Perfectionism, over commitment to work, and burnout in employees seeking workplace counselling

Abstract

Perfectionism is linked to a range of psychological disorders. Burnout is an important construct in the working population and has been found to overlap with related constructs such as depression. There has been limited research conducted between perfectionism and burnout. This study aimed to examine the link between perfectionism and burnout and if procrastination and over commitment to work were mediators of the relationship. There were 69 participants recruited who were seeking counselling from a counselling firm, providing employee assistance services. The results indicated that procrastination was not a significant mediator of the relationship between concern over mistakes (CM) perfectionism and burnout. There was a significant mediating relationship found, where overcommitment to work mediated the relationship between CM perfectionism leading to burnout. The implications of the study in terms of interventions for employees suffering burnout, and directions for future research in regards to the relationship between perfectionism and burnout are considered.

Key words:

• burnout
• employees
• perfectionism
• work

Recent research has examined the effects of perfectionism and burnout in employees in the workforce (Leonard & Harvey, 2008; Mitchelson & Burns, 1998; Stoeber & Rennert, 2008). Burnout was originally defined as a stress reaction from employees whose primary work was human services (Maslach, 1978). It has been defined as a three dimensional construct comprising of exhaustion, depersonalisation, and reduced personal accomplishment as a result of work (Friedman, 2000). It has been argued that perfectionism may be an important intrinsic variable in employees who are experiencing burnout. Perfectionism is elevated in individuals reporting anxiety disorders, eating disorders, and depression compared with controls (Egan, Wade, & Shafran, 2011; Shafran & Mansell, 2001). While there has been debate if perfectionism can be positive (Shafran & Mansell, 2001), there is some evidence that achievement striving is related to positive adjustment in non‐clinical samples (see Stoeber & Otto, 2006 for a review). There has also been debate regarding the definition of perfectionism. Shafran, Cooper, and Fairburn (2002) proposed a definition of clinical perfectionism as the pursuit of personally demanding standards despite negative effects and sense of self‐worth being based on achievement. Their cognitive‐behavioural model was developed to account for the factors that maintain clinical perfectionism and has recently been updated (Shafran, Egan, & Wade, 2010). One of the core components of the model is avoidance, which Shafran et al. (2002, 2010) state that individuals with clinical perfectionism use as a way of ameliorating the distress from failing to live up to high standards.

Santanello and Gardner (2007) suggested that avoidance may be expressed by perfectionist behaviours such as working excessive hours or overly scrutinising work as a way of avoiding negative thoughts about inadequacy. Santanello and Gardner argued that once perfection becomes unattainable, the perfectionist will then engage in avoidance by procrastinating, an interpretation based on their finding of a significant correlation on the Frost multidimensional perfectionism scale (FMPS; Frost, Marten, Lahart, & Rosenblate, 1990) and avoidance in a sample of college students. Ferrari (1992) also reported a significant positive correlation between procrastination and socially prescribed perfectionism on the Hewitt and Flett multidimensional perfectionism scale (HMPS; Hewitt & Flett, 1991a) in a sample of college students.

There has been limited research that has directly examined the relationship between burnout and perfectionism. Burnout is an important variable to understand however, as it has been associated with absenteeism from work, physical health problems, and as a possible precursor to depression. Ahola et al. (2008) found that individuals experiencing burnout were more likely to experience a medical absence from work lasting greater than 9 days. Stoeber and Rennert (2008) examined the relationship between perfectionism, as measured by an adapted version of multidimensional inventory on perfectionism in sports scale, and burnout, as measured by the German version of the Maslach burnout inventory in a sample of secondary school teachers. They found a correlation between the teachers' negative reactions to failure to live up to high personal standards (PS) and burnout. Mitchelson and Burns (1998) reported a correlation between perfectionism, as measured by the FMPS total (Hewitt & Flett, 1991a), and burnout, as measured by the Maslach burnout inventory—general survey (Maslach, Jackson, & Leiter, 1996), in working mothers, in terms of exhaustion and cynicism. Flett, Hewitt, and Hallett (1995) reported a correlation between self‐oriented perfectionism and a measure of professional distress in a sample of teachers. However, the Flett et al. study did not contain a measure of burnout. The authors found more associations with socially prescribed perfectionism and measures of teacher stress, possibly because the items from the teacher stress inventory loaded on external stressors at work.

