On the nature, predictors, and outcomes of work passion profiles: A generalisability study across distinct types of employees

Abstract

Thus far, little research has adopted a person‐centred approach to investigate the nature of work passion profiles. As a result, our understanding of the most commonly occurring combinations of harmonious passion (HP) and obsessive passion (OP) in the workplace remains limited. To achieve a more refined understanding of the nature of these work passion profiles, our first aim was thus to identify the configurations of HP and OP for work observed among five samples, including 11 subsamples, of employees (N = 7258). Then, we also considered the extent to which these profiles and their associations with theoretically‐relevant predictors (work‐home segmentation and organisational support) and outcomes (work engagement, work‐family conflict, turnover intentions, presenteeism, and counterproductive work behaviours) generalised across all subsamples. We identified a total of five profiles with a structure that differed slightly across samples: High OP Dominant, High HP Dominant, Average HP Dominant, Low HP Dominant (i.e., low levels of passion dominated by higher levels of HP relative to OP), and Moderately Low Passion. The High OP Dominant profile was systematically the most prevalent (37.5%–54.1% of the sample), whereas the High HP Dominant was the least prevalent (2.1%–7.7%). Across all samples, work‐home segmentation was related to a higher likelihood of membership into the profiles characterised by higher, relative to lower or moderate, levels of passion (HP and OP), whereas organisational support also helped employees to stay away from the High OP Dominant profile. Lastly, the least desirable outcomes were observed in the High OP Dominant profile, whereas the most desirable outcomes were observed in the High HP Dominant profile. Interestingly, work engagement levels where comparable in these two profiles. Beyond their theoretical implications for research on work passion, these results highlight how work passion has highly similar implications across contexts.

1. INTRODUCTION

Prior studies have examined employees' passion for their work (Vallerand & Houlfort, 2019) based on the acknowledgement of its multiple benefits for organisational (e.g., Vergauwe et al., 2022) and individual (e.g., Pollack et al., 2020) functioning. The Dualistic Model of Passion (DMP; Vallerand, 2010, 2015; Vallerand et al., 2003) defines passion as a strong inclination towards a specific activity, such as work (Vallerand, 2015; Vallerand & Houlfort, 2019). Passionate workers invest substantial effort and time in their work, consider it central to their identity, love it, and regard it as important (Vallerand & Houlfort, 2019). However, not all forms of passion are equally desirable. The DMP (Vallerand, 2015; Vallerand & Houlfort, 2019) highlights the need to differentiate passion based on its harmonious (HP; a strong psychological investment in an activity [job] that is freely chosen by the individual) or obsessive (OP; strong psychological investment in an activity [job] that is driven by internal or external contingencies associated with the activity) nature. Research has reported well‐differentiated associations between these two forms of passion and a variety of predictors and outcomes (e.g., Breu & Yasseri, 2023; Laurent et al., 2023), showing that the benefits of passion were accompanied by undesirable consequences when it became obsessive (e.g., Gillet et al., 2023c; Schellenberg et al., 2019). However, although both types of passion have never been proposed as mutually exclusive (Gillet et al., 2023c), research on work passion has thus far primarily neglected the combined influence of both types of work passion.

Beyond recognising that some employees can jointly experience both types of passion (Vallerand, 2015), person‐centred investigations have started to investigate how harmonious passion (HP) and obsessive passion (OP) combine within distinct profiles of employees (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b). This approach is linked to the recent development of the Quadripartite Model of Passion (QMP; Schellenberg et al., 2019), which highlights the role of distinctive combinations of HP and OP. By considering the unique configurations of work passion present among distinct types, or profiles, of employees, this approach should help us to achieve a clearer understanding of the most optimal work passion profiles for employees and their organisations. For instance, is high OP as problematic when combined with similarly high HP? This approach should help shed light on the mechanisms (e.g., Perceived organisational support[POS]) involved in the development and maintenance of more or less desirable work passion profiles rather than considering how to separately stimulate HP while limiting OP.

However, investigations of the QMP have so far been mainly limited to non‐work domains (e.g., undergraduate students and video gamers: Schellenberg et al., 2019; students: Schellenberg et al., 2021a; athletes: Schellenberg et al., 2021b; marijuana users: Dolan et al., 2021), with a single variable‐centred study conducted among employees (Gillet et al., 2023c). Future QMP research is thus needed to document whether and how previous results may apply to this distinct and critical life context, involving engagement decisions potentially less driven by free choice than previous contexts (education, sports, recreational drug use) given the economic necessity of work for most people.

Theoretically underpinned by the DMP (Vallerand, 2015) and QMP (Schellenberg et al., 2019), this study aims to: (1) achieve a more refined person‐centred understanding of the work passion profiles observed among different samples of French employees; (2) document the construct validity of these profiles by examining their associations with theoretically‐relevant predictors and outcomes; and (3) determine whether the nature of the profiles, as well as their associations with predictors and outcomes, differ as a function of various work characteristics. The three research questions guiding this study are: (a) Can distinct work passion profiles be identified, and are these profiles consistent with the predictions of the DMP (Vallerand, 2015) and QMP (Schellenberg et al., 2019) and with previous research findings (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b)? (b) Will the strength and direction of the associations between work passion profiles, predictors, and outcomes align with theoretical expectations? (c) To which extent will the results to the two previous questions generalize to employees working in various work settings?

This research seeks to contribute to the research literature on work passion in four important ways. First, to achieve a better alignment with the person‐centred propositions of the QMP (Schellenberg et al., 2019) than previous variable‐centred studies (Schellenberg et al., 2021a, 2021b), we rely on a person‐centred approach to identify distinct profiles of employees experiencing different configurations of HP and OP. Variable‐centred studies focus on the isolated, additive, or interactive associations between both types of passion, predictors, and outcomes, assuming these relations to generalize to the whole sample. More specifically, variable‐centred analyses assume that all participants come from the same population for which results can be summarised by a set of “average” parameters. In contrast, person‐centred analyses (Morin et al., 2018) identify subpopulations of workers presenting qualitatively distinct configurations of HP and OP, and is thus entirely consistent with the theoretical underpinnings of the QMP (Schellenberg et al., 2019) which also expects employees to display distinctive configurations of HP and OP. However, with few exceptions (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b), all previous investigations of this model have relied on a [suboptimal] variable‐centred approach. A person‐centred approach allows for testing whether workers will correspond to one of the four theoretical configurations highlighted in the QMP, rather than simply assuming that they will, providing a more stringent test of validity. Moreover, it provides a way to go beyond QMP predictions by allowing for the identification of profiles characterised by moderate (or moderately high or low) levels of HP and/or OP rather than simply assuming that these configurations will necessarily entail high or low levels of HP and/or OP. This matches prior studies showing that employees tend to more frequently display moderate, rather than high or low, levels of work passion (e.g., Birkeland & Buch, 2015; Breu & Yasseri, 2023). This is also in line with Li et al.’s (2020) results, who identified work passion profiles characterised by moderate levels of HP and OP. Identifying employees' work passion profiles should enable us to provide validity evidence for the QMP (Schellenberg et al., 2019) in the work context, as well as to consider the possible existence of additional work passion profiles not highlighted in this model.

Second, we investigate whether the work passion profiles will be replicated across 11 subsamples of French employees: (a) working in the administrative, nursing, or sales domains (Sample 1); (b) working full‐time or part‐time (Sample 2); (c) working in the private or public sector (Sample 3); (d) enrolled in a permanent or temporary position (Sample 4); and (e) with a high (experienced employees) or low (novice employees) tenure in their position (Sample 5). In this way, we can find out whether the identified profiles are generalisable to the work context in the broadest sense of the term, or whether work will be associated with unique work passion profiles. From an applied standpoint, this is an important contribution, as it enables us to envisage whether or not work passion interventions developed in other contexts could be generalised to employees.

Third, we replicate and extend the preliminary findings of the few other person‐centred QMP studies (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b) which has investigated the work passion profiles of workers from different cultures (i.e., Gillet, Morin, et al., 2023: British Isles and United States; Li et al., 2020: China; and Morin et al., 2023b: Indigenous vs. non‐Indigenous Australians) by relying on other (i.e., French) samples of employees. Person‐centred research is cumulative, so that multiple studies are needed to differentiate the core profiles that emerge across all situations, the additional ones that only emerge in some situations, and the idiosyncratic profiles that reflect random sampling variations (Morin et al., 2016b). There is thus a need for additional QMP investigations of the generalisability of work passion profiles, especially in a Western cultural context—which differs substantially from the cultural contexts (e.g., China) considered in the few previous studies of work passion profiles (e.g., Li et al., 2020). For instance, in Western cultures, the emphasis is often placed on individual rights and individualism, whereas in China's collectivist culture, the smooth running of work teams and organisations is usually considered as more important than individual considerations (Brew et al., 2011). These cultural differences could have a significant influence on the nature, development, predictors, and outcomes of employees' work passion profiles. Although few studies have focused on cross‐cultural similarities or differences in work passion (O’Keefe et al., 2022; Weng et al., 2022), and none with a specific focus on French employees, limited evidence suggests that past results (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b) may generalize to French employees. For instance, Slemp et al.’s (2021) results (from a model including job crafting, autonomy support, work passion, engagement, and burnout) generalised to Australian and Chinese employees.

Fourth, we expand upon prior findings (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b) by considering predictors and outcomes not previously examined. More precisely, these authors only considered a limited set of predictors (e.g., work centrality, role ambiguity) and outcomes (e.g., work‐family conflict, counterproductive work behaviours, resilience at work, family life satisfaction) associated with these profiles, highlighting the need for a more comprehensive understanding of predictors and outcomes. In the present study, we investigate the role of employees' perceptions of organisational support (i.e., employees' perceptions that their organisation values their contribution and cares about their well‐being; Eisenberger et al., 1986) and work‐home segmentation (i.e., the presence of clear physical, temporal, and behavioural boundaries between employees' professional and personal roles; Kreiner, 2006) as predictors of profile membership, as well as the implications of these profiles for employees' levels of: (a) Work engagement (i.e., “a positive, fulfilling, work‐related state of mind that is characterised by vigour, dedication, and absorption”; Schaufeli et al., 2002, p. 74); (b) work‐family conflict (i.e., “a form of interrole conflict in which the role pressures from the work and family domains are mutually incompatible in some respect”; Greenhaus & Beutell, 1985, p. 77); (c) turnover intentions (i.e., employees' self‐reported intentions of leaving their organisation in the near future; Jaros, 1997); (d) presenteeism (i.e., employees' report of having attended their work when their state of health prevented them from working properly; Miraglia & Johns, 2016); and (e) counterproductive work behaviours (i.e., workplace behaviours that are intentional, deliberate, and detrimental to the functioning of the organisation and its members; Spector et al., 2010). These outcomes were selected to replicate while also expanding upon prior research. These predictors are likely to play a role in driving employees to allocate ―willingly or not― more or less of their energy and resources to their work role (Hobfoll, 2011), and can thus be theoretically expected to play a role in the emergence of specific work passion profiles (Vallerand, 2015; Vallerand & Houlfort, 2019).

Similarly, all outcomes considered in this research are closely linked to the preservation or deterioration of employees' resources (Hobfoll, 2011). For instance, some employees may intend to leave their organisation because they feel no longer having sufficient resources to cope with their job demands. Leaving the organisation may then be a way for them to preserve their few remaining resources, while hoping that their new work environment will enable them to rebuild their resources. In addition, these outcomes are known to be highly relevant to employees' professional success and ability to experience satisfactory career trajectories. Thus, work engagement is recognised as a key driver of job performance, well‐being, and positive functioning (e.g., Meyer & Schneider, 2021) and helps workers better cope with the demands of their work (Schaufeli et al., 2019). In contrast, work‐family conflict has often been related to reduced levels of performance because it takes a toll on employees' personal life, thus encouraging them to withdraw from their work role to avoid further losses of resources (Huyghebaert‐Zouaghi et al., 2022). Turnover intentions have long been recognised as a direct, and critically important, precursor of a wide variety of undesirable work outcomes (e.g., actual turnover: Fukui et al., 2019; reduced performance: Haque, 2021), whereas presenteeism is known to share negative associations with learning, performance, executive functioning, and social skills (e.g., Miraglia & Johns, 2016). Finally, counterproductive work behaviours are associated with negative outcomes (e.g., lower job satisfaction and work engagement; Gillet et al., 2023b), in addition to reflecting volitional behaviours seeking to harm organisations or organisation members.

We finally investigate the generalisability of these associations between predictors, work passion profiles, and outcomes across all subsamples of French employees. By providing evidence of generalisability (or variability), we seek to inform research in relation to the nature, predictors, and outcomes of work passion profiles, and to guide practice by highlighting the extent to which generic interventions are likely to yield similar benefits for different types of workers. Importantly, person‐centred results are naturally aligned with managers and practitioners' tendency to think about employees in terms of categories (person‐centred) than in terms of complex associations among variables (variable‐centred; Morin et al., 2011). Our findings may thus have important implications for practice at a time when many organisations are rethinking the way they can preserve and enhance employees' workplace well‐being, motivation, and performance. For instance, documenting the outcomes of work passion profiles should help decide which profiles should be prioritised for intervention, whereas identifying predictors should help identify actionable levers of intervention.

1.1. A person‐centred perspective on work passion

The DMP (Vallerand, 2015; Vallerand et al., 2003) posits that a passion may be harmonious or obsessive depending on how it is internalised into one's identity. HP workers freely choose to engage in the work that they love, which has come to be autonomously internalised into their identity. Work occupies an important, but not overpowering, place in their identity. As a result, HP for work represents a strong, but controllable, motivational force that can exist in harmony with other facets of employees' lives (Vallerand & Houlfort, 2019). In contrast, although OP employees also love their work, they feel pressured to engage in it as a result of internal or external pressures and contingencies (Vallerand, 2015). Such contingencies might be a boost of self‐esteem, social approval, or the avoidance of negative emotions (Lafrenière et al., 2011). As a result of this more externally‐driven form of internalisation, OP entails an uncontrollable urge to work (Houlfort et al., 2018), leading obsessively passionate workers to struggle with establishing adaptive boundaries between work and other life domains (Vallerand & Houlfort, 2019).

The DMP (Vallerand, 2015) explicitly positions HP and OP on two separate continua, thus acknowledging the possibility that some workers may simultaneously experience high levels of OP and HP, while others may experience only one, or neither, form of passion. From this perspective, the QMP (Schellenberg et al., 2019) proposes to differentiate among four prototypical configurations of passion: OP Dominant (low HP; high OP), HP Dominant (high HP; low OP), Mixed Passion (high HP; high OP), and Low Passion (low HP; low OP). This representation highlights the need to move beyond the examination of the additive and independent effects of HP and OP to consider their combined role. For example, is it better to display no passion at all for one's work (i.e., Low Passion) than to be obsessively passionate about it (i.e., OP Dominant)? Are workers protected against the negative effects of high OP when they also display high HP (i.e., Mixed Passion vs. OP Dominant)? Lastly, is it always better to display high HP coupled with low OP (i.e., HP Dominant), or are there times, contexts, or settings in which it is more adaptive to also display high OP (i.e., Mixed Passion)? Indeed, it is quite conceivable that an employee could be characterised by high levels of both HP and OP. This is not necessarily because both types of passion are simultaneously present at the same time in the same individual, but rather because employees may have different levels of passion at work, depending on the period and tasks involved. For instance, the same employee could very well display obsessive behaviour during a period of intense pressure to meet deadlines, and much more harmonious behaviour at another time, when the pressure to achieve results is less acute.

Li et al. (2020) identified three work passion profiles among two samples of Chinese employees: (1) Mixed Passion (high HP and OP; 38% of sample 1% and 59% of sample 2); (2) HP Dominant (moderate HP and low OP; 60% of sample 1% and 29% of sample 2); and (3) OP Dominant (low HP and moderate OP; 2% of sample 1% and 12% of sample 2). Additional studies identified three similar profiles coupled with a Low Passion profile, respectively among a sample of workers from the United States and the British Isles (Gillet, Morin, et al., 2023) and across subsamples of Indigenous and non‐Indigenous Australian workers (Morin et al., 2023b). In this study, we assess the generalisability of past results (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b) within five samples, and 11 subsamples, of French employees. Based on prior findings and the QMP, we expect that:

Hypothesis 1

At least four profiles will be identified: (1) Low Passion (Low HP and OP); (2) Mixed Passion (High HP and OP); (3) HP Dominant (High HP and Low OP); and (4) OP Dominant (Low HP and High OP).

1.2. Predictors of profile membership

According to the conservation of resources theory (Hobfoll, 2011) and work‐home resources model (ten Brummelhuis & Bakker, 2012), POS and work‐home segmentation can be seen as core drivers of intense forms of work involvement such as passion. More precisely, POS seeks to help workers face contextual demands while retaining their personal resources (Gillet et al., 2018b; Morin et al., 2023a), Managers and practitioners can easily act to improve employees' perceptions of organisational support to help improve employees' functioning (Caesens et al., 2020). Similarly, work‐home segmentation is known to represent another critical driver of job performance, work engagement, and well‐being (Kreiner, 2006; Kubicek & Tement, 2016). Work‐home segmentation helps build, accumulate, and recover their personal work‐related resources as a result of the higher levels of psychological detachment they are able to experience during their off‐job time (Derks et al., 2014). As these work‐related resources accumulate, they are likely to become available to support employees in meeting their job and family demands (ten Brummelhuis & Bakker, 2012).

1.3. Perceived organisational support

Meta‐analyses support the role of POS as a positive driver of individual and organisational outcomes (Kurtessis et al., 2017; Riggle et al., 2009). Perceived organisational support helps workers maintain and develop work‐related resources (Hobfoll, 2011), thereby helping them recover more quickly from work and protecting them against the undesirable effects of job and family demands (Spurk et al., 2016; ten Brummelhuis & Bakker, 2012). Moreover, employees who feel supported by their organisation tend to display higher levels of work engagement (Caesens et al., 2014)—a construct sharing commonalities with HP (Tóth‐Király et al., 2021)—while being less likely to rely on destructive forms of overinvestment (e.g., OP, workaholism; Spurk et al., 2016). Perceived organisational support is also associated with higher levels of autonomous motivation and lower levels of controlled motivation (e.g., Gillet et al., 2013). Yet, HP is anchored in autonomous forms of motivation (i.e., motivated desire, interest, or choice), whereas OP is anchored in more controlled forms of motivation (i.e., internal and/or external pressures; Vallerand, 2015). By enabling employees to experience enjoyable and satisfying work experiences, POS may thus facilitate the development of HP. In contrast, employees perceiving high levels of organisational support should not feel a sense of obligation that characterises OP. Perceived organisational support is also incompatible with a complete lack of investment and energy that characterises non‐passionate employees (Schellenberg et al., 2019). For these reasons, POS should increase the likelihood of membership into a profile characterised by high levels of HP and low levels of OP (HP Dominant profile) compared to one with low levels of HP coupled with high (OP Dominant profile) or low (Low Passion profile) levels of OP.

However, POS can also decrease employees' personal resources (Caesens et al., 2023) by encouraging them to repay their organisations (i.e., Kurtessis et al., 2017) for its support by increasing their level of work investment beyond what might be reasonable (Gillet, Morin, et al., 2017), making it harder for them to detach from work (Gillet et al., 2021). Thus, by creating an external contingency to employees' willingness to dedicate themselves to their work, POS can set the stage for the emergence of OP (Vallerand, 2015; Vallerand et al., 2003). In such situations, employees may display high levels of OP as they experience a controlled internalisation of the work activity in their identity (Vallerand, 2015). Moreover, for obsessively passionate employees, work is a very important source of self‐esteem. These employees are thus highly sensitive to all work‐related factors that could be associated with an increase in their self‐worth (Mageau et al., 2011). Workers perceiving high levels of organisational support may feel that their organisation values them, counts on them, and that their work is appreciated (Kurtessis et al., 2017). All these elements may lead to an increase in self‐esteem, but also indirectly foster the development of OP. Interestingly, although past studies have generally conceptualised organisational support as a predictor of adaptive outcomes in a “the more, the better” perspective (Caesens et al., 2014), recent results provide a more nuanced picture, suggesting that high levels of organisational support may also be harmful in a “too much of a good thing” manner (Caesens et al., 2023; Harris & Kacmar, 2006).

To the best of our knowledge, no research has yet examined the effects of POS on work passion. However, Birkeland et al. (2018) showed that perceived supervisor support was positively related to HP and negatively associated with OP. Kong and Ho (2018) also found a positive effect of leader‐member exchange (a construct encompassing perceptions of leader support) on HP and OP, even if the former relationship was stronger than the latter. Likewise, Gillet et al. (2017b, Gillet et al., 2016, 2022) found positive associations between leader‐member exchange and employees' likelihood of displaying higher levels of workaholism, a construct closely related to OP (Birkeland & Buch, 2015) which also entails a high level of work investment (Schaufeli et al., 2009). Thus, we expect that POS should increase the likelihood of membership into a profile with high levels of HP and low levels of OP (HP Dominant profile) compared to a profile with high levels of HP and OP (Mixed Passion profile), a profile with low levels of HP and high levels of OP (OP Dominant profile), and a profile with low levels of HP and OP (Low Passion profile). In addition, POS should increase the likelihood of membership into the Mixed Passion and OP Dominant profiles compared to the Low Passion profile.

Hypothesis 2

Perceived organizational support will be associated with a higher likelihood of membership into the HP Dominant profile relative to the OP Dominant, Mixed Passion, and Low Passion profiles, as well as into the OP Dominant and Mixed Passion profiles relative to the Low Passion profile.

1.4. Work‐home segmentation

Organisations vary in the degree to which they expose employees to an environment that promotes the segmentation or integration between employees' professional and personal roles (Kreiner, 2006). By pressuring employees to maintain a persistently high level of work investment, work‐home integration (i.e., a lack of boundaries between roles) leads to decreases in employees' positive feelings towards work by making it harder for them to recover and to invest their personal resources towards the fulfilment of their personal goals (Althammer et al., 2021). As a result, organisations that force or aggressively promote work‐home integration (e.g., mandatory cell phone wearing, mandatory availabilities on weekends) place strong external contingencies upon their employees, thereby increasing their likelihood of developing OP (Kreiner, 2006; Vallerand, 2015). More generally, employees facing high job demands tend to spend an excessive amount of time and effort in their work at the expense of their family life, have difficulties disengaging from work, experience negative mood when prevented from working, and remain preoccupied with their work at home (Gillet et al., 2017b; ten Brummelhuis & Bakker, 2012). These consequences of job demands are defining characteristics of OP (Vallerand, 2015), suggesting that experiencing the former may lead to the latter.

Conversely, organisations facilitating work‐home segmentation (e.g., by clearly defining work hours and by discouraging work‐related communications occurring during evenings or weekends), make it easier for employees to do their work without experiencing such a persistent drain of their personal resources. As a result, employees have more flexibility and may more easily fulfil their various personal roles (Althammer et al., 2021). Working then becomes internalised in a more autonomous and voluntary manner, thereby increasing employees' likelihood of experiencing HP (Vallerand, 2015). More generally, employees who benefit from high work‐home segmentation are thus more likely to build psychological, personal, and social resources within their work and nonwork domains (Xanthopoulou et al., 2007). In turn, it becomes possible to capitalize on these resources to meet challenges across multiple life settings, thereby helping employees increase the quality of their work recovery and to cope with job and family demands (Sonnentag & Fritz, 2015; ten Brummelhuis & Bakker, 2012). We thus expect that work‐home segmentation should increase the likelihood of membership into profiles characterised by high levels of HP (HP Dominant and Mixed Passion profiles) compared to profiles with lower levels of HP (OP Dominant and Low Passion profiles).

Hypothesis 3

Work‐home segmentation will be associated with a higher likelihood of membership into the HP Dominant and Mixed Passion profiles relative to the OP Dominant and Low Passion profiles.

1.5. Outcomes of profile membership

The core theoretical models underpinning this study (i.e., the DMP: Vallerand, 2015; Vallerand & Houlfort, 2019; the QMP: Schellenberg et al., 2019) note that passionate workers allocate more resources to their work activity in a more (HP) or less (OP) balanced manner, while gaining personal resources as a result of the fulfilment and enjoyment derived from spending time in an activity about which they are passionate. It is thus not surprising that HP (Forest et al., 2012; Houlfort et al., 2018), as well as membership into a HP Dominant profile (Gillet, Morin, et al., 2023; Schellenberg et al., 2019, 2021b), tend to be associated with a variety of positive outcomes encompassing the professional and personal domains (Vallerand & Houlfort, 2019). In contrast, OP (Fernet et al., 2014; Pollack et al., 2020) and membership into an OP Dominant profile (Gillet, Morin, et al., 2023; Schellenberg et al., 2019, 2021b) tend to be associated with generally less desirable outcomes (Vallerand & Houlfort, 2019).

When we turn our attention more specifically to the outcomes considered in this study, we first note that because passionate workers (HP and/or OP) tend to invest a lot of time, effort, and energy in their work, they also tend to display high levels of work engagement (Birkeland & Buch, 2015; Tóth‐Király et al., 2021). However, this high level of engagement does not need to be negative, as some employees' engagement stems from the experience of enjoyable and satisfying work experiences (Schaufeli et al., 2009, 2019), thus sharing natural connections with HP (Vallerand & Houlfort, 2019). In contrast, work engagement may also stem from feelings of obligation, albeit linked to enjoyable activities, and thus also share natural connections with OP (Vallerand & Houlfort, 2019). In contrast, employees with no passion (Low Passion), because they do not enjoy their work and do not feel the need to be productive, are unlikely to display high levels of engagement (Vallerand, 2010). We thus expect that the Low Passion profile would be associated with the lowest levels of work engagement.

Regularly engaging in an activity about which we are passionate (e.g., work), even if anchored in OP, may lead to higher levels of functioning relative to that of employees without a passion in their lives (Schellenberg et al., 2019). Though OP is often linked to detrimental consequences (Vallerand & Houlfort, 2019), Amiot et al. (2006) found that it could be associated with psychological adjustment in competitive environments, whereas Lafrenière et al. (2009) showed that it was positively related to life satisfaction following success in one's passionate activity. In contrast, non‐passionate employees are unable to experience these periodic boosts in adjustment resulting from passion (Schellenberg et al., 2019). These results do not suggest that an OP Dominant profile will necessarily be desirable. Rather, they suggest that it might be preferable to be passionate about one's work, no matter the dominant type of passion, than to experience a complete lack of passion for it (Gillet, Morin, et al., 2023). Schellenberg et al. (2021b) demonstrated that non‐passionate individuals displayed the lowest levels of goal attainment. Morin et al. (2023b) also found that the Low Passion profile displayed the lowest levels of psychological well‐being and resilience at work. Yet, employees' psychological well‐being and health are closely related not only to work‐family conflict (e.g., Huyghebaert‐Zouaghi et al., 2022), but also to work behaviours (e.g., turnover, presenteeism, counterproductive behaviours; Gillet et al., 2023b). Thus, we propose that:

Hypothesis 4

The Low Passion profile will be associated with lower work engagement and higher work‐family conflict, turnover intentions, presenteeism, and counterproductive work behaviours than the three other profiles.

Schellenberg et al. (2019, 2021b) further demonstrated that workers reporting the highest levels of HP tended to experience more positive outcomes (e.g., physical health and psychological well‐being) than those reporting lower levels of HP. They also found that employees with the highest levels of OP tended to experience more negative outcomes (e.g., physical symptoms) than those with low levels of OP (see Gillet, Morin, et al., 2023; Morin et al., 2023b for similar results). From a theoretical perspective, for employees with a HP Dominant profile, work can typically co‐exist harmoniously with other facets of their life (Vallerand, 2010, 2015). This harmonious co‐existence allows them to establish adaptive boundaries between their work and other life areas, thus reducing their risk of experiencing conflicts between their personal and professional lives, in turn promoting their psychological well‐being and functioning at work (Vallerand & Houlfort, 2019). HP should also help employees to build up, or accumulate, work‐related resources as a result of the enjoyment (e.g., well‐being; Vallerand et al., 2003) they derive from working (e.g., better health, more positive mood, greater knowledge and skills; Hobfoll, 2011). As these resources accumulate, they become available to support the demands of employees' personal lives, allowing them to experience positive outcomes not only at work (e.g., high work engagement, low presenteeism) but also in their life in general (e.g., low work‐family conflict; Vallerand, 2015; Vallerand & Houlfort, 2019). Interestingly, Gong et al. (2020) showed that the negative effects of HP on turnover intentions were mediated by psychological well‐being. Moreover, harmoniously passionate employees tend to behave more appropriately in the work context (e.g., more supportive behaviours towards others than behaviours aimed at harming others' or organisational interests) because they do not feel constantly judged and evaluated, and their self‐esteem is not solely linked to their work achievements (Mageau et al., 2011). In other words, they do not feel pressured, but free to act (Vallerand, 2015). Their sense of competence and identity are thus not under constant threat, and harmoniously passionate employees do not feel the need to harm or diminish others to demonstrate their value and maintain a good self‐image (Gillet et al., 2023c).

In contrast, OP Dominant employees engage in their work with a rigid persistence, making it harder to establish boundaries between work and other life domains, thereby increasing the likelihood that work will interfere with other life areas and well‐being (Gillet, Morin, et al., 2023; Houlfort et al., 2018). These employees are thus less likely to accumulate personal resources to support their work. Likewise, because they tend to spend most of their personal resources at work, employees with high levels of OP should be more likely to adopt defencive strategies to protect themselves from further loss of resources in other life domains (Hobfoll, 2011). As a result, they are more likely to experience the demands of their personal life as a threat to their work functioning. By prioritising their work role obsessively, these employees should be less willing to capitalize on resources gained at work to support their psychological functioning, leading them to experience negative work outcomes (e.g., decreasing work engagement) and at the work‐family interface (e.g., high work‐family conflict). In addition, obsessively passionate employees have a constant need to feel valued and to demonstrate their superiority, which can translate into more counterproductive behaviours towards others (e.g., aggressiveness, disrespect) especially if they feel their self‐esteem is threatened (Gillet et al., 2023c; Mageau et al., 2011). They also devote an inordinate amount of time to their work, which is their core source of self‐esteem (Vallerand, 2010), and are thus determined to continue working even if their state of health does not allow them to (i.e., high presenteeism). Finally, OP may be associated with high levels of turnover intentions due to the fatigue, irritability, and tension experienced by obsessively passionate employees (Vallerand & Houlfort, 2019).

We also expect workers displaying high levels of HP and OP (Mixed Passion) to experience some of the benefits of HP, but without the detrimental outcomes of OP. Indeed, HP is theoretically associated with positive emotions (Vallerand, 2015; Vallerand & Houlfort, 2019) that help increase workers' well‐being and functioning by attenuating the undesirable effects of negative emotions. In doing so, the presence of HP, alone or in combination with OP, should lead to a more adaptive functioning than an OP Dominant profile, which is consistent with the protective role of HP identified in previous research (Gillet, Morin, et al., 2023; Schellenberg et al., 2019). We thus expect that:

Hypothesis 5

The HP Dominant profile will display higher work engagement and lower work‐family conflict, turnover intentions, presenteeism, and counterproductive work behaviours than the OP Dominant and Mixed Passion profile.

Hypothesis 6

The OP Dominant profile will display lower work engagement and higher work‐family conflict, turnover intentions, presenteeism, and counterproductive work behaviours than the Mixed Passion profile.

2. METHOD

2.1. Participants and procedure

This study relies on data collected, using the same procedures but different data collections, among five independent (i.e., non‐overlapping) samples of employees. More specifically, across all samples, we recruited participants aged between 18 and 70 and currently employed by a French organisation (self‐employed workers were excluded as we assessed employees' perceptions of organisational support). In Sample 1, participants also had to work as administrative workers, nurses or sales workers. In each of these five convenience samples (described shortly), workers from various French organisations were first approached by research assistants using social networks. Before completing the questionnaire (10 min), participants were provided online information about the objectives of the research, were informed that participation was voluntary and confidential, and were ensured that they would be able to freely withdraw from the study at any time. All participants who agreed to participate completed an online questionnaire. Written informed consent (i.e., clicking “agree”) was obtained from all individual participants involved in the study. More generally, all procedures implemented in the present research follow the ethical standards and principals of the Declaration of Helsinki (World Medical Association, 2013). No incentive was offered to take part in the study.

Our primarily goal in relying on five distinct samples is to assess the generalisability of our results. However, as these samples remained mainly comparable to one another, we wanted to conduct an even more robust test of generalisability. Based on previous research having uncovered variations in employees' work passion as a function of work settings (Fernet et al., 2014; Vallerand & Houlfort, 2019), we thus decided to focus on possible differences related to work characteristics presenting at least some theoretical or empirical links with work passion or with how it relates to our predictors or outcomes. More precisely, we examine whether work passion profiles and their associations with predictors and outcomes generalize to employees working: (a) in the sales, administrative, or nursing domains (Sample 1, including 1908 employees); (b) full‐time or part‐time (Sample 2, including 1298 employees); (c) in the private or public sector (Sample 3, including 1609 employees); (d) in a permanent or temporary position (Sample 4, including 1211 employees); and (e) in their current job for at least 20 years (experienced employees) or less than one year (novice employees) (Sample 5, including 1232 employees). Additional demographic characteristics are reported in Table 1 for all subsamples.

TABLE 1.

Demographic characteristics of all subsamples.

	Sample 1			Sample 2		Sample 3		Sample 4		Sample 5
	A: Administrative	B: Nurses	C: Sales	A: Full‐time	B: Part‐time	A: Private	B: Public	A: Permanent	B: Temporary	A: Experienced	B: Novice
Sample size	N = 460	N = 850	N = 598	N = 673	N = 625	N = 675	N = 934	N = 642	N = 569	N = 519	N = 713
Sex
% Female	93.7%	94.8%	69.1%	60.5%	75.8%	66.5%	66.4%	69.6%	68.2%	59.7%	54.5%
Age
M years	43.87	38.16	33.33	33.06	28.72	34.99	43.37	38.68	26.41	51.44	27.52
SD years	10.81	11.12	12.76	11.12	12.96	11.97	10.54	11.69	9.28	5.71	10.10
Education
No diploma	3.0%	0.5%	2.7%	2.4%	2.2%	3.1%	1.8%	2.5%	2.5%	4.2%	1.5%
Vocational	20.0%	17.2%	15.9%	12.2%	13.1%	20.4%	17.7%	15.4%	10.7%	38.7%	16.2%
Secondary	41.7%	22.4%	39.0%	24.5%	52.8%	27.1%	27.8%	23.2%	34.3%	20.8%	38.9%
University	35.2%	60.0%	42.5%	60.9%	31.8%	49.3%	52.7%	58.9%	52.5%	36.2%	43.3%
Position
Permanent	84.8%	82.2%	82.1%	75.2%	67.4%	78.2%	100%	100%	0%	97.4%	64.0%
Full‐time	80.7%	83.4%	75.1%	100%	0%	87.3%	100%	56.5%	87.3%	81.3%	86.8%
Supervisory	10.9%	18.0%	19.6%	31.9%	9.9%	16.7%	40.7%	19.3%	14.6%	23.1%	28.7%
Private sector	89.1%	82.0%	80.4%	94.1%	88.0%	100%	0%	87.7%	78.4%	83.0%	92.6%
Tenure (organisation)
M years	12.57	9.64	6.25	4.52	3.36	6.33	13.96	9.29	1.38	25.93	2.19
SD years	11.31	9.26	8.07	6.66	5.79	7.36	10.33	8.75	1.93	7.55	4.75
Tenure (position)
M years	7.96	6.28	4.81	1.94	2.50	4.13	5.84	5.20	1.17	25.08	0.50
SD years	8.55	7.00	6.56	2.18	3.52	4.89	4.89	4.85	1.73	5.55	0.15
Profiles
High HP dominant	(2) 7.6%	(2) 7.6%	(2) 7.6%	(3) 2.1%	(3) 2.9%	(3) 4.1%	(3) 4.1%	(4) 7.7%	(4) 7.7%	(3) 3.6%	(3) 3.6%
High OP dominant	(1) 44.1%	(1) 44.1%	(1) 44.1%	(4) 54.1%	(4) 37.5%	(2) 50.0%	(2) 50.0%	(2) 42.8%	(2) 42.8%	(4) 47.9%	(4) 47.9%
Moderately low passion		(4) 34.5%		(1) 30.6%	(1) 35.1%
Average HP dominant	(4) 34.5%		(4) 34.5%	(2) 13.2%	(2) 24.5%	(1) 16.5%	(1) 16.5%	(1) 16.0%	(1) 16.0%	(2) 31.8%	(2) 31.8%
Low HP dominant	(3) 13.8%	(3) 13.8%	(3) 13.8%			(4) 29.5%	(4) 29.5%	(3) 33.6%	(3) 33.6%	(1) 17.2%	(1) 17.2%

Abbreviations: HP, harmonious passion; M, mean; OP, obsessive passion; SD, standard deviation.

Our decision to contrast our results as a function of these specific characteristics (domain, full‐time/part‐time, sector, permanent/temporary, and tenure) is predicated on the following considerations. First, Carver and Scheier's (1990) control theory suggests that employees' functioning and work‐family balance could be impaired in a setting (i.e., nursing) which limits their ability to work efficiently due to insufficient access to the materials and support required for their work, inadequate staffing, and emotional and physical demands of caring for others (Gillet et al., 2020). For these reasons, nurses may feel the need to devote more time to their work to ensure the quality and safety of patient care, while at the same time feeling frustration and resignation about a healthcare system that does not live up to their expectations and does not allow them to work properly (Wang et al., 2016). For these reasons, the nursing context could amplify the undesirable effects of OP (e.g., Lapalme & Guerrero, 2019), while also decreasing the likelihood of developing a strong passion for work. In contrast, the higher person‐environment fit experienced by many other employees (e.g., sales and administrative workers) could possibly decrease the undesirable effects of OP by helping them maintain clearer boundaries between the work and family domains (De Gieter et al., 2022) while also increasing their work engagement (e.g., Huyghebaert‐Zouaghi et al., 2022).

Second, research also shows that temporary employees often feel more pressure to demonstrate their value to the organisation, with the hope of becoming permanent, which leads them to invest more time to their work than they should, often at the expense of personal life (Chambel & Farina, 2015). This context may thus be more favourable to the emergence of OP relative to HP (Mauno & Ruokolainen, 2017). In contrast, by spending less time at work than full‐time employees, part‐time employees should be less likely to develop high levels of passion for their work (HP and OP; Chambel et al., 2017). Third, public sector employees may be less sensitive to the impact of POS, and more sensitive to that of other sources of support (i.e., colleagues and supervisors; Caesens et al., 2023), relative to private sector employees, who also tended to be more strongly attached or identified to their organisation (Brunetto et al., 2010). Fourth, more experienced employees have had more time to develop a work routine and a variety of work habits allowing them to maintain a better balance between their professional and personal lives (Bradley, 2007), which may make them less sensitive to the impact of work‐home segmentation, and protect them against some of the undesirable effects of OP, relative to novice employees (Spurk et al., 2019).

2.2. Measures

All questionnaires were self‐reported and administered in French. With one exception (i.e., work‐home segmentation), these questionnaires were either originally created in French, or had been previously validated in French and found to display psychometric properties comparable to those of the original versions (Fouquereau et al., 2019; Gillet et al., 2018a; Huyghebaert et al., 2018b; Huyghebaert‐Zouaghi et al., 2021). The work‐home segmentation measure was adapted to French using a classical translation back‐translation procedure involving independent bilingual translators and members of the research team.

Work passion (profile indicators) was assessed using Philippe et al.’s (2017; originally developed in French) questionnaire covering HP (three items; e.g., Work is in harmony with the other things that are part of me; α = 0.70–0.91 in Philippe et al., 2017; α = 0.78–0.84 across the present samples) and OP (three items; e.g., I have almost an obsessive feeling for work; α = 0.69–0.85 in Philippe et al., 2017; α = 0.70–0.72 across the present samples). Items were rated on a seven‐point scale (Strongly disagree to Strongly agree).

Perceived organisational support (predictor) was assessed using the four‐item version (Caesens et al., 2014; α = 0.82; originally developed in French; e.g., My organisation really cares about my well‐being; α = 0.74–0.77 across the present samples) of Eisenberger et al.’s (1986) instrument. Items were rated on a seven‐point scale (Strongly disagree to Strongly agree).

Work‐home segmentation (predictor) was measured with a four‐item subscale from Kreiner's (2006) measure (e.g., My workplace lets people forget about work when they are at home; α = 0.94 in Kreiner, 2006; α = 0.78–0.82 across the present samples). Items were rated on a seven‐point scale (Strongly disagree to Strongly agree).

Work engagement (outcome) was assessed using a three‐item scale (i.e., When working, I feel bursting with energy, I am enthusiastic about my job, and I am immersed in my work; α = 0.77–0.81 across the present samples) from Schaufeli et al. (2019; α = 0.77–0.85; French version by Huyghebaert‐Zouaghi et al., 2021), all rated on a seven‐point scale (Never to Always).

Work‐family conflict (outcome) was assessed with three items (e.g., My work schedule makes it difficult for me to fulfil my family obligations; α = 0.84–0.89 across the present samples) from Huyghebaert et al. (2018a; α = 0.92; originally developed in French). Items were rated on a seven‐point scale (Strongly disagree to Strongly agree).

Turnover intentions (outcome) were assessed with three items (e.g., I often think about quitting this organisation; α = 0.89–0.92 across the present samples) from Jaros et al. (1997; α = 0.82–0.85; French version by Gillet et al., 2018a). Items were rated on a seven‐point scale (Strongly disagree to Strongly agree).

Presenteeism (outcome) was assessed using a six‐item scale developed by Koopman et al. (2002; α = 0.80; French version by Huyghebaert et al., 2018b; for example, Because of my health problems, the stresses of my job were much harder to handle; α = 0.95–0.97 across the present samples). Items were rated on a seven‐point scale (Strongly disagree to Strongly agree) while referring to the past month.

Counterproductive work behaviours (CWB; outcome) were assessed using five items focussing on social interactions (e.g., Insulted someone about their job performance; α = 0.70–0.72 across the present samples) developed by Spector et al. (2010; α = 0.79; French version by Fouquereau et al., 2019), all rated on a seven‐point scale (Never to Always).

3. ANALYSES

3.1. Preliminary analyses

The psychometric properties of all multi‐item measures were verified as part of preliminary factor analyses. We relied on exploratory structural equation modelling to represent participants' responses to the work passion questionnaire (Marsh et al., 2013) and on confirmatory factor analyses for the predictors and outcomes. For both types of models, we then tested, separately in each of our five samples, the invariance of the measurement model across all subsamples. Details on these analyses (factor structure, measurement invariance across subsamples, composite reliability, and factor correlations) are reported in the online supplements (text and Tables S1–S5). The main analyses relied on factor scores saved from these preliminary analyses (Morin et al., 2016b). To ensure comparability across groups, factor scores were obtained from models specified as invariant across subsamples (Millsap, 2011), and estimated in standardised units (SD = 1; M = 0). Factor scores provide a partial control for unreliability (Skrondal & Laake, 2001) and preserve the structure of the measurement model (e.g., invariance; Morin et al., 2016a).

3.2. Model estimation

All models were estimated in Mplus 8.8 (Muthén & Muthén, 2022) using the maximum likelihood robust estimator and full information maximum likelihood (FIML; Endres, 2010) procedures to handle the limited amount of missing responses (0%–1.6% in Sample 1, 0%–2.8% in Sample 2, 0%–4.9% in Sample 3, 0%–3.5% in Sample 4%, and 0%–2.9% in Sample 5). Latent profile analyses (LPA) were estimated using 5000 random start values, 1000 iterations, and 200 final stage optimisations (Hipp & Bauer, 2006; Morin & Litalien, 2019).

3.3. Latent profile analyses

LPA summarize the multivariate distribution of scores on a set of profile indicators via the identification of a finite set of latent subpopulations, or profiles, displaying distinct configurations, while allowing for within‐profile variability on all indicators (McLachlan & Peel, 2000). These profiles are similar to prototypes and called latent to reflect their probabilistic nature (Morin et al., 2018). Each participant is assigned a probability of membership in each of the latent profiles, resulting in a solution controlled for classification errors. Solutions including one to eight profiles were estimated separately across subsamples while allowing the means and variances of the indicators (HP and OP) to be freely estimated (Morin & Litalien, 2019).

3.4. Model comparison and selection

The decision of how many profiles to retain relies on a consideration of whether the profiles are meaningful, aligned with theory, and statistically adequate (Marsh et al., 2009; Morin, 2016). Statistical indicators (McLachlan & Peel, 2000) can also be consulted. A lower value on the Akaïke Information Criterion (AIC), Consistent AIC (CAIC), Bayesian Information Criterion (BIC), and sample‐size Adjusted BIC (ABIC) indicates better fitting models. Statistically significant p‐values on the adjusted Lo et al. (2001) Likelihood Ratio Test (aLMR), and Bootstrap Likelihood Ratio Test (BLRT) also suggest better fit relative to a model with one fewer profile.

Statistical research has shown that the BIC, CAIC, ABIC, and BLRT, but not the AIC and aLMR, were efficient to indicate the true number of latent profiles (e.g., Diallo et al., 2016, 2017). The AIC and aMLR are thus only reported for purposes of transparency but are not used for model assessment. These tests all present a strong sample size dependency (Marsh et al., 2009), and thus often fail to converge on a specific number of profiles. When this happens, a graphical display (i.e., an elbow plot) can be used to locate the point at which the decrease in the value of the information criteria reaches a plateau (Morin et al., 2011). In practice (e.g., Morin et al., 2016a; Morin & Litalien, 2019), the statistical indicators are considered first to help pinpoint a range of acceptable solutions, which are then examined to eliminate those that are statistically improper, before being contrasted in terms of meaningfulness and theoretical conformity. The classification accuracy (from 0 to 1) is summarised by the entropy, which should not be used to select the number of profiles (Lubke & Muthén, 2007).

3.5. Tests of profile similarity

Assuming that the same number of profiles are extracted across groups (Morin & Wang, 2016), the group‐specific LPA solutions will be combined into a single multi‐group LPA model for tests of profile similarity across subsamples (Morin et al., 2016b). These sequential tests start by assessing whether the same number of profiles can be identified in each subsample. The subsample‐specific solutions are then combined in a multi‐group model of configural similarity. Equality constraints are then imposed on the within‐profile means (structural similarity), variances (dispersion similarity), and size (distributional similarity). These tests rely on the CAIC, BIC, and ABIC, so that each type of similarity can be considered supported as long as two indicators decrease following the integration of equality constraints (Morin et al., 2016b). ¹

3.6. Predictors and outcomes of profile membership

Starting from the most similar multi‐group model, we then assessed the extent to which the relations between the profiles, the predictors (predictive similarity), and the outcomes (explanatory similarity) remained the same across subsamples. To realize these tests, the predictors (i.e., POS and work‐home segmentation) and outcomes (i.e., work engagement, work‐family conflict, turnover intentions, presenteeism, and CWB) had to be directly included (i.e., direct inclusion, or one‐step, approach) into the final LPA solution (Morin et al., 2016b). The predictors were first considered in three predictive models in which their associations with profile membership were specified using a multinomial logistic regression function. First, a null effects model assumed no relations between the predictors and the profiles. Second, the effects of the predictors were freely estimated and allowed to vary across subsamples. Finally, a model of predictive similarity constrained these associations to be equal across subsamples. Outcome measures were allowed to vary as a function of profile membership. Explanatory similarity was assessed by constraining these associations to be equal across subsamples. The multivariate delta method was used to test the significance of between‐profile differences (Raykov & Marcoulides, 2004).

4. RESULTS

4.1. Latent profile analyses

The statistical indicators associated with each of the subsample‐specific LPA solutions are reported in Tables S7–S9, and graphically displayed in Figures S1 and S2, in the online supplements. These indicators failed to converge on a clearly dominant solution across subsamples, but tended to support solutions including four to seven profiles (with marked variations across subsamples and indicators). However, the elbow plots associated with these solutions systematically suggested that the flattening of the decrease in the value of the information criteria tended to happen earlier than that, generally around solutions including three to five profiles across all subsamples. These solutions were thus more carefully examined. This examination first revealed that all solutions were already quite similar across subsamples, thus providing early evidence of configural similarity. Moreover, this examination revealed that adding a fourth profile always resulted in a theoretically meaningful contribution across all subsamples. In contrast, adding a fifth (or sixth) profile generally resulted in the arbitrary separation of one already identified profile into smaller ones with a comparable shape, or in the estimation of empty profiles. The four‐profile solution was thus retained across all subsamples.

The results from the tests of profile similarity conducted across subsamples within each of our five samples are reported in Table 2 (Samples 1 and 2) and 3 (Samples 3–5). Starting with a model of configural similarity including four profiles in each group, equality constraints were progressively integrated to the solution. In Samples 3, 4, and 5, all of these additional equality constraints resulted in a systematic decrease in CAIC, BIC, and ABIC, and were thus supported by the data, leading us to retain a model of distributional similarity in these three samples. In contrast, in Samples 1 and 2, the model of structural similarity resulted in an increase in BIC and ABIC (and CAIC in Sample 1) relative to the model of configural similarity and was thus rejected. An examination of the parameter estimates obtained across all subsamples in the configural solution highlighted differences limited to a single indicator (HP) in a single profile that differed in shape between nurses relative to sales or administrative employees (Sample 1), as well as between full‐time and part‐time employees (Sample 2). Once equality constraints placed on this indicator were relaxed, the model of partial structural similarity was supported by the data, as well as the next model of dispersion similarity, for both samples. The final model of distributional similarity was also supported in Sample 1, but rejected in Sample 2, suggesting that the size of the profiles differed between full‐time and part‐time employees.

TABLE 2.

Results from the multi‐group latent profile analyses (samples 1 and 2).

Model	LL	#fp	Scaling	AIC	CAIC	BIC	ABIC	Entropy
Sample 1
Configural similarity	−6446.551	59	1.046	13,011.101	13,397.497	13,338.497	13,151.054	0.720
Structural similarity	−6527.768	43	1.069	13,141.536	13,423.147	13,380.147	13,243.536	0.762
Partial structural similarity	−6496.618	44	1.003	13,081.235	13,369.395	13,325.395	13,185.607	0.733
Dispersion similarity	−6539.675	28	1.149	13,135.349	13,318.723	13,290.723	13,201.767	0.706
Distributional similarity	−6548.730	22	1.029	13,141.459	13,285.539	13,263.539	13,193.645	0.708
Predictive similarity: Predictors
Null effects model	−11598.767	10	0.947	23,217.534	23,283.072	23,273.072	23,241.302	0.705
Free relations with predictors	−11502.027	28	1.028	23,060..054	23,243.561	23,215.561	23,126.605	0.719
Equal relations with predictors	−11522.914	16	0.983	23,077.828	23,182.689	23,166.689	23,115.857	0.715
Explanatory similarity
Free relations with outcomes	−18540.295	70	1.042	37,220.590	37,679.357	37,609.357	37,386.966	0.806
Equal relations with outcomes	−18610.086	30	1.114	37,280.173	37,476.787	37,446.787	37,351.477	0.819
Sample 2
Configural similarity	−3821.114	39	1.062	7720.228	7960.380	7921.380	7797.496	0.783
Structural similarity	−3850.200	31	1.042	7762.400	7953.290	7922.290	7823.818	0.746
Partial structural similarity	−3845.700	32	1.059	7755.399	7952.447	7920.447	7818.799	0.748
Dispersion similarity	−3792.576	24	0.925	7633.153	7780.938	7756.938	7680.703	0.769
Distributional similarity	−3810.265	21	0.908	7662.529	7791.842	7770.842	7704.135	0.766
Predictive similarity: Predictors
Null effects model	−7110.115	12	0.986	14,244.230	14,318.253	14,306.253	14,268.135	0.763
Free relations with predictors	−7035.396	24	1.140	14,118.791	14,266.837	14,242.837	14,166.601	0.772
Equal relations with predictors	−7040.686	18	1.085	14,117.371	14,228.406	14,210.406	14,153.229	0.771
Explanatory similarity
Free relations with outcomes	−11980.574	52	1.015	24,065.147	24,385.913	24,333.913	24,168.734	0.838
Equal relations with outcomes	−12747.588	32	0.879	25,559.176	25,757.590	25,725.590	25,623.940	0.812

Abbreviations: #fp, Number of free parameters; ABIC, Sample size adjusted BIC; AIC, Akaïke information criteria; BIC, Bayesian information criteria; CAIC, Constant AIC; LL, Model loglikelihood; Scaling, Scaling correction factor associated with robust maximum likelihood estimates.

The final solutions retained for all samples are illustrated in Figure 1, while parameter estimates can be consulted in Tables S10 and S11 of the online supplements. Consistent with the high entropy of these solutions (0.708–0.750), the results (Table S10) suggest that they all present a relatively high level of classification accuracy, ranging from 70.8% to 96.4% across profiles and subsamples. Turning first our attention to the profile that differed in shape across subsamples in Samples 1 and 2 (i.e., Profile 4), the global shape of this profile remained similar across subsamples. Indeed, whereas the levels of HP observed in Profile 4 were slightly above the sample mean among administrative, sales, and part‐time employees, they were slightly under the sample mean among nurses and full‐time employees. Taking a step back from these specific results, the similarity in the nature of the profiles identified across samples is quite striking. To facilitate referencing, we have summarised the names of all profiles identified at the bottom of Table 1. Although a four‐profile solution was retained in each sample, the whole set of profiles identified across the five samples can be summarised by a series of five main configurations, partially supporting Hypothesis 1.

FIGURE 1.

Final four‐profile solutions observed in all samples. Profile indicators are factor scores estimated in standardised units (M = 0; SD = 1); Sample 1: Profile 1: High OP Dominant; Profile 2: High HP Dominant; Profile 3: Low HP Dominant; Profile 4 (Admin/Sales): Average HP Dominant; and Profile 4 (Nurses): Moderately Low Passion; Sample 2: Profile 1: Moderately Low Passion; Profile 2: Average HP Dominant; Profile 3: High HP Dominant; and Profile 4: High OP Dominant; Sample 3: Profile 1: Average HP Dominant; Profile 2: High OP Dominant; Profile 3: High HP Dominant; and Profile 4: Low HP Dominant; Sample 4: Profile 1: Average HP Dominant; Profile 2: High OP Dominant; Profile 3: Low HP Dominant; and Profile 4: High HP Dominant; Sample 5: Profile 1: Low HP Dominant; Profile 2: Average HP Dominant; Profile 3: High HP Dominant; and Profile 4: High OP Dominant.

First, a profile dominated by high level of HP (High HP Dominant) was identified across all subsamples, and this profile was systematically the smallest (corresponding to 2.1%–7.7% of the employees). Second, a profile dominated by high levels of OP was also identified across all subsamples (High OP Dominant), and this profile was systematically the largest (37.5%–54.1%).

Two additional profiles, respectively characterised by average (Average HP Dominant) or low (Low HP Dominant) levels of passion dominated by higher levels of HP relative to OP were also identified in most samples. Thus, an Average HP Dominant profile was identified in 10 out of 11 subsamples, with the sole exception of the nurses from Sample 1. This profile corresponded to roughly a fifth (13.2%–24.5%) of the full‐time, part‐time, private sector, public sector, permanent, and temporary employees, and to roughly a third (31.8%–34.5%) of the administrative, sales, experienced, and novice employees. Then, a Low HP Dominant profile was identified in 9 out of 11 subsamples (it was not identified in Sample 2) and corresponded to roughly a fifth (13.8%–17.2%) of the administrative, sales, nurses, experienced, and novice employees, versus roughly a third (29.5%–33.6%) of the private sector, public sector, permanent, and temporary employees. As the variation in size related to these profiles seemed far more related to the nature of the samples (which are roughly comparable to one another), than to the subsamples selected to differ systematically from one another, they can be interpreted as reflecting random sampling variations rather than meaningful differences. However, these variations suggest that the prevalence of these two profiles tends to oscillate between a fifth to a third of the employees, with one of these two profiles being generally more prevalent than the other in any given sample.

The last profile was only identified among three subsamples (nurses, full‐time, and part‐time employees) among whom both types of passion were lower than the sample average (Moderately Low Passion) and corresponded to roughly a third of the employees. Interestingly, this profile only emerged when one of the two previous profiles did not (Average HP Dominant or Low HP Dominant), suggesting that close to, or slightly below, average levels of passion can either be dominated by HP in some situations, or not dominated by any form of passion in other situations. Finally, when we consider differences in the prevalence of the profiles between the full‐time and part‐time employees forming Sample 2, these differences mainly show that the High OP Dominant profile was more prevalent among full‐time relative to part‐time employees, among whom the Moderately Low Passion and Average HP Dominant profiles were slightly more prevalent.

4.2. Predictors of profile membership

As shown in Tables 2 and 3, our predictive results generalised across subsamples within our five samples. Indeed, for all samples, the CAIC, BIC, and ABIC were lower for the model of predictive similarity than for alternative models. The results from these solutions of predictive similarity are reported in Table 4 and are remarkably consistent across samples.

TABLE 3.

Results from the multi‐group latent profile analyses (samples 3, 4, and 5).

Model	LL	#fp	Scaling	AIC	CAIC	BIC	ABIC	Entropy
Sample 3
Configural similarity	−4860.472	39	1.072	9798.943	10,047.676	10,008.676	9884.780	0.760
Structural similarity	−4869.817	31	1.158	9801.634	9999.345	9968.345	9869.864	0.764
Dispersion similarity	−4871.608	23	1.104	9789.216	9935.905	9912.905	9839.838	0.748
Distributional similarity	−4872.179	20	1.110	9784.358	9911.913	9891.913	9828.377	0.748
Predictive similarity: Predictors
Null effects model	−8904.383	9	0.963	17,826.766	17,884.216	17,875.216	17,846.625	0.744
Free relations with predictors	−8832.190	21	1.004	17,706.379	17,840.430	17,819.430	17,752.717	0.756
Equal relations with predictors	−8837.596	15	0.990	17,705.192	17,800.942	17,785.942	17,738.290	0.755
Explanatory similarity
Free relations with outcomes	−15024.864	49	0.984	30,147.728	30,460.513	30,411.513	30,255.850	0.836
Equal relations with outcomes	−15057.266	29	1.020	30,172.532	30,357.650	30,328.650	30,236.522	0.835
Sample 4
Configural similarity	−3646.742	39	1.027	7371.485	7608.900	7569.900	7446.021	0.715
Structural similarity	−3663.883	31	1.019	7389.767	7578.482	7547.482	7449.014	0.709
Dispersion similarity	−3663.613	23	1.127	7373.226	7513.240	7490.240	7417.183	0.749
Distributional similarity	−3663.859	20	1.140	7367.718	7489.469	7469.469	7405.941	0.750
Predictive similarity: Predictors
Null effects model	−6666.106	9	0.994	13,350.212	13,405.105	13,396.105	13,367.517	0.743
Free relations with predictors	−6595.625	21	0.990	13,233.250	13,361.333	13,340.333	13,273.629	0.755
Equal relations with predictors	−6604.872	15	0.983	13,239.744	13,331.233	13,316.233	13,268.586	0.754
Explanatory similarity
Free relations with outcomes	−11387.160	49	0.996	22,872.321	23,171.181	23,122.181	22,966.538	0.810
Equal relations with outcomes	−11423.934	29	1.061	22,905.868	23,082.745	23,053.745	22,961.629	0.806
Sample 5
Configural similarity	−3727.973	39	0.988	7533.946	7772.136	7733.136	7609.255	0.734
Structural similarity	−3747.896	31	0.993	7557.791	7747.121	7716.121	7617.653	0.740
Dispersion similarity	−3752.355	23	1.209	7550.710	7691.181	7668.181	7595.123	0.713
Distributional similarity	−3756.453	20	1.002	7552.906	7675.054	7655.054	7591.526	0.735
Predictive similarity: Predictors
Null effects model	−6884.206	9	1.007	13,786.412	13,841.467	13,832.467	13,803.879	0.729
Free relations with predictors	−6803.972	21	1.067	13,649.943	13,778.404	13,757.404	13,690.699	0.745
Equal relations with predictors	−6810.313	15	1.052	13,650.627	13,742.385	13,727.385	13,679.738	0.743
Explanatory similarity
Free relations with outcomes	−11565.456	49	0.987	23,228.913	23,528.656	23,479.656	23,324.011	0.814
Equal relations with outcomes	−11617.431	29	1.023	23,292.862	23,470.661	23,441.661	23,349.144	0.811

Abbreviations: #fp, number of free parameters; ABIC, Sample size adjusted BIC; AIC, Akaïke information criteria; BIC, Bayesian information criteria; CAIC, Constant AIC; LL, Model loglikelihood; scaling, scaling correction factor associated with robust maximum likelihood estimates.

TABLE 4.

Results from the predictive analyses (all samples).

Predictors	Coef. (SE)	OR	Coef. (SE)	OR	Coef. (SE)	OR	Coef. (SE)	OR	Coef. (SE)	OR	Coef. (SE)	OR
	High HP Dom. Versus high OP Dom.		High HP Dom. Versus Average HP Dom. or moder. Low passion a		High HP Dom. Versus low HP dominant		High OP Dom. Versus average HP Dom. or Moder. Low passion a		High OP Dom. Versus low HP Dominant		Low HP dominant versus average HP Dom. or moder. Low passion a
Sample 1
POS	0.060 (0.127)	1.062	0.763 (0.185)**	2.146	0.643 (0.162)**	1.902	0.101 (0.084)	1.106	0.582 (0.116)**	1.790	0.205 (0.135)	1.227
Segmentation	1.210 (0.184)**	3.353	0.161 (0.127)	1.175	0.559 (0.223)*	1.748	−0.446 (0.084)**	0.640	−0.651 (0.121)**	0.521	−0.482 (0.125)**	0.618

Sample 2	High HP Dom. versus high OP Dom.		High HP Dom. versus average HP Dom.		High HP Dom. versus moder. Low passion		High OP Dom. versus average HP Dom.		High OP Dom. versus moder. Low passion		Moder. low passion versus average HP Dom.
POS	0.412 (0.390)	1.510	0.813 (0.396)*	2.255	0.893 (0.393)*	2.442	0.401 (0.110)**	1.493	0.481 (0.105)**	1.618	−0.080 (0.113)	0.923
Segmentation	1.351 (0.470)**	3.861	0.143 (0.481)	1.154	0.777 (0.470)	2.175	−1.207 (0.159)**	0.299	−0.574 (0.111)**	0.563	−0.634 (0.168)**	0.530

Sample 3	High HP Dom. versus high OP Dom.		High HP Dom. versus average HP Dom.		High HP Dom. versus low HP Dominant		High OP Dom. versus average HP Dom.		High OP Dom. versus low HP Dominant		Low HP dominant versus average HP Dom.
POS	0.273 (0.263)	1.314	0.631 (0.266)*	1.879	0.732 (0.265)**	2.080	0.358 (0.110)**	1.430	0.459 (0.102)**	1.583	−0.101 (0.116)	0.904
Segmentation	0.698 (0.172)**	2.010	−0.347 (0.204)	0.707	−0.090 (0.181)	0.914	−1.045 (0.141)**	0.352	−0.788 (0.104)**	0.455	−0.256 (0.154)	0.774

Sample 4	High HP Dom. Versus high OP Dom.		High HP Dom. Versus average HP Dom.		High HP Dom. Versus low HP Dominant		High OP Dom. Versus average HP Dom.		High OP Dom. Versus low HP Dominant		Low HP Dominant versus average HP Dom.
POS	0.194 (0.169)	1.214	0.756 (0.187)**	2.130	0.953 (0.172)**	2.593	0.561 (0.136)**	1.752	0.758 (0.116)**	2.134	−0.197 (0.142)	1.217
Segmentation	0.744 (0.190)**	2.104	−0.360 (0.231)	0.698	0.104 (0.200)	1.110	−1.104 (0.176)**	0.332	−0.640 (0.122)**	0.527	−0.464 (0.190)*	1.590

Sample 5	High HP Dom. Versus high OP Dom.		High HP Dom. Versus average HP Dom.		High HP Dom. Versus low HP Dominant		High OP Dom. Versus average HP Dom.		High OP Dom. Versus low HP Dominant		Low HP Dominant versus average HP Dom.
POS	0.201 (0.334)	1.223	2.592 (0.452)**	13.356	2.866 (0.468)**	17.567	0.538 (0.109)**	1.713	0.812 (0.138)**	2.252	0.232 (0.162)	1.261
Segmentation	2.054 (0.443)**	7.799	−0.463 (0.351)	0.629	−0.695 (0.378)	0.499	−0.664 (0.123)**	0.515	−0.896 (0.156)**	0.408	−0.274 (0.138)*	0.760

Note: The coefficients and OR reflect the effects of the predictors on the likelihood of membership into the first listed profile relative to the second listed profile; indicators of perceived organisational support and work‐home segmentation are factor scores with a standard deviation of 1 and a mean of 0.

Abbreviations: OR, odds ratio; POS, perceived organisational support; SE, standard error of the coefficient.

^a In Sample 1, the Average HP Dominant profile was identified among administrative and sales employees, whereas the Moderately Low Passion profile was identified among nurses.

*p < 0.05; **p < 0.01.

Perceptions of organisational support were associated with a higher likelihood of membership into the High HP Dominant and High OP Dominant profiles relative to the Average HP Dominant, Low HP Dominant, and Moderately Low Passion profiles across most samples. The only exception was Sample 1, where these perceptions were not differentially associated with membership into the High OP Dominant profile relative to the combined Average HP Dominant (administrative and sales employees)/Moderately Low Passion (nurses) profiles. These results partially support Hypothesis 2.

Perception of work‐home segmentation norms were systematically associated with a higher likelihood of membership into the High HP Dominant, Average HP Dominant, Low HP Dominant, and Moderately Low Passion profiles relative to the High OP Dominant profile across all samples. These perceptions were also associated with a higher likelihood of membership into the Average HP Dominant (in combination with the Moderately Low Passion profile in Sample 1) profile relative to the Low HP Dominant profile in Samples 1, 4, and 5 or to the Moderately Low Passion profile in Sample 2. The only non‐systematic effect occurred in Sample 1, in which work‐home segmentation perceptions were also associated with a higher likelihood of membership into the High HP Dominant profile than into the Low HP Dominant profile. These results partially support Hypothesis 3.

4.3. Outcomes of profile membership

As shown in Tables 2 and 3, the associations between the profiles and the outcomes generalised across most subsamples considered within our five samples. Indeed, in Samples 1, 3, 4, and 5, at least two indicators out of the CAIC, BIC, and ABIC were lower (two in Sample 5, but all three in Samples 1, 3, and 4) for the model of explanatory similarity than for the alternative model. In contrast, these three indicators rather supported the presence of associations with outcomes that differed between subsamples of full‐time and part‐time employees in Sample 2. The results from these solutions are reported in Table 5, partially support Hypotheses 4–6, and are again quite consistent across samples.

TABLE 5.

Associations between profile membership and the outcomes (all samples).

	[1] High HP Dominant	[2] High OP Dominant	[3] Low HP Dominant	[4] Average HP Dominant	[5] Moderately Low Passion	Statistically significant differences
Work engagement	M [CI]	M [CI]	M [CI]	M [CI]	M [CI]
Sample 1	0.104 [−0.013; 0.221]a	0.165 [0.097; 0.232]a	−1.161 [−1.468; −0.854]	0.072 [0.003; 0.141]a, §		[3] < [1] = [2] = [4/5]
Sample 2 (full‐time)	0.880 [0.669; 1.092]	0.378 [0.286; 0.470]		−0.035 [−0.229; 0.160]a	0.075 [−0.059; 0.210]a	[4] = [5] < [2] < [1]
Sample 2 (part‐time)	−0.102 [−0.560; 0.355]b,c	0.231 [0.065; 0.396]b		−0.608 [−0.782; −0.433]a	−0.474 [−0.664; −0.283]a,c	[4] = [5] < [1]; [4] < [1] = [2]; [1] = [5]
Sample 3	0.305 [0.174; 0.436]b	0.280 [0.213; 0.348]b	−0.235 [−0.348; −0.121]a	−0.390 [−0.536; −0.245]a		[3] = [4] < [1] = [2]
Sample 4	0.349 [0.251; 0.447]a	0.293 [0.202; 0.384]a	−0.416 [−0.554; −0.279]	−0.077 [−0.262; 0.109]		[3] < [4] < [1] = [2]
Sample 5	0.908 [0.810; 1.005]	0.219 [0.136; 0.302]	−0.480 [−0.670; −0.290]	−0.195 [−0.299; −0.091]		[3] < [4] < [2] < [1]

Work‐family conflict	M [CI]	M [CI]	M [CI]	M [CI]	M [CI]
Sample 1	−0.999 [−1.039; −0.959]	0.730 [0.664; 0.795]	−0.584 [−0.696; −0.472]	−0.463 [−0.511; −0.415] §		[1] < [3] < [4/5] < [2]
Sample 2 (full‐time)	−1.021 [−1.129; −0.912]	0.748 [0.641; 0.856]		−0.767 [−0.874; −0.659]	−0.387 [−0.467; −0.307]	[1] < [4] < [5] < [2]
Sample 2 (part‐time)	−0.971 [−1.181; −0.761]a	0.598 [0.430; 0.767]		−0.860 [−0.924; −0.796]a	−0.253 [−0.382; −0.124]	[1] = [4] < [5] < [2]
Sample 3	−0.508 [−0.611; −0.404]a	0.720 [0.647; 0.794]	−0.498 [−0.550; −0.447]a	−0.861 [−0.917; −0.806]		[4] < [1] = [3] < [2]
Sample 4	−0.577 [−0.668; −0.487]	0.668 [0.566; 0.770]	−0.329 [−0.419; −0.238]	−0.733 [−0.811; −0.655]		[4] < [1] < [3] < [2]
Sample 5	−0.764 [−0.848; −0.679]a	0.678 [0.588; 0.768]	−0.749 [−0.824; −0.674]a	−0.384 [−0.448; −0.320]		[1] = [3] < [4] < [2]

Turnover intentions	M [CI]	M [CI]	M [CI]	M [CI]	M [CI]
Sample 1	−0.519 [−0.640; −0.397]	0.189 [0.112; 0.265]	0.760 [0.417; 1.103]	−0.283 [−0.363; −0.203] §		[1] < [4/5] < [2] < [3]
Sample 2 (full‐time)	−0.638 [−1.093; −0.183]a	−0.016 [−0.129; 0.097]		−0.410 [−0.623; −0.197]a	−0.247 [−0.393; −0.101]a	[1] = [4] = [5] < [2]
Sample 2 (part‐time)	−0.432 [−0.839; −0.025]	0.195 [−0.004; 0.394]a		0.058 [−0.125; 0.242]a	0.281 [0.061; 0.501]a	[1] < [2] = [4] = [5]
Sample 3	−0.464 [−0.604; −0.324]	0.150 [0.068; 0.232]	−0.025 [−0.144; 0.094]a	−0.241 [−0.395; −0.088]a		[1] < [3] = [4] < [2]
Sample 4	−0.517 [−0.660; −0.374]	0.036 [−0.067; 0.139]a	0.180 [0.033; 0.327]a	−0.275 [−0.475; −0.075]		[1] < [4] < [2] = [3]
Sample 5	−0.637 [−0.729; −0.546]	0.150 [0.051; 0.249]a	0.111 [−0.081; 0.302]a	−0.019 [−0.133; 0.095]a		[1] < [2] = [3] = [4]

Presenteeism	M [CI]	M [CI]	M [CI]	M [CI]	M [CI]
Sample 1	−0.467 [−0.579; −0.356]	0.365 [0.285; 0.445]	−0.216 [−0.360; −0.072]a	−0.258 [−0.324; −0.192]a, §		[1] < [3] = [4/5] < [2]
Sample 2 (full‐time)	−0.400 [−0.807; 0.007]a	0.227 [0.103; 0.352]		−0.460 [−0.590; −0.330]a	−0.371 [−0.458; −0.284]a	[1] = [4] = [5] < [2]
Sample 2 (part‐time)	−0.480 [−0.790; −0.170]a	0.526 [0.329; 0.723]		−0.316 [−0.442; −0.189]a	−0.005 [−0.178; 0.168]	[1] = [4] < [5] < [2]
Sample 3	−0.285 [−0.415; −0.155]a	0.222 [0.140; 0.303]	−0.099 [−0.200; 0.002]	−0.341 [−0.445; −0.236]a		[1] = [4] < [3] < [2]
Sample 4	−0.404 [−0.515; −0.292]a	0.251 [0.139; 0.363]	−0.058 [−0.183; 0.067]	−0.307 [−0.458; −0.155]a		[1] = [4] < [3] < [2]
Sample 5	−0.227 [−0.465; 0.011]a	0.291 [0.189; 0.394]	−0.282 [−0.395; −0.169]a	−0.200 [−0.283; −0.116]a		[1] = [3] = [4] < [2]

CWB	M [CI]	M [CI]	M [CI]	M [CI]	M [CI]
Sample 1	−0.315 [−0.417; −0.213]	0.181 [0.112; 0.250]a	0.223 [0.017; 0.430]a	−0.190 [−0.249; −0.132] §		[1] < [4/5] < [2] = [3]
Sample 2 (full‐time)	−0.626 [−0.706; −0.545]	0.190 [0.079; 0.300]		−0.208 [−0.374; −0.042]a	−0.251 [−0.318; −0.183]a	[1] < [4] = [5] < [2]
Sample 2 (part‐time)	−0.486 [−0.634; −0.338]	0.347 [0.125; 0.569]		−0.218 [−0.343; −0.093]a	−0.048 [−0.230; 0.134]a	[1] < [4] = [5] < [2]
Sample 3	−0.247 [−0.404; −0.089]a,b	0.205 [0.128; 0.282]	−0.116 [−0.197; −0.036]b	−0.268 [−0.350; −0.186]a		[1] = [4] < [2]; 1 = [3] < [2]; [4] < [3]
Sample 4	−0.270 [−0.361; −0.178]a	0.134 [0.034; 0.235]	0.040 [−0.063; 0.143]b	−0.294 [−0.384; −0.203]a		1 = [4] < [2] = [3]
Sample 5	−0.406 [−0.552; −0.260]	0.216 [0.130; 0.302]	−0.125 [−0.261; 0.011]a	−0.150 [−0.231; −0.069]a		[1] < [3] = [4] < [2]

Note: Results (M, CI) associated with the same subscript within a single line do not differ from one another in a statistically significant manner (p ≤ 0.05); CWB: Counterproductive work behaviours; indicators of work engagement, work‐family conflict, turnover intentions, and presenteeism are factor scores with a mean of 0 and a standard deviation of 1.

Abbreviations: CI, 95% confidence interval, M, mean.

^§ In Sample 1, the Average HP Dominant profile was identified among administrative and sales employees, whereas the Moderately Low Passion profile was identified among nurses.

Across all samples, levels of work engagement were the highest among the High HP Dominant and High OP Dominant profiles, which only differed from one another among full‐time employees (Sample 2) and in Sample 5 where these levels where slightly higher in the High HP Dominant profile than in the High OP Dominant one. These levels were also at their lowest in the Low HP Dominant and Average HP Dominant profiles, which differed from one another in Samples 1 and 4, showing that these levels were lower in the Low HP Dominant profile than in the Average HP Dominant one. No differences were found between the Average HP Dominant and Moderately Low Passion profiles in Sample 2, the only sample where these two profiles were identified.

Levels of work‐family conflict were the highest in the High OP Dominant profile across all samples, and the lowest in the High HP Dominant profile (Samples 1, 2 full‐time, and 5), Average HP Dominant profile (Samples 3 and 4) or both (Sample 2 part‐time) profiles. The levels of work‐family conflict observed in the Low HP Dominant and Moderately Low Passion profiles fell in between these two extremes (with the exception of Sample 5, where these levels were as low in the Low HP Dominant and High HP Dominant profiles). Turnover intentions were systematically the lowest in the High HP Dominant profile across all subsamples, and the highest among the High OP Dominant and Low HP Dominant profiles, although the turnover intentions observed in these last two profiles sometimes differed from one another (higher in the Low HP Dominant profile in Sample 1 but higher in the High OP Dominant one in Sample 3). The turnover intentions associated with the Average HP Dominant and Moderately Low Passion profiles fell in between these two extremes.

Presenteeism was the highest in the High OP Dominant profile and the lowest in the High HP Dominant profile across all subsamples. Presenteeism was also lower in the Average HP Dominant (part‐time employees in Sample 2) or Low HP Dominant (Samples 3 and 4) profiles than in the Moderately Low Passion profile. In most samples (except Sample 1), levels of presenteeism were comparably lower in the High HP Dominant profile and in the Average HP Dominant one. Counterproductive work behaviours were the highest in the High OP Dominant profile (and in the Low HP Dominant profile in samples 1 and 3) and the lowest in the High HP Dominant profile across all subsamples. Levels of CWB fell in between these two extremes in the Average HP Dominant and Moderately Low Passion profiles, which did not differ from one another.

Lastly, differences between full‐time and part‐time employees (Sample 2) were limited to a subset of comparisons, but this did not change the overall picture of these results. Work engagement did not differ between the High HP Dominant and High OP Dominant profiles among part‐time employees, whereas they were higher in the former profile (High HP Dominant) for full‐time employees. Similarly, levels of work‐family conflict did not differ between the High HP Dominant and Average HP Dominant profiles for part‐time employees, whereas they were higher in the former profile (High HP Dominant) for full‐time employees. In both subsamples, turnover intentions showed few variations across profiles. However, they were at their highest in the High OP Dominant profile (relative to all other profiles) among full‐time employees, and at their lowest in the High HP Dominant profile (relative to all other profiles) among part‐time employees. Lastly, levels of presenteeism were similar across the High HP Dominant, Average HP Dominant, and Moderately Low Passion profiles among full‐time employees, but higher in the last of these profiles (Moderately Low) for part‐time employees. No differences were found for CWB.

5. DISCUSSION

This study sought to increase our theoretical understanding of work passion via the identification of the HP and OP configurations observed among distinct profiles of employees. To help guide practice into selecting which profiles are most likely to benefit from work‐related interventions, we also documented the associations between these profiles and a set of theoretically‐relevant predictors (i.e., work‐home segmentation and organisational support) and outcomes (i.e., work engagement, work‐family conflict, turnover intentions, presenteeism, and CWB). Lastly, capitalising on a multi‐sample research design, we examined the generalisability of these profiles and of their associations with predictors and outcomes across five samples, including 11 subsamples, of employees. Results demonstrating consistency across samples are summarised in Table 6. In addition to their practical implications, these results have theoretical implications for the DMP (Vallerand, 2015; Vallerand & Houlfort, 2019) and QMP (Schellenberg et al., 2019).

TABLE 6.

Summary of the characteristics of the main profiles identified in this study.

	High HP dominant	Average HP dominant	Low HP dominant	High OP dominant	Moderately low passion
HP	High	Average to moderately high	Average to moderately low	Average to moderately low	Moderately low
OP	Moderately low to low	Moderately low to low	Low	High	Moderately low
Size	2.1%–7.7%	13.2%–34.5%	13.8%–33.6%	37.5%–54.1%	30.6%–35.1%
Subsamples	All	10 out 11 subsamples (not for the nurses from sample 1)	9 out 11 subsamples (not in sample 2)	All	3 out 11 subsamples (nurses, full‐time, and part‐time employees)
Systematic predictions:	Higher perceived organisational support		Lower work‐home segmentation norms	Higher perceived organisational support Lower work‐home segmentation norms
Systematic outcomes	Highest work engagement Lowest work‐family conflict Lowest turnover intentions Lowest presