1 Introduction

Academic staff in the Higher Education / University sector are at risk of both subclinical and clinical mental health problems – ranging from anxiety and depression to suicide ideation and enactment (see e.g., Guthrie et al., 2018; Kinman et al., 2006; Shin & Jung, 2014). The sector has recognised this as an industry-wide issue which has implications for employee wellbeing (i.e., avoiding burnout, physical and mental health impacts on the individual and those around them) and for organisational fitness (lost work days, negative job churn). These impacts occur across individuals’ careers, from PhD training (Mackie & Bates, 2019) through to senior appointments (see e.g., Woodward, 2007). Workplace factors such as job autonomy, participatory decision making and managerial support and time spent doing research are linked with job satisfaction (see e.g. Dougall et al., 2021a; Guthrie et al., 2018). Many of these factors are routinely captured in the UKs Health and Safety Executive (HSE) surveys of workplace experience (Cousins et al., 2004) and are linked to wellbeing and turnover intention (e.g., Mahfooz et al., 2017; Mosadeghrad, 2013). However, despite a growing recognition of the issues and the related harms, little progress has been made in reducing the prevalence of poor mental health or turnover intention in the academy.

The current study aims to explore how two factors, social identity (one’s identification with one’s social groups, i.e., as ‘academics’) and identity incompatibility (when the multiple identities an individual holds sit more or less comfortably with one another) function in this context. It also aims to test the relative impact of identity-based factors relative to those traditionally identified as linked to poor mental health and turnover intention.

1.1 Social identity and well being

The social identities we hold have been shown to be potentially protective in a variety of domains. These include recovery from major surgery, addiction recovery, depression recovery, life transitions (such as retirement), anxiety management and pain management (Cruwys et al., 2013, 2015; Cruwys & Gunaseelan, 2016; Gleibs, Haslam, Jones et al., 2011; Gleibs, Haslam, Haslam et al., 2011 Gleibs et al., 2014; Jones et al., 2012). A body of research also suggests that having access to multiple group identities is particularly protective (Blondé et al., 2022; Jetten et al., 2015; Jones et al., 2011; Jones & Jetten, 2011; Killian, 1952). For instance, having multiple group identities appears to buffer the impact of traumatic emergency surgery (Bule & Frings, 2016). Interventions which support the generation of social connections generation is also show improved outcomes for issues such as anxiety and depression (i.e., Haslam et al., 2016, 2019).

In workplace contexts, social identification with ones’ occupation has been linked to both well-being and resilience. For instance, in a quasi-experimental study, Haslam and O’Brien (2005) asked both bomb-disposal operatives and bar staff to rate the relative stresses of their own and the other jobs. Those with higher levels of identification rated their own job as less stressful, both absolutely and relative to the other group. This suggests that occupational identities (social identities linked to work) may serve to contextualise and reduce the impact of stressors (something observed more recently in other domains, e.g. understanding relapse risk, see Frings, Wood et al., 2019; Frings, Gleibs et al., 2019). If this is the case, having a strong identity related to being an academic may buffer the stressors of the role, resulting in better mental health and lower turnover intention. However, this account may be complicated by the possibility of identity incompatibility (clashes in the identity based and practical demand features of competing identities within an individual, see Ethier & Deaux, 1994).

1.2 Identity incompatibility

While identities are often protective, they are not always so. Some identities can be associated with risky behaviours (i.e., in addiction, sexual health, preventative health, etc., Cruwys et al., 2020; Frings & Albery, 2021; Hsieh & Ruther, 2016; Tombor et al., 2015). Moreover, the nature of our psychosocial lives is that people typically hold numerous identities (i.e., religious, gender, sports, family role-based, occupational, educational, etc.). These identities hold different prominence depending on the current context and how easily accessible they are cognitively (Frings & Albery, 2017). While in some circumstances different identities can be held comfortably (i.e., perhaps as a pacifist and a Buddhist), in others they may not be, if at all (perhaps being a vegan and an abattoir worker, see Héliot et al., 2020; Frings et al., 2022). Such incompatibility has been shown to be linked to negative outcomes. For instance, Jetten found that identities linked with ethnic background and with being a university student could be incompatible (see also Jetten et al., 2008) and that increased incompatibility was linked to lower perceived readiness to join university (Jetten et al., 2008). Likewise, having multiple compatible identities is linked with better life transitions (i.e., Iyer, 2008). In other contexts, (Morgenroth et al., 2021) observed that differences in perceived work-life balance (reflecting different roles or social selves, was driven by perceived fit with leaders, (i.e. amongst women in male dominated professions).

