The Mediating Effect of University Teaching Staff’s Psychological Well-being between Emotional Intelligence and Burnout

Introduction

In recent years, the work done by the Teaching and Research Staff (TRS) has evolved in response to the new teaching guidelines of the European Higher Education Area (EHEA). These guidelines require that teaching staff continuously update their professional skills and gain new knowledge. These demands and a lack of personal resources to address them may cause tension, stress, or anxiety. If this situation continues over an extended period it may provoke the recognised syndrome of burnout at work (Rabasa et al., 2016).

According to Maslach and Jackson (1986), burnout manifests itself through three symptoms: 1) emotional exhaustion or tiredness (situation in which workers perceive that they can no longer give more of themselves emotionally); 2) depersonalisation or a cynical attitude towards others (development of negative attitudes and feelings towards the people receiving the work); and 3) reduced personal accomplishment, characterised by feelings of being incompetent and a failure as a professional (tendency of workers to value themselves negatively, this assessment affecting their ability to work).

Maslach and Jackson (1986) studied burnout from a psychosocial perspective, considering it a “process” developed through the interaction between workplace characteristics and personal traits. It starts with exhaustion, then with depersonalization or cynicism, and finally with reduced personal accomplishment. Notably, Maslach et al. (1996) replaced this third dimension with professional efficacy, giving the model two negative dimensions and a positive one. Thus, burned-out people have high scores in the negative dimensions (exhaustion and cynicism) and low scores in the positive dimension (professional efficacy) (Bresó et al., 2007).

As the reader can observe, these symptoms have a very significant emotional component. This is why not all teachers handle demands in the same way. There are some personality traits that protect university teachers from burnout (Mérida-López & Extremera, 2017; Jennings & Greenberg, 2009).

One of those personality traits that protect university teaching staff from burnout is emotional intelligence (EI) (Andrés, 2012; Brackett et al., 2006; Ilaja & Reyes, 2016; Mérida-López & Extremera; 2017).

Nowadays, there are different theoretical conceptualisations about this construct. However, one of the most prominent is the one referred to the TMMS-24 instrument (Aranguren & Irrazabal, 2015; Fernández-Berrocal et al., 2004; González, 2011), where EI is considered as the meta-knowledge that people have their abilities to perceive, assimilate, understand, and manage their own emotions and others’, distinguishing three elements: 1) emotional perception: ability to feel and pay attention to feelings appropriately; 2) emotional understanding: identification and adequate understanding of emotional states; and 3) emotional regulation: ability to regulate negative emotional states and maintain positive ones. Of these dimensions described, the most basic is the one related to emotional perception that once developed allows the person to develop emotional understanding that will allow them to reach the highest level, emotional regulation (Mestre-Navas & Fernández-Berrocal, 2015).

Emotionally intelligent people pay attention to emotions derived from their surroundings, understand the possible causes and consequences of such emotions, and finally develop strategies to control or handle feelings (Mayer & Salovey, 1995). Based on the above findings, EI can be thought of an important personal resource to perceive, accept, and regulate negative emotions (exhaustion and cynicism) and as a protection factor against burnout (professional efficacy).

This construct has highlighted how important the role of understanding and controlling emotions can be for psychological well-being (PWB), since the appearance of the first theory on EI changed the study of emotions (Salovey & Mayer, 1990). Researchers went from thinking that emotions were a distraction for cognitive processes to understanding them as a vital phenomenon that provides useful information on how to solve everyday problems, including problems derived from work demands, stirring up interest in studying EI in relation to PWB. Ryff (1989) understands PWB as the result of an evaluation made by an individual about how he or she has lived, addressing the challenges faced while trying to function in a positive way (Ryff & Keyes, 1995), and overcoming the demands of his or her work environment. According to Aranguren and Irrazabal (2015), the PWB model by Ryff is made up of the following dimensions: 1) self-acceptance: people try to feel good about themselves even when they are aware of their own limitations; 2) positive relationships: individuals need to establish social relationships and have friends they can trust; 3) autonomy: to be able to sustain their own individuality in different social contexts; autonomy makes it possible to resist social pressure to a greater extent and better self-regulate behaviour; 4) environment mastery: it is the personal skills to choose or create favourable environments to satisfy an individual’s own desires and needs; 5) personal growth: consists of the effort to develop an individual’s own competences; and 6) purpose in life: people’s need to set goals, defining a series of objectives that allow their life to be endowed with a certain sense.