A construct that appears to be relevant to understanding if there is a link between perfectionism and burnout is over commitment to work. Siegrist et al. (2004) argued that Over Commitment to Work (OCW) is a coping mechanism for some individuals, which is characterised by the individual seeking higher demands at work or extending efforts at work beyond what is formally required by the organisation. Siegrist et al. define OCW as excessive work‐related commitment and a high need for approval. Items on their measure of OCW relate largely to an inability to withdraw from work obligations. They found a moderate, positive correlation between the OCW and the concern over mistakes (CM) subscale of the German‐language version of the multidimensional perfectionism scale (MPS; Frost et al., 1990; Wirtz, Siegrist, Rimmele, & Ehlert, 2008). This is an interesting finding; however, given the correlation nature of the research, there is no explanation of what antecedents of OCW are. Consequently an aim of the present research was to examine if CM is an antecedent to over commitment to work, leading to burnout.

The purpose of this study was to examine potential mediators of the relationship between perfectionism and burnout to further understand it. The FMPS subscales of CM and PS were used to represent perfectionism. It was hypothesised that individuals high in perfectionism as measured by the CM and PS subscales of the FMPS (Frost et al., 1990) will have higher levels of work‐related burnout. Santanello and Gardner (2007) suggest that excessive work hours and scrutinising one's work are coping mechanisms to avoid painful emotions with perceived failure of achieving high standards. Thus, the second prediction is that high PS will be correlated with high OCW. Shafran et al. (2002) posit that another form of coping with the potential failure of meeting high standards is to engage in procrastination and avoidance. The third prediction is that CM will be correlated with procrastination. The mediation model proposes that procrastination should mediate the relationship between CM and work‐related burnout, and OCW should mediate the relationship between PS and work‐related burnout. Fig. 1 depicts the proposed mediation model, with CM and PS as the exogenous variables and procrastination, OCW, and work‐related burnout as the endogenous variables.

Figure 1 Model 1, a saturated model containing all possible recursive pathways.

METHOD

Participants

Participants were recruited from a private counselling firm that provides employee assistance counselling for a variety of large organisations in Western Australia. The participants were all currently employed. There were 69 participants (71% female, 29% male); and 70 participants completed the questionnaire, but one case was removed because of significant missing data. Sixteen per cent of participants were aged between 18 and 25, 17% of participants were aged between 26 and 35, 33% or participants were aged between 36 and 45, 23% of participants were aged between 46 and 55, and 10% of participants were aged over 56. The median education level was an undergraduate university degree (32%). The median occupation type was professional (58%), 19% were identified working in an administrative role, 13% were identified having a semi‐skilled occupation, and 10% were identified working in a trade or skilled occupation.

Measures

Multidimensional perfectionism scale (FMPS; Frost et al., 1990)

Perfectionism has predominately been measured by the two multidimensional perfectionism scales, the 35‐item FMPS (Frost et al., 1990), and the 45‐item HMPS (Hewitt & Flett, 1991). The FMPS consists of six subscales: PS, CM, doubts about actions (DA), parental expectations (PE), parental criticism (PC), and organisation (O). The HMPS consists of three subscales: self‐oriented perfectionism, socially prescribed perfectionism, and other‐oriented perfectionism. Factor analyses have shown a two‐factor solution: maladaptive evaluative concerns (CM, DA, PC, PE, and self‐oriented perfectionism) and achievement striving (PS, O, and other‐oriented perfectionism; Bieling, Israeli, & Antony, 2004; Frost, Heimberg, Holt, Mattia, & Neubauer, 1993). The CM and PS subscales of the FMPS was chosen as the measures of perfectionism in the present study as, when combined, these are the most relevant to ‘clinical’ perfectionism (Egan et al., 2011).

Participants rated their endorsement of each statement on the FMPS on a 5‐point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The overall score of perfectionism in the FMPS (excluding the O subscale) is reliable as measured by a Cronbach alpha of 0.90 (Frost et al., 1993). In this study, Cronbach's alpha was 0.93 for CM and 0.81 for PS. An example item for PS is ‘If I do not set the highest standards for myself, I am likely to end up a second‐rate person’, and for and example item for CM is ‘If I fail at work, I am a failure as a person’.