Building on the above work, Frings and colleagues (Frings, Gleibs et al., 2019, 2022) observed a differentiation between practical incompatibility and identity-based incompatibility and provide clear definitions of these. Practical incompatibility is defined as ‘behavioural or resource (including time) based demands or norms stemming from one identity which conflicts with those stemming from other identities. Identity incompatibility arises from are challenges arising from ‘value-based norms stemming from one identity which conflict with another identity’ (Frings, Gleibs & Ridley, 2019b). Frings et al., (2019b) observed these constructs to be both distinct and linked with actual outcomes (University degree class attainment). More recently, identity incompatibility (between faith identities and being a drinker) has been linked to intentions to change problematic drinking behaviour (Frings et al., 2022, see also Dingle et al., 2015).

Within the context of academia, identity incompatibilities are likely. When faced with the interfaces between work and the rest of life, balancing roles (such as parent and academic) has been documented as difficult (e.g., Toffoletti & Starr, 2016). Academia also still faces issues with equality of access to the profession based on social class, and class has been linked to differences in academics’ wellbeing (Dougall et al., 2021b). Within their jobs, academics are required to fulfil multiple core functions in their occupation, providing high quality, often carefully monitored, student experiences and outcomes (being an educator), while also delivering research outcomes in terms of income, research outputs and real-world impact (being a researcher). The volume of work means these can often involve zero-sum resource trade-offs (i.e., an hour providing student support is an hour less research, or vice-versa). This makes some level of incompatibility between these identities perhaps inevitable. Academics may also assign different values to these sub-roles and have preferences for how they would like to use their time which are more or less realisable. At an institutional level, organisations also have different perspectives on the relevant balance of work, and/or segregate staff according to their workload (i.e., designating ‘research’ and ‘teaching’ as the core role).

With this evidence base a series of hypotheses can be generated. Since incompatibilities have been linked to poor outcomes, it is likely that academics experiencing identity incompatibility will have poorer mental health, a poor experience of their workplace and increased turnover intention. Given that one’s identity contextualises experience of practical difficulties (see e.g., Frings et al., 2019a; Haslam et al., 2005), a less pronounced link may be expected for practical incompatibility. We can also speculate on the impact of relative levels of identity in situations where incompatibility is perceived. To the extent that incompatibility impacts more valued outcomes and deeply held values, it is likely that observed effects may be more pronounced amongst those who value their identities most highly. In addition, conflicts are also likely when both identities are equally important as the needs and values of one cannot be easily disregarded to meet the other (relative to the situation when one identity domain is not valued). These effects could be expected to concatenate such that incompatibility has the highest impact when both social categories are (i) highly valued and (ii) valued to a similar extent.

In summary, the current study aimed to test four pre-registered (https://osf.io/uta95/) hypotheses:

H1

Higher levels of identity incompatibility between educator and researcher identities will be linked with (i) poorer mental wellbeing, (ii) poorer experience of work and (iii) increased turnover intention.

H2

These effects will be most pronounced amongst those who identify highly with being an academic (i.e., a moderation effect of academic identity will be present).

H3

H1 effects will be most pronounced when both researcher and educator identities are similar in strength (i.e., a moderation effect of identity difference will be present).

H4

H1 effects will be most pronounced when both academic identity is high and educator/researcher the difference in levels of identification is high (i.e., an interaction effect will be present between the two moderators).

