeweerd, & van Breukelen, [15]). Answers are given on a 5-point scale varying from very few possibilities to very many possibilities. A high score indicates a high level of autonomy, α=.83 (n = 362).

Quantitative demands were measured with six items from the Vragenlijst Organisatiestress Doetinchem [Organizational Stress Questionnaire] (VOS-D; Bergers, Marcelissen, & de Wolff, [4]). Respondents are asked to indicate how often they experience difficult quantitative demands on a 5-point Likert scale varying from "seldom" to "very often." A high score indicates high quantitative demands, α=.80 (n = 362).

Correlations between the study variables were calculated. All variables were centred before entering them into a regression analysis (Aiken & West, [1]). We performed a separate stepwise hierarchical regression analysis for each of the three burnout symptoms. In the first step demographic variables were entered as predictors, in the second the three variables from the DCS model and the interaction of demand and control, in the third step emotional labour and emotional consonance.

Table I presents a description of the sample. A comparison with a heterogeneous sample of teachers from another Dutch study (Brouwers & Tomic, [11]) indicated that our sample reported much lower levels of emotional exhaustion and depersonalization and similar levels of personal accomplishment.

Table I. Demographic characteristics, job characteristics, and burnout symptoms of the sample.

Correlations between the dimensions of burnout and the job characteristics and emotional labour are given in Table II. Both emotional exhaustion and depersonalization were related to more quantitative demands, less control, and less social support. Furthermore, both emotional exhaustion and depersonalization were related to more surface acting. More depersonalization was also related to more suppression. Finally, personal accomplishment was positively correlated with control and support, and also positively correlated with emotional consonance. Deep acting was not correlated with any of the burnout symptoms.

Table II. Correlations between demographic variables, work characteristics and emotional labour.

The strength of the relationship between emotional labour and burnout symptoms, after taking into account the relationship with the variables from the DCS model, was assessed in a hierarchical regression analysis. The results of the separate analyses for each burnout dimension are given in Table III. In the first step, the demographic variables age and gender were entered in the analysis. Age was found to be negatively related to personal accomplishment and was not related to emotional exhaustion or depersonalization.

Table III. Hierarchical regression analysis of predictors of emotional exhaustion, depersonalization, and personal accomplishment (n=345).

In the second step, the separate variables of the DCS model were entered and the interaction of demand with control. More demands, less control, and less support were all related to more emotional exhaustion. More control and more social support were related to less depersonalization; the interaction of more control and more demands was also separately related to less depersonalization. More control and more support were positively related to personal accomplishment

In the third step the emotional labour variables were entered. Hypothesis 1a and 1b were confirmed, as more surface acting was significantly related to more emotional exhaustion and to more depersonalization. Hypothesis 2a was not supported, as suppression was not related to emotional exhaustion. Hypothesis 2b was supported as more suppression was related to more depersonalization. Hypothesis 3a was supported as emotional consonance had a negative relationship to depersonalization. Hypothesis 3b was also confirmed, as emotional consonance had a positive relationship to personal accomplishment.

Because others have reported a positive relationship between deep acting and personal accomplishment (Brotheridge & Grandey, [7]), we examined this relationship in more detail. We performed an additional regression analysis with personal accomplishment as criterion in which emotional consonance was entered in a separate fourth step. This analysis revealed that deep acting was a significant predictor of personal accomplishment in step 3, β=.15, p=.006, but after entering emotional consonance in the regression model in step 4, this relationship disappeared.

In line with a multitude of earlier studies that have used the DCS model, our study indicates that quantitative demands, control, and support are significantly related to emotional exhaustion. In addition to these relationships, the emotional labour dimension of surface acting was, as hypothesized, also found to be related to emotional exhaustion. The relationship between surface acting and emotional exhaustion is a replication of earlier findings (Brotheridge & Grandey, [7]; Brotheridge & Lee, [8]; Grandey, [21]; Zammuner & Galli, [42]; Zammuner & Lotto, [43]) and supports the idea that having to pretend certain emotions may come at a personal cost in terms of feeling emotionally drained or exhausted (Hochschild, [23]). This finding adds to the evidence that surface acting in particular should be regarded as stressful (Mann & Cowburn, [29]).

With the exception of quantitative demands, the variables from the DCS model were also found to be associated with depersonalization. Of the three emotional labour strategies that we measured, surface acting and suppression were also found to be significantly related to depersonalization. This result replicates but also extends findings on surface acting in various professions and students (Brotheridge & Grandey, [7]; Brotheridge & Lee, [9]). As our study takes the DCS model into account, it is possible to give an estimate of the relative importance of emotional labour and the DCS model to our understanding of depersonalization. A comparison of the amount of variance explained in steps 2 and 3 indicated that the DCS model was more important than emotional labour in explaining emotional exhaustion in our sample of teachers. Emotional labour and the DCS model seem, however, to be equally important when trying to understand depersonalization.