Copenhagen burnout inventory (CBI; Kristensen, Borritz, Villadsen, & Christensen, 2005)

The CBI is a 19‐item measure that assesses three dimensions of burnout—personal burnout (fatigue), work‐related burnout, and client‐related burnout. The CBI uses frequency or intensity measures for relevant items on a 5‐point Likert scale, from 1 (a very low degree) to 5 (a very high degree) for intensity and from 1 (never/almost never) to 5 (always) for frequency. The reliabilities for the three subscales have been found to be good, ranging from 0.85 to 0.87 (Kristensen et al., 2005) and from 0.79 to 0.89 (Winwood & Winefield, 2004). In this study, Cronbach's alpha for work‐related burnout was 0.86. As this study was only interested in work‐related burnout, the seven‐item work‐related burnout subscale was administered.

The procrastination scale (Tuckman, 1991)

The procrastination scale is a 16‐item measure that was designed to assess procrastinating tendencies in college students. It was chosen as a measure of avoidant behaviour. It is measured on a 4‐point Likert scale from 1 (that's not me for sure) to 4 (that's me for sure). Four items are reversed scored. It has a reliability of 0.86, and good concurrent validity with a similar measure of procrastination, the general self‐efficacy scale (Scherer et al., 1982). Cronbach alpha in this study was 0.89.

Intrinsic effort scale (Siegrist et al., 2004)

The intrinsic effort scale measures the construct of Over Commitment to Work. It is a six‐item subscale of the Effort‐Reward Ratio Model (Siegrist et al., 2004). It is measured on a four point Likert scale from 1 (strongly disagree) to 4 (strongly agree). The internal consistency in numerous studies ranges from 0.64 to 0.81 (Siegrist et al.). In this study the Cronbach alpha was 0.82.

Procedure

The study received approval from the Curtin University Human Research Ethics Committee. Clients presenting to Prime Psychology were offered to complete an anonymous work‐stress survey by the receptionist prior to entering their appointment for counselling. The questionnaire contained a participant information sheet and consent form in addition to the measures. A return box was available at the reception desk as well as the option of returning the survey by mail. No surveys were returned by mail. One hundred and ten surveys were distributed and 69 participants completed questionnaires returned giving a 64% response rate. Seventy questionnaires were initially returned; however, one individual was excluded because of significant missing data in not completing questions.

RESULTS

Table-1 indicates the means and standard deviations for the measures. To determine the relationship between the variables, structural equation modelling (SEM) was performed using LISREL, version 8.8 (Jöreskog & Sörbom, 2007).

Table 1 Means and standard deviations for measures of perfectionism, procrastination, burnout and over‐commitment to work

Measure	M	SD
Procrastination	33.06	8.61
Burnout	23.16	5.49
Overcommitment	15.51	4.06
Concern over mistakes	25.25	8.59
Personal standards	20.33	4.16

SEM assumption testing

SEM requires multivariate normality (Kline, 1998) and a chi‐square test was not significant (χ² = 0.098, p = .952), indicating the variables were multivariate normal. Examination of the scatterplots of the bivariate relationships indicated that the assumption of linearity that is required for SEM was satisfied. Finally, all measures had relatively high tolerance values indicating that multicolinearity in the data was not a problem (Stevens, 2002).

Testing of the saturated model

A diagram of the saturated model (model 1) can be seen in Fig. 1. The model has a measurement component and a structural component. The measurement component assumes that each of the five observed measures (fmps1, fmps2, ps, ies, and cbi) are driven by a latent variable (CM, PS, Procrastination, OCW, and work‐related burnout, respectively). To overcome the problem of having only one measure associated with a latent variable that LISREL requires to compute measurement errors and factor loadings, the measurement error associated with the observed variables was set to one minus their reliability coefficients, and the factor loadings to the square root of reliability coefficients for each measure (Goodwin & Plaza, 2000).

The χ² goodness of fit statistic for model 1 was non‐significant (χ²[1] = 2.646, p = .101) indicating that model 1 fits the data reasonably well. This is not surprising as the model includes all possible pathways and will therefore fit the data better than any of its nested counterparts. To determine if pathways could be removed from the saturated model to create a nested model that is less complex, path coefficients for the structural aspect of the saturated model were examined to identify non‐significant pathways that were candidates for removal, as seen in Table-2. Four of the pathways were non‐significant and were therefore removed, resulting in a reduced Model 2, which is a mediator model, as seen in Fig. 2.