In the current welfare society, the interest in studying this relationship may be owed to evidence suggesting that EI can help people adapt to important life contexts and strengthen PWB. That is, EI could be understood as a useful tool for psychological adjustment and optimal interpersonal relationships in people’s lives (Augusto-Landa et al., 2012; Barraca, 2010; de los Remedios et al., 2017; Muñoz, 2016). As García et al. (2018) contend, job satisfaction (professional efficacy) contributes positively to workers’ PWB and can reduce levels of burnout (exhaustion and cynicism).

There is little research around the study of the three constructs indicated together, and none focused on university teachers. However, as has been stated, there is evidence relating EI levels to PWB (Aranguren & Irrazabal, 2015; Mayer et al., 2000; Rey & Extremera, 2011) and also to burnout (Arrogante, 2014; Ilaja & Reyes, 2016). Similarly, there are studies that relate PWB to burnout (Martínez-Ramón, 2015; Portilla & Alva, 2019). These studies conclude the following: teachers with high scores in the highest-level dimension of EI (emotional regulation) and PWB should show lower levels of exhaustion and cynicism, as well as higher levels of professional efficacy, that is why this study aims to:

• - Validate the multidimensional structure of the instruments used.
• - Analyse the levels of EI (differentiating between its dimensions, from lowest to highest level), PWB, and burnout (distinguishing between negative and positive dimensions).
• - Determine the mediating effector between the constructs studied.

Therefore, this study is aimed at verifying the following hypotheses, summarised in Figure 1:

H1: The higher the level of emotional perception in teaching staff, the higher the level of emotional understanding.

H2: The higher the level of emotional understanding, the higher the level of emotional regulation.

H3a: The higher the level of emotional regulation, the lower the level of negative burnout, in terms of exhaustion and cynicism.

H3b: The higher the level of emotional regulation, the higher the level of professional efficacy.

H4: The higher the level of emotional regulation, the higher the level of PWB.

H5a: The higher the level of PWB, the lower the level of negative burnout, in terms of exhaustion and cynicism.

H5b: The higher the level of PWB, the higher the level of professional efficacy.

Hypotheses H3a, H3B, H4, and H5a encourage thinking within a mediation framework. Specifically, the relationship between emotional regulation and negative burnout (H3a) and the relationship between emotional regulation and professional efficacy (H3b) might be mediated by PWB. Other conditions for these two mediation effects are that emotional regulation relates to PWB (H4), that PWB relates to negative burnout (H5a), and that PWB relates to professional efficacy (H5b). These assumptions of mediation will be investigated in hypothesis H6a and H6b, as follows:

H6a: The relationship between emotional regulation and negative burnout is mediated by PWB.

H6b: The relationship between emotional regulation and professional efficacy is mediated by PWB.

Method

Participants

The study was carried out with the TRS of the University of Las Palmas de Gran Canaria. Three hundred and four teachers participated in the study. The sociodemographic profile analysis reveals that the distribution is slightly skewed towards male teachers (168 men vs. 136 women). As for age, the range is between 29 and 72 years old with a higher representation of people who are 51 to 60 years old (n = 118), being the mean age 49.5 years old and the standard deviation 7.2. Concerning the categories of the teaching staff, there is a high percentage of full professors, and there is a large representation of staff from Social and Legal Sciences (n = 89), as well as Health Sciences (n = 73). Most teaching staff have 21 to 30 years of experience (n = 137). Finally, there is a deviation of almost 10 percentage points between the teaching staff who also hold management posts at the university (45.7%) and those who do not (56.3%).

Instruments

In this study we have used the measurement scales found in the literature which have been widely used to evaluate the constructs included in this model.

Emotional intelligence. To measure EI, we used the TMMS-24 scale by Salovey et al. (1995), in the Spanish version adapted by Fernández-Berrocal et al. (2004). The TMMS-24 scale is a reduced version of TMMS-48 adapted to Spanish which evaluates EI levels, with psychometric qualities that are very similar to those in the original scale. TMMS-24 preserves the three original components while also halving the number of items and keeping items with higher internal consistency: emotional perception (8 items), emotional understanding (8 items), and emotional regulation (8 items). The reduced version is based on a Likert format with five answer options. Table 1 shows some of the items that make up the instrument. TMMS-24 was published before the theoretical reformulation of the model by Mayer and Salovey (1995). For this reason, there is a lack of correspondence between its factors and the dimensions proposed in the new model. Despite this inconsistency, TMMS-24 is undeniably useful to evaluate reflexive processes of emotion. It is also the most widely used tool among the Spanish-speaking population, as well as being related to the model by Mayer and Salovey (1995), even if it does not comprise all of its dimensions (Extremera & Fernández-Berrocal, 2005).

Psychological well-being (PWB). To measure this dimension, we used the Scales of Psychological Well-Being (SPWB) by Van Dierendonck et al. (2008). The original version was made by Ryff (1989). It is aimed at measuring the dimensions described in the multifactor model by Ryff (1989) and Ryff and Keyes (1995). SPWB Spanish version are comprised of 29 items with 6-point Liker-type answers, which have been written both in affirmative and negative sentences. The items are grouped into six factors: self-acceptance (6 items), positive relationships (6 items), autonomy (8 items), environment mastery (6 items), personal growth (7 items), and purpose in life (6 items). Table 1 shows some of the items that make up the instrument.

Burnout. To measure burnout, we used the Maslach Burnout Inventory – General Survey (MBI-GS) by Schaufeli and Leiter (1996), in the version adapted by Bresó et al. (2007). This is a generic measurement tool that is useful to evaluate burnout in all sorts of jobs, independently from the specific tasks involved, and with internal consistency that is higher than .70. MBI-GS assesses dimensions of exhaustion (5 items evaluating feeling of emotional exhaustion, tiredness, fatigue, and tension), cynicism (4 items representing an attitude of indifference, devaluation, and alienation with a person’s job), and professional efficacy (6 items related to professional expectations, expressing beliefs about work capacity and an effective contribution to the organisation). Items are answered using a Likert-type scale with points ranging from 0 to 6 points. Table 1 shows some of the items that make up the instrument.

Procedure

This study is based on the quantitative approach by Díaz-Barriga and Luna (2014), employing a non-experimental methodological design through the use of a completely structured online survey (G-Suite) completed by university teaching staff. Simple information was provided, presenting the research and guaranteeing anonymity and data protection. The TRS was contacted twice; a questionnaire was sent via a mailing list to the staff’s institutional e-mail addresses. The first contact was at the beginning of the university year (192 answers) and the second one was 5 months later, at the end of the university year (112 answers).

Data Analysis

Firstly, we describe the analysis of the validity of the different measurement scales used in this research to know its psychometric properties. Secondly, we test the different proposed hypotheses using structural equation modelling (SEM). SEM was applied using the variance-covariance matrix as input data and based on the maximum likelihood estimation method (MLM). Finally, we test the mediator effect of PWB in the relationship between emotional regulation and negative burnout. The statistical analysis was conducted using SPSS 22 and AMOS 22.

A confirmatory factor analysis was conducted to determine goodness of fit of each measurement scale (EI, PWB, and MBI). The purpose of this analysis was to test the hypothesis that all items on each scale were measuring a common construct. Because of this, a nonsignificant chi-square valuable is desirable, but since the chi-square statistic is dependent on sample size, a pattern of results across a number of goodness of fit tests was considered. As recommended in the literature (e.g., Aldás & Uriel, 2017; Hair et al., 1999), the following tests were included: comparative fit index (CFI), normed fit index (NFI), Tucker Lewis index (TLI), and root mean square error of approximation (RMSEA). A model is considered to provide better fit to the data as more of the following criteria are met: chi-square is nonsignificant, CFI and NFI are higher than .95, TLI is higher than .90, and RMSEA is lower than .08. Anyway, values for both CFI and NFI higher than .90 are indicative of acceptable model fit.

In addition, construct validity, which refers to the extent to which a scale provides a reasonable assessment of the construct that it purports to measure, was evaluated. In this study, internal consistency, individual item reliability, composite reliability, convergent validity, and discriminant validity are evaluated.