2 Methodology

2.1 Participants

A sample of 141 UK resident academics was recruited (see Procedure, below, for recruitment methods). Of these 41% self-reported as male and 59% as female (M age = 38.69 years, SD = 11.60). The mean number of years in the industry was 10.56 (SD = 9.00, minimum = 0, max = 55) and the sample represented PhD candidates (n = 19), Post-doctoral researchers (n = 2), Lecturers (n = 47), Senior Lecturers (n = 18), Assistant Professors (n = 14), Associate Professors (n = 8), Readers (n = 17), Principal Lecturers (n = 5) and Professors (n = 11). The sample included people whose ethnic origin was self-reported as White British (n = 100), Other White (n = 18), Other (inc. mixed) (n = 6), Asian or Asian British (Indian) (n = 5), Asian or Asian British (Pakistani) (n = 4), Black or Black British (African) (n = 3), Chinese (n = 2), Black or Black British (Caribbean) (n = 1), Asian or Asian British (Bangladeshi) (n = 1) and White Irish (n = 1). In terms of mental health, GAD7 (see below) scores in the sample ranged from 1 to 4 (M = 2.07, SD = 0.76). 4 is a recognised cut-off of anxiety symptoms for this scale. For PHQ8 (see below) the sample ranged from 1 to 4 (M = 1.96, SD = 0.73), with 5 being a recognised cut off for clinically depressive symptoms.

For the planned moderation analysis, power analysis using G-Power (3.1.9.2) suggests the sample would allow detections of differences between two sloped in the region of ∆ slope = 0.04, with an achieved power of 0.84. For the mediation analysis, power analysis using the Schoemann et al. (2017) approach suggested, assuming moderate (r =.4) correlations between predictors, mediators and outcomes, a low correlation between mediators (0.2) and a moderate direct effect (0.5), a sample of 141 yielded a power = > 0.85.

2.2 Design

A cross sectional design was employed measuring social identities, identity incompatibility, mental distress, experience of work and turnover intention.

2.3 Measures

2.3.1 Identities

Levels of identity associated with being an academic, an educator and a researcher were measured using the item ‘I identify with [group] as a group’. Participants marked on seven-point Likert type scales the degree to which they identified with each group anchored at 1 (Not at all) and 7 (Very Much) (see Postmes et al., 2013). Higher scores indicated greater identification. Identity difference was calculated by subtracting educator identity scores from researcher identity scores.

2.3.2 Identity incompatibility

Using the eight item scale devised by Frings, Gleibs et al., (2019b) identity incompatibility was measured using the following items; (1) ‘being an educator means there are practical difficulties that get in the way of me being a researcher’, (2) ‘being an educator puts demands on my time which may make it difficult to be a researcher’, (3) ‘It is difficult to balance the practical demands of being an educator and being a researcher’, (4) ‘As a researcher, being an educator is not encouraged’, (5) ‘Sometimes my colleagues don’t understand why, as a researcher, I want to be an educator’, (6) ‘I worry that the values of being a researcher are incompatible with my identity as a researcher’, (7) ‘I worry that what makes me an educator is incompatible with what makes me a researcher’, and (8) ‘I cannot talk to my researcher colleagues about education’. Participants marked on Likert type scales anchored at 1 (Not at all) and 7 (Very much). The scale comprises practical incompatibility (items 1–3 as listed above) and identity incompatibility (items 4–8) sub components. Factor analysis supported this division of subscales and internal reliability (see supplementary materials on the OSF at https://osf.io/uta95/). Higher scores indicate greater incompatibility. Cronbach αs for subscales were both > 0.87.

2.3.3 Mental distress

The GAD7 and PHQ8 were used to measure mental wellbeing (Kroenke et al., 2009; Spitzer et al., 2006). These measures ask participants to indicate how often they had experienced different outcomes over the last 2 weeks. Items included ‘Not being able to stop or control worrying’, ‘Being so restless that it’s hard to sit still’, ‘Feeling tired or having little energy’, ‘Feeling bad about yourself — or that you are a failure or have let yourself or your family down’. Response options were Not at all (coded 0), Several days (+ 1), More than half the days (+ 2), Nearly every day (+ 3), The self-harm/suicide ideation question from the full PHQ9 scale was not included. In line with Kroenke et al. (2016), the GAD and PHQ scores were collapsed into a single score. The internal reliability was Cronbach α = 0.94. Higher scores indicate greater mental distress.