Table 2 Path coefficients for model 1 (N = 69)

	Path coefficients	Significance
Model 1: Saturated model
CM → procrastination	0.852	p < .01
CM → overcommitment	0.644	p < .01
CM → burnout	−0.120	p > .05
PS → procrastination	1.009	p < .01
PS → overcommitment	−0.214	p > .05
PS → burnout	0.128	p > .05
Procrastination → burnout	−0.007	p > .05
Overcommitment → burnout	0.889	p < .01

Figure 2 Model 2, the mediator model.

The χ² goodness of fit statistic for model 2 was non‐significant, χ²[3] = 2.924, p = .404, indicating that model 2, like model 1, fit the data reasonably well. The path coefficients for model 2 are listed in Table-3. The χ²_difference statistic was computed by subtracting the χ²‐value for Model 1 (χ² = 2.695) from the χ²‐value for model 2 (χ² = 2.924). The degrees of freedom (dfs) for the χ²_difference statistic was derived by subtracting the dfs for the model 1 χ² (= 1) from the dfs for the model 2 χ² (= 3). Thus χ²_difference = 0.229 with two degrees of freedom, which returned a p‐value of 0.892 and therefore shows there was a non‐significant difference between the fit of the models and thus it can be concluded that the two models fit the data equally well. For the purpose of scientific parsimony in accepting the most simply model, the more parsimonious model 2 was accepted.

Table 3 Path coefficients for model 2 (N = 69)

	Path coefficients	Significance
Model 2: Mediator model
CM → procrastination	0.853	p < .01
CM → overcommitment	0.630	p < .01
PS → procrastination	−1.011	p > .01
PS → overcommitment	−0.201	p > .05
Procrastination → burnout	−0.067	p > .05
Overcommitment → burnout	0.873	p < .01

As can be seen in Table-3, the PS → overcommitment pathway and the procrastination → burnout pathway were once again non‐significant as seen also for model 1. This leaves one significant mediating pathway CM → overcommitment → burnout. As seen in Table-3, the path coefficient for the non‐significant procrastination → burnout pathway is close to zero and is therefore unlikely to reach significance with increases in sample size, which rules out a mediating role for procrastination.

DISCUSSION

The results demonstrated that CM leads to OCW which results in burnout. The results suggest that if individuals have a high CM, which is clearly recognised a prominent part of unhelpful or clinical perfectionism (e.g., Shafran & Mansell, 2001), they have to achieve high standards at work are associated with symptoms of burnout.

The prediction that perfectionism would be associated with burnout was confirmed. The findings add support to the research on the relationship between perfectionism and burnout (Flett et al., 1995; Mitchelson & Burns, 1998; Stoeber & Rennert, 2008). Shafran et al. (2002) have argued that PS and CM is in line with the original construct of perfectionism being a self‐focused process.

Our prediction that high PS would be associated with OCW was also supported. The finding supports the results of Wirtz et al. (2008) who reported an association between OCW and perfectionism. The moderate association between PS and OCW offers an underlining link between the two constructs. OCW was also strongly correlated with work‐related burnout. The results support the model proposed by Shafran et al. (2002) by demonstrating that individuals with perfectionism pursue high standards despite significant negative outcomes. In this study, it has been demonstrated that the pursuit of high standards at work is associated with adverse outcomes of difficulty separating from work and symptoms of burnout. Shafran et al. posited that the pursuit of high standards is maintained in order to avoid failure and self‐criticism. They also hypothesised that the individual may reappraise the standards once they are achieved and may set higher standards to attain. In the work environment, this may contribute to difficulty separating from work in order to fulfil these standards.

Shafran et al. (2002) proposed that the fear of failing to meet high standards in some individuals leads to avoidant behaviours. This study employed procrastination as a measure of this type of avoidant behaviour. Procrastination was not found to be correlated with CM or work‐related burnout, which is in contrast to previous findings using similar measures (Ferrari, 1992; Frost et al., 1990; Walsh & Ugumba‐Agwunobi, 2002). There are several possible reasons for this. Previous studies have only reported a moderate association between procrastination and perfectionism ranging from 0.22 (Frost et al.) to 0.40 (Walsh & Ugumba‐Agwunobi). It is possible that the sample size of this study was too small to detect a moderate population effect between procrastination and CM. Another possible reason is that the measure of procrastination does not adequately measure avoidant behaviour. Dunn, Whelton, and Sharpe (2006) administered the COPE inventory (Carver, Scheier, & Weintraub, 1989) as a measure of avoidant coping, which contains subscales of mental and behavioural disengagement. Dunn et al. reported a mediation effect of avoidant coping between perfectionism and psychological distress. Finally, it is possible that the population sampled in this study is not representative of the broader working population in terms of coping behaviours. The workers attending counselling may demonstrate less avoidant coping styles as they are actively seeking help.