Internal consistency was analysed through the Cronbach’s alpha coefficient (Cronbach, 1951). The internal consistency of the scale is acceptable if α ≥ .70.

The individual reliability of each item on the scale is examined through factor loading with its construct. As a general rule, it is established that an item can be part of a scale when it presents a factor loading ≥ .70, although this rule should not be so rigid, allowing items with lower reliability if their presence improves the validity of the construct (Hair et al., 2014).

The composite reliability of each scale allows to test, as the Cronbach’s alpha coefficient, the internal consistency of the items of a construct, that is, its capacity to measure a concept together. The acceptable value is ≥ .70.

The convergent validity of a scale implies that the set of items that form this construct represents the same underlying concept (Henseler et al., 2009), that is to say, all the items tend to measure the same reality and nothing more than that. The analysis of convergent validity is carried out from mean extracted variance (average variance extracted, AVE; Hair et al., 2011), stating that there is convergent validity in a construct when the AVE value is higher than .50 (Fornell & Larcker, 1981), that is, more than 50% of the variance of the construct is due to its indicators.

Finally, according to Fornell and Larcker (1981), a construct has discriminant validity if AVE is higher than square correlations between this construct and the others that form the model and indicates that one construct is different from another.

Results

Reliability and Validity Analysis of Measurement Scales

In this section, we show the results obtained in the analyses used for testing reliability and validity of the scales (EI, PWB, and MBI). Confirmatory factor analyses were used to determine the dimensional character of these scales, which showed that all the constructs are multidimensional. Table 1 shows final items of the scales and their dimensions.

Emotional intelligence. The results of the confirmatory factor analysis applied to the EI scale shed light on the existence of three dimensions, which have been labelled as “emotional perception”, “emotional understanding”, and “emotional regulation”, as stated in the literature (see Table 2).

Although the results of this measurement model indicate statistical significance, χ²(51) = 136.829, p = .000, it should be noted that these statistics depend on the size of the sample, thereby the need to analyse other indicators of fit. In this regard, the results obtained show that the other indicators of global fit of the model are found to be within values recommended by the literature (CFI = .955, NFI = .930, TLI = .941, RMSEA = .075), so we can conclude that the specified model adequately reproduces the observed covariance matrix.

The model demonstrates an acceptable level of individual reliability, since the relationship between each item and its respective dimension is statistically significant, with standardized regression weights higher than .70 in all cases, and with t-statistic values, t being also significant.

Internal consistency was estimated by means of construct reliability, which must reach a minimum value of .70 according to recommendations set by Nunnally (1978). In that respect, the results enable us to confirm internal consistency in all cases since values exceed the minimum. Moreover, the convergent validity of the dimensions was estimated by calculating average variance extracted (AVE) proposed by Fornell and Larcker (1981). Since values were above the critical value of .50, we can accept the convergent validity of this construct. In addition, the Cronbach alpha statistic was used to analyse the internal consistency of measures.

Psychological well-being. The results of the confirmatory factor analysis applied to this scale show the existence of six dimensions as stated in the literature: self-acceptance, positive relations, autonomy, environment mastery, personal growth, and purpose in life (see Table 3). Since the results of this measurement model indicate statistical significance, χ²(104) = 259.584, p = .000, we analysed other indicators of fit too. In this regard, the results obtained show that the other indicators of global fit of the model are found to be within values recommended by the literature (CFI = .943, NFI = .909, TLI = .925, RMSEA = .070), so we can conclude that the specified model adequately reproduces the observed covariance matrix. The model demonstrates an acceptable level of individual reliability, since the relationship between each item and its respective dimension is statistically significant, with standardized regression weights higher than or very close to .70 in all cases, and with t-statistic values, t also being significant.

Also, the internal consistency was estimated by means of construct reliability. The results show this internal consistency in all cases, except for two dimensions – autonomy and environment mastery. Moreover, the convergent validity was estimated by calculating AVE. The results indicate that all critical values were above .50, except for the “environment mastery” dimension. In addition, the Cronbach alpha statistic was used to analyse internal consistency of the measures. Results suggest that future research should include new items in PWB “autonomy” and “environment mastery” dimensions to improve internal consistency and convergent validity.