A new finding in our study is that suppression is significantly related to depersonalization. Most studies of emotional labour have focused on the production of emotions or emotional expressions and have not measured suppression separately. Some have used a measure of "self-focused behaviour" in which suppression is assumed but, taken together with surface acting, Pugliesi ([35]) and others have used a measure of "emotional dissonance" that also explicitly comprises both suppression and surface acting (Zapf, Seifert, Schmutte, Mertini, & Holz, [44]). Brotheridge and Lee ([10]) saw suppression as an activity that is an inherent part of surface acting. As suppression measured with a separate scale was only weakly correlated with surface acting in their study (Brotheridge & Lee, [10]), this assumption does not hold. Apparently people do not necessarily have to suppress an emotion before they act another one. Because several studies report that suppression is used as a separate strategy in various professions (Mann, [28]; Ybema & Smulders, [41]), we also measured it separately in our questionnaire. Other researchers also recently added a subscale to measure suppression to the Emotional Labor Scale (Lee & Brotheridge, [27]), the instrument that is most often used in the USA.

Although it is tempting to conclude that surface acting and suppression contribute to a feeling of depersonalization, our study design does not allow any conclusions about causality. It is just as plausible to assume that teachers who are exhausted or who feel detached from their students do not respond to certain situations in a way that feels natural. In such a state, teachers might have to use surface acting and suppression more.

The third dimension of burnout, personal accomplishment, was strongly related to emotional consonance, with a beta weight of.50. Emotional labour measures the effort to deal with the experience or the expression of emotions and is therefore thought to contribute to stress or strain. Emotional consonance measures the absence of such effort and is a useful extension of the nomological network or constructs related to emotional labour (Diefendorff, Croyle, & Gosserand, [17]). Emotional consonance can be useful in gaining an understanding of personal accomplishment.

Deep acting was not found to be related to personal accomplishment in our study, which seems to be in contrast with findings in another study (Brotheridge & Grandey, [7]). After entering emotional consonance in the regression model for personal accomplishment, the relationship of deep acting with personal accomplishment disappeared. Brotheridge and Lee re-analysed data from a previous study (Brotheridge & Grandey, [7]) and presented a path model in which authenticity of emotional expression was a stronger predictor of personal accomplishment than deep acting (Brotheridge & Lee, [9]). These findings on authenticity parallel our findings on emotional consonance. It therefore seems that emotional consonance, the absence of emotional labour, is more strongly associated with personal accomplishment than deep acting.

The varying results that have been obtained on the relationship between deep acting and emotional exhaustion and personal accomplishment might result from samples that differ in the skilfulness of the individual to apply deep acting. Studies that assess emotional labour repeatedly during the training and supervision of teachers or nurses might give us an understanding of how people change in their use of emotional labour. Over time, emotional labour may vary in amount, in type, and, especially for deep acting, in the degree to which the desired emotional result is obtained. By studying trainees we will get a better understanding of what constitutes adequate professional behaviour and how it develops. Future studies may also point out which levels of deep acting, faking, and suppression can be regarded as adequate or healthy and which levels are harmful.

An indication that some of our findings can be generalized to other European countries can be derived from a comparison from the beta weights in the regression of emotional exhaustion on quantitative demand. This beta weight is.44 in our study, which is almost identical to that of an Italian sample,.46 (Pisanti, Gagliardi, Razzino, & Bertini, [33]) and a Greek sample,.45, (Pomaki & Anagnostopoulou, [34]) of secondary school teachers.

The DCS model has been very influential. The three concepts from this model are still important for burnout researchers, but the scope is getting broader. Job demands now are sometimes subsumed under the header workload; support can be taken in a broader sense to comprise the community (Maslach et al., [30]). Reward in financial or other ways, (un)fairness or inequity of workload or pay, and finally personal values all seem to play a role in the aetiology of burnout symptoms, but workload is most directly related to exhaustion (Maslach et al., [30]).

Some limitations of our study need to be discussed. The emotional labour instrument that we developed and used in this study (Briët et al., [6]) does not measure the acting of negative emotions such as anger and aggression. A scale to measure the faking of negative emotions has been developed by others (Glomb & Tews, [19]). For studies in professions in which such emotions are also expressed (Fitness, [18]), a measure of the (non)expression of negative emotions in the workplace will have to be included.

It should also be noted that we reported on one specific type of teacher (mathematics) in one specific setting (secondary schools). It is not known whether such teachers have special personality characteristics or whether teaching mathematics entails different social interactions from teaching other subjects. What is known from other studies is that the emotional "geography" in primary school is one of emotional intensity, whereas teachers in secondary school seem to behave in more distant ways than their primary school colleagues (Hargreaves, [22]). The situation in secondary school may, however, become very demanding, for example when pupils refuse to do assignments, or worse, when pupils like to act "bad" because the reaction of the peer group makes them feel "good" (Paulle, [32]). Clearly, some schools may suffer more from such problems than others, which is a factor that should be taken into account in further studies.

An important finding in our study is that emotional consonance gives a clue towards understanding personal accomplishment. While it is important to investigate what makes people feel exhausted, there is an increasing interest in what makes people enthusiastic about their work and what makes them feel competent. In theoretical models that focus on engagement (Schaufeli & Bakker, [36]) and resources (de Jonge, Dormann, & van Vegchel, [13]), emotional consonance may play an important role.

In conclusion, the DCS model has been very valuable for our understanding of emotional exhaustion. The concept of emotional labour seems to be an additional useful perspective. The study of surface acting, especially, might give us a better understanding of emotional exhaustion and depersonalization.

The authors appreciate the cooperation of the board and members of the Dutch Association of Teachers of Mathematics. The first author would like to thank the University of Lisbon for their hospitality during the preparation of this manuscript.