The finding of an association between OCW and work‐related burnout is important, as OCW measures the psychological difficulty of withdrawing from work commitments (Siegrist et al., 2004) as opposed to working long hours. This finding is supported in the literature, with the number of hours worked unrelated to physical health symptoms, psychological health, or fatigue (Akerstedt et al., 2004; Tucker & Rutherford, 2005). Burnout, as defined by Kristensen et al. (2005), is the attribution of fatigue to work stressors. OCW has been found to be moderately correlated with chronic stress (Wirtz et al., 2008) and fatigue (Preckel, Meinel, Kudielka, Haug, & Fischer, 2007), and burnout has been linked to poor physiological outcomes such as absenteeism because of illness (Ahola et al., 2008; Firth & Britton, 1989; Honkonen et al., 2006) and depression (Ahola et al., 2005). Thus, individuals with high PS are more likely to feel unable to disengage from work in order to meet their high standards, which may contribute to symptoms of burnout.

Limitations and future research

It is important to note the limitations of this study. First, the sample size was relatively small. At a minimum, five subjects per fixed parameter is required for SEM (Kline, 1998), which for the current study would be a sample size of 75, thus our sample of 69 participants resulted in a minimally acceptable sample size. Consequently, results should be interpreted with caution. For example, the path coefficient for the PS → overcommitment pathway was approaching significance, and if the sample size was larger, it may have been significant, which would have produced a second mediating pathway running through overcommitment. Therefore, the results must be interpreted as preliminary data where the nature of the proposed relationships and mediating model should be investigated with a larger sample.

The second limitation was an inability to make causal links between perfectionism, OCW, and burnout. As suggested by Shafran and Mansell (2001), high PS often begin with positive outcomes, such as achieving good results or praise at work, but over time become more negative. Although SEM can give clues as to the direction of relationships between variables, a longitudinal study would be required to examine the causal nature of CM leading to OCW and consequently burnout.

A further limitation was the biased sample of the population being individuals who were seeking workplace counselling. Consequently, it is highly likely that the sample comprised individuals who were already experiencing significant burnout. It is likely that a different result would have occurred in a random sample of workers, not just those individuals seeking counselling. Notwithstanding this limitation in generalisability, there is also a strength as it was of interest to determine for those individuals who seek counselling because of workplace stress what individual factors, such as perfectionism, may be relevant. Furthermore, it was an aim of the study to understand relevant factors leading to burnout in workers; thus, it was intended to sample individuals who may potentially have higher levels of burnout as they are seeking counselling. The results are important as it suggests that screening for and targeting high CM through relevant treatment programs (e.g., Shafran et al., 2010) that are known to reduce these problematic aspects of perfectionism (Egan et al., 2010) may be useful for individuals who are seeking workplace counselling. Nonetheless, future research should aim to see if this mediating model found among individuals seeking counselling is replicated in a broader sample of workers.

Another limitation is that this study did not include external variables, such as work load or home life, as factors in the relationship between perfectionism and burnout. The difficulty in including external variables is quantifying the impact on the individual. This study, however, has identified that perfectionism should be included in future research into burnout. Finally, in research examining burnout, the contribution of individual factors is an area that requires further research attention. It would have been useful in this study to have examined work‐related data, and future research should investigate this to help understand the relationship between work‐related and individual factors.

CONCLUSION

This study has made an important contribution to the limited research between perfectionism and burnout. We identified partial explanations for the factors involved in burnout. Perfectionism and OCW are two important factors when considering the impact of burnout on workers. The ability for an employee to separate from work at the end of the working day is linked with the high standards that the worker holds and the dichotomous rules in which they derive self‐worth. Clinicians working with clients presenting with symptoms of burnout would benefit from examining whether the client displays perfectionism traits and the impact of this on their working behaviour.