Burnout. The results of the confirmatory factor analysis applied to this scale show the existence of three dimensions as stated in the literature: exhaustion, cynicism, and professional efficacy (see Table 4). Since the results of this measurement model show statistical significance, χ²(32) = 94.989, p = .000, we have analysed other indicators of fit too. In this regard, the results obtained show that the other indicators of global fit of the model are found to be within the values recommended by the literature (CFI = .963, NFI = .945, TLI = .947, RMSEA = .081), so we can conclude that the specified model adequately reproduces the observed covariance matrix. The model demonstrates an acceptable level of individual reliability, since the relationship between each item and its respective dimension is statistically significant, with standardized regression weights higher than or very close to .70 in all cases, and with t-statistic values, t also being significant. Also, the results of construct reliability (CR) and convergent validity (AVE) indicate that all values exceed recommended values. In addition, the Cronbach alpha statistic confirmed the internal consistency of the scale.

Hypothesis Testing

Given the high number of dimensions in each construct, the specification issues of the structural model have led us to validate PWB latent constructs, negative burnout, and professional efficacy jointly. Thus, we needed a global construct of each one to test the proposed model. For this purpose, PWB indicators were calculated as weighted averages of scores given by respondents to the items that make up each dimension, weighted for regression weights for each of those in CFA, and variables were labelled with their own names. We have followed the same process with negative burnout, whose dimensions correspond to exhaustion and cynicism. In Table 5, we collected the results of this model on the basis of which we can deduce that (1) the model has a good fit (CFI = .971, NFI = .942, TLI = .962, RMSEA = .056), (2) values of construct reliability (CR) and convergent validity (AVE) exceed the recommended values and the Cronbach alpha statistic too, and (3) the model demonstrates an acceptable level of individual reliability, since the relationship between each item and its respective dimension is statistically significant, with standardized regression weights higher than or very close to .70 in all cases, except in four of the cases. For this, in the structural model it is possible to use these new variables to simplify and specify the model.

We tested for discriminant validity of the three latent constructs included in this model (PWB, negative burnout, and professional efficacy). In Table 6 it can be seen that square roots of all AVE are greater than elements not on the diagonal. Therefore, it can be affirmed that the three latent constructs also possess discriminant validity.

To validate the theoretical model proposed, SEM were used, employing the maximum likelihood method. The results indicated that this model was satisfactory, χ²(86) = 168.874, p = .000, CFI = .955, RMSEA = .056, as the CFI value exceeds .95 and the RMSEA value does not exceed the recommended maximum of .08. As shown in Figure 2, these results demonstrate that:

- The higher the university teacher’s level of emotional perception, the higher the level of emotional understanding (b = .320, p = .000). At the same time, the higher the level of emotional understanding, the higher the level of emotional regulation (b = .345, p = .000). Therefore, hypotheses H1 and H2 are accepted.

- Emotional regulation is not associated with negative burnout or professional efficacy (b = .002, p = .981 and b = -.055, p = .367, respectively). These results do not support hypotheses H3a and H3b.

- The higher the level of emotional regulation, the higher the level of PWB (b = .554, p = .000). This result supports hypothesis H4.

- Finally, the higher the PWB, the lower the level of negative burnout and the higher the level of professional efficacy (b = -.537, p = .000 and b = .752, p = .000). Therefore, hypotheses H5a and H5b are accepted.

- The percentages of variance explained by the model of PWB, professional efficacy, and negative burnout are 30.7%, 52.2%, and 28.8% respectively.

Mediation Effect Testing

In order to demonstrate the existence of a partial mediation model explaining an indirect additional effect of emotional regulation on negative burnout and professional efficacy through PWB, the procedure established by Mathieu and Taylor (2006) was followed. We fit only direct (no mediation) model to verify that our model is indeed a full mediation model. The direct model estimates a direct link between emotional regulation and professional efficacy, with no path leading to or stemming from PWB (although PWB remains as one latent variable in the model). The results of this model demonstrate that emotional regulation influences negative burnout and professional efficacy (b = -.306, p = .000 and b = .362, p = .000), although this model exhibited deficient fit indices, χ²(89) = 435.197, p = .000, CFI = .811, RMSEA = .113. Therefore, these results and those obtained in the previous section are consistent with the assumption of a full mediational effect of PWB on the relationship between emotional regulation and negative burnout and professional efficacy, because including PWB in the model eliminates the direct relationship between emotional regulation and negative burnout and professional efficacy. Therefore, these results empirically support H6a and H6b.

Discussion

Concerning the first of the aims of this study, this paper verifies the dimensional structure of the three constructs studied, supporting the three-dimensional nature of EI as explained in the literature: emotional perception, emotional understanding, and emotional regulation (Fernández-Berrocal et al., 2004; González et al., 2011). It also corroborates the multidimensional structure of PWB as indicated in the literature: self-acceptance, positive relationships, autonomy, environment mastery, personal growth, and purpose in life (Aranguren & Irrazabal, 2015; Ryff, 1989; Ryff & Keyes, 1995). Finally, burnout, in its double aspect, has also been corroborated once given its three-dimensional structure: exhaustion, cynicism, and professional efficacy (Cruz et al., 2015; Maslach et al., 1996).

EI three dimensions are sequentially correlated, which matches what has been explained in the literature (Mayer & Salovey, 1995; Mestre-Navas & Fernández-Berrocal, 2015; Salovey & Mayer, 1990).

Emotional regulation has no effect on negative burnout or professional efficacy contrary to what is stated by Mérida-López and Extremera (2017) or Jennings and Greenberg (2009) that emotional regulation is a component which protects university teachers from burnout. Also, these results do not support findings by Augusto-Landa et al. (2012), who concluded that EI predicts occupational burnout and proposed a model that explains 80% of burnout variability.

The results confirm that emotional regulation influences PWB (Augusto-Landa et al., 2012; de los Remedios et al., 2017; Muñoz, 2016). It should be noted that these results coincide with results presented by Muñoz (2016) who states that EI is a personal resource that favours and predicts teachers’ PWB. Also, PWB has a direct, negative influence on negative burnout and a positive effect on professional efficacy. These results match results collected in the study by Barraca (2010), who found a negative correlation between PWB and teaching staff’s discomfort, given that being a university teacher is a heavily stressful occupation. García et al. (2018) also supports this idea, but in this case the study was carried out with healthcare workers.

This study also confirmed the mediating role of PWB in the relationship between emotional regulation and negative burnout (exhaustion and cynicism) and professional efficacy. This result is consistent with previous studies that have shown that PWB has a positive effect as a mediating variable between individual dimensions and psychological adjustment indicators (Arrogante, 2014; Rey & Extremera, 2011). By confirming the existence of a total mediation, this study makes an innovative, valuable contribution, shedding light on the relationship between the variables analysed. To sum up, PWB is a mediator variable in a four-step model, with emotional regulation being the initial variable, and negative burnout and professional efficacy being the result variables. Therefore, the teaching staff with higher levels of emotional regulation and PWB is less likely to suffer from negative burnout and more likely to achieve greater professional efficacy.

This study lets us draw significant practical implications from our findings, for instance, the development of programmes for promoting EI or PWB among teaching staff. An example can be found in Hué (2007), who developed the method of emotional thought, whose objective is that teaching staff develops PWB based on EI principles. With regards to PWB viewed as a mediating variable, we should highlight the programme developed by Lyubomirsky (2008). Although this programme is not explicitly directed at teaching staff, it is indeed aimed at improving PWB among adults. To this end, it includes activities which, in most cases, can be carried out in the workplace, focusing, among others, on the following aspects: 1) expressing gratitude to cope with stress; 2) promoting optimism as an interpretation of the world from a more positive perspective; 3) avoiding excessive thinking, by means of strategies aimed at discouraging negative thoughts; 4) exercising kindness to be attentive to others; 5) taking care of social relationships to communicate, give support, and strengthen loyalty; 6) developing strategies to cope with problems; and 7) letting oneself flow to experience periods of intense concentration and view one’s tasks as an exciting challenge.

This study is not without limitations, which should be taken into account in future research. Firstly, although the sample distribution is very similar to the total defined population, the sample size should be expanded to achieve as much representation as possible. Secondly, we should emphasise the need to carry out longitudinal studies to confirm the results presented in this paper. Other evaluation instruments should also be used, e.g., execution tests, in order to corroborate the results based on different methodologies, thus finding out whether they can be generalised or not.