2.3.4 Experience of work

The scale devised by Cousins et al., (2004) and used by the Health and Safety Executive in the UK, was used to measure experience of work. This 35 item scale includes items such as ‘I am clear what is expected of me at work’, I am subject to personal harassment at work’, ‘I am clear what my duties and responsibilities are’, ‘I have to work very fast’, ‘I can rely on my line manager to help me out with a work problem’, and ‘My colleagues are willing to listen to my work related problems’. Participants are asked to ‘Please answer each question by indicating how much the experience applies to you’, with response options as Never (coded 1), Seldom (2), Sometimes (3), Often (4), Always (5). Dimensions of workplace experience captured comprise task demands, control (over content of work, pacing, etc), managerial support, work colleague support, role clarity, relationships and pace of change. Items are coded or reverse coded such that higher scores indicate a more positive experience. Cronbach α = 0.90.

2.3.5 Turnover intention

Turnover intention was measured using a scale outlined in Roodt (2004 cited Bonds, 2017). This asks participants to rate how often during the past nine months they had experienced each of a number of thoughts and feelings related to their job including; ‘How often have you considered leaving your job?’ (1 = never, 5 = always), ‘To what extent is your current job satisfying your personal needs?’ (1 = to no extent, 5 = to a large extent), ‘How often are you frustrated when not given the opportunity at work to achieve your personal work-related goals?’ (1 = never, 5 = always), ‘How often do you dream about getting another job that will better suit your personal needs?’ (as previous), ‘How likely are you to accept another job at the same compensation level should it be offered to you?’ (1 = highly unlikely, 5 = highly likely), and ‘How often do you look forward to another day at work?’ (1 = never, 5 = always). Items were scored such that higher scores indicate higher turnover intentions. Cronbach α = 0.81.

2.3.6 Demographics

Age, gender, academic discipline, ethnicity, perceived social capital, and details of current role (level, hours worked per week, contract type, estimated percentage of time spent on research, teaching, academic management). See supplementary materials (OSF, link above) for full details of scale wording.

2.4 Procedure

Participants were recruited from a combination of community/snowball recruitment (emails to colleagues in different disciplines, tweets, etc.) which made up 19% of the sample, and commissioned recruitment (via Qualtrics) for the remainder. Participants were given information and offered informed consent. They then completed an anonymous questionnaire online (via the Qualtrics platform) containing the measures above (presented in the order listed above). Finally, they were thanked and debriefed. All participants were paid an agreed incentive (~£10 depending on recruitment method) to take part.

2.5 Ethics and pre-registration

The project received ethical oversight from London South Bank University (application number ETH2122-0007). The methods and analysis plan were registered on the Open Science Framework (see link above) prior to any data being collected.

3 Results

3.1 Relationship between variables

As per the pre-registered analysis plan, to test H1 we conducted zero order correlations between practical and identity incompatibilities, mental distress, experience at work, turn over intentions, identification with being (i) an academic (ii) an educator and (iii) a researcher and number of hours worked a week. We took a bootstrap approach (n = 1000 samples). Simulation and conceptual work by methodologists has shown such an approach minimises the impact of possible skew or outlying / overly influential data within the sample (see Cribari-Neto & Zarkos, 1999; Efron, 1979). In this vein, the resulting 95% confidence intervals indicate the potential impact of overly influential data points (i.e. influential outliers). Means for each dimension along with correlations are shown in Table 1.

As can be seen in Table 1, levels of identification themselves did not link to mental distress, experience at work or turnover intentions. Practical incompatibility was positively related to levels of academic, educator and research identification. In contrast, identity incompatibility was related to mental distress. Mental distress was in turn linked with poorer experience at work and higher turnover intentions. Those who worked longer hours also tended to have higher identification with being an academic and a researcher (but no relationship with educator identity) and tended to experience more practical incompatibility.

Table 1 Descriptive statistics for and zero order correlations between key study variables
Full size table

3.2 Mediation analysis

To unpack the relationships observed in the correlations further, an exploratory mediation analysis testing the relationship between identity incompatibility and turnover intention, mediated by (i) mental distress and (ii) workplace experience was conducted (see Fig. 1). The Process macro model 4 was used testing the model described above using 5000 Bootstrapped samples. 95% confidence intervals are reported.

Fig. 1
figure 1

Mediation model tested. * = p <.05, **, p <.010, *** p <.001

Full size image

The overall model was significant, R2 =.34, F (3,137) = 23.74, p <.001. Turnover intention was not directly related to identity incompatibility, c’ = 0.01, t = 0.17, p =.87, CIs = -0.08, 0.09. Mental distress was directly related to turnover intention in the model, b1 = 0.23, t = 2.62, p =.010, CIs = 0.20, 0.59. Work experience was related to turnover intention, with better work experience linked to lower turnover intention, b2 = − 0.80, t = 6.81, p <.001, CIs = -1.03, -0.57. Identity incompatibility was related to mental distress, a1 = 0.11, t = 2.74, p =.007, CIs = 0.03, 0.20. Identity incompatibility was not related to workplace experience, a2 = − 0.05, t = 1.65, p =.10, CIs = − 0.11, 0.01. The indirect (mediated) effect of identity incompatibility on turnover intention via mental distress was significant, a1b1 = 0.03, CIs = 0.01, 0.06. The indirect effect of identity incompatibility on turnover intention via work experience was non-significant, a2b2 = 0.7, CIs = − 0.91, 10. In sum, a mediated relationship between identity incompatibility and turnover intention via mental distress was observed.

3.3 Moderation analysis

In line with our pre-registered analysis plan to test H2, H3 and H4, three moderation analyses which tested the relationships between identity incompatibility and (i) mental distress, (ii) experience at work and (iii) turnover intention were conducted. For each model the moderating influence of levels of identification as an academic and the difference in levels of identity between researcher and educator were tested. In each case, the overall model was not significant, ps > 0.198. H2-4 were not supported.

4 Discussion

In common with those in other occupations, academics hold multiple roles in their occupation which may form distinct social identities. These can be more or less incompatible in terms of both practicality and identity-based factors. The current study aimed to explore the impact of such incompatibilities in the context of work-related identities (specifically being an ‘educator’, a ‘researcher’ and an ‘academic’) on mental health, experience of the work environment and turnover intention. It was hypothesised that higher levels of incompatibility between educator and researcher identities would be linked with (i) poorer mental wellbeing, (ii) poorer experience of work and (iii) increased turn-over intention. It was also hypothesised these effects would be most pronounced amongst those who identify highly with being an academic, when both researcher and educator identities were similar in strength and when both academic identity is high and difference in identification between educator/researcher was high.

These hypotheses were partially supported. Positive relationships were observed between identity (but not practical) incompatibility and mental distress. Practical (but not identity) incompatibility was linked with absolute levels of identifications. Workplace experience and turnover intention were not directly related to levels of identification (with the exception of hours worked, which was higher amongst those with higher incompatibility). These relationships (or lack thereof) appeared constant at different levels of identification with each category. However, an exploratory analysis observed an indirect relationship between identity incompatibility and turnover intention mediated by workplace experience and mental health.

These findings have a number of theoretical and practical implications. On a theoretical level, the interactive nature of identities on outcomes highlights the importance of considering the role of a given identification in the context of others which are also salient to the individual in the same context. This reinforces a previously documented caveat to the observation that having multiple identities (or therapeutically generating new ones) is beneficial – specifically that it is the nature of the relations between valued identities which moderates effects. A related point is the absence of moderation effects. While caution needs to be exercised in the interpretation of null results, it should be noted that the effect was observed regardless of levels of identities. This may suggest a possibility that level of incompatibility is as important, if not more so, in predicting the outcome of having multiple identities, and this warrants further exploration. Whether incompatibility and compatibility mirror one another in their impacts in such contexts remains an open question.

The findings support the notion of different dimensions of incompatibility – one based on identity and one on practical issues. This was observed in Frings, Gleibs et al., (2019b), but not linked to differential outcomes. However, in contrast, no clear distinction between incompatibility types was observed in Frings et al. (2022). The current study observed two distinct factors of incompatibility and also observed disassociation (the factors related differentially to outcomes).

Considerable variance in both types of incompatibility was observed in the current study. An interesting next step for the current line of inquiry is to explore the extent to which incompatibilities are situationally and/or psychologically defined. Situational factors are likely to narrow the range of ways an identity can be experienced, and the relationships between them (for instance, the potential zero-sum relationships between teaching and research in academia). As both basic social identity perspectives (e.g., self categorisation theory, see Turner et al., 1987) and recent research (e.g., Frings et al., 2019b) show that identity contextualises experience, one intriguing possibility is that the formulations of each identity affect outcomes. For instance, if one’s identity as an academic contains elements of combining teaching and research, the ‘zero-sum’ belief may not hold sway, and incompatibility may (psychologically at least) be reduced. An interesting (but not hypothesised around) observation of the current study is a lack of direct association between identity and outcomes such as mental distress, turnover intention etc. Noting the interpretation of null effects (albeit in this case well powered zero order correlations), this contrasts with other work, summarised in a recent meta analysis (Steffens et al., 2017) and perhaps suggests work exploring if this population differs from others in ability to use workplace identity as a buffer from stress would be of value.

One motivation of the current study was to explore predictors of, and highlight potential changes to be made, around the issue of occupational stress amongst academics. From a practical perspective, the findings suggest a number of important implications for staff, managers and organisations. Axiomatically, mental health is linked to workplace experience and turnover intention. It is also likely that reducing incompatibility between job roles will improve mental health. One possible strategy for change is to encourage staff to recognise the competing roles of their job and be empowered to either reduce the incompatibility and/or be encouraged to have and hold boundaries around time allocated to each aspect of the role. However, the bulk of responsibility for such change arguably lies at a system or institutional level. This latter focus contrasts with common current emphasis which suggest providing ‘stress management’ tools (ranging from free exercise classes to employee assistance programs providing basic mental health support) to create a ‘resilient’ workforce. While these interventions are seemingly beneficial, they do not address the identity based (and arguably other) root causes of poor workplace experience in in complex role environments. Specifically, our current observation highlights that not only identity (internal, which can perhaps be partially self-managed) but also practical (external, which individuals have less autonomy over, at least in the short term) incompatibility factors are important and proposes a key role of organisations. The latter are strongly impacted by the organisational policies set and norms encouraged. For example, providing genuine opportunities for career progression via teaching, research or hybrid career paths (anecdotally more common in newer institutions with a greater teaching focus), adopting transparent and accurate workload modelling (an emerging trend in the sector) which ensure sufficient time to complete tasks (increasingly difficult under low resource conditions), combined with clearly defined outcome expectations may also serve to allow employees in academia to balance needs on both practical and identity dimensions. For those with outside of work responsibilities (or those who wish, not unreasonably, to keep to set contracted hours), a clear sense of when goals are achieved vs. when goals are not achieved may also support and achieve a balance between academic and outside identities (and the incompatibilities this raises). It should also be recognised that such changes at an organisational level may be hampered by perceptions of flexibility in working arrangements, workload management requirements etc., as well as by externally driven pressures placed on the organisation (for example, falling real world value of fixed student course fees).

The study has a number of limitations. The data are cross sectional, and as such no assumptions about causational direction can be evaluated. It also focussed on a sample of academics from the UK. This is important as cultural differences in the impact of multiple groups on wellbeing have been shown to vary cross-culturally e.g., Chang et al., (2016), and it is thus likely that incompatibility may also be experienced in different ways at a cultural group level. The sample also had subclinical levels depression and anxiety, which may limit generalisability to more sectors of the workforce with more intense symptomology. Finally, the sample was not sufficient to test if these differences change over the career journey (i.e. do people resolve incompatibilities over time if they stay in the profession).

In conclusion, the current study highlights the importance of incompatibility in academics job roles with identity-based factors being particularly important. This has theoretical implications around the role of multiple identifications and wellbeing, and points to practical ways to reduce mental distress and staff turnover in organisations where roles are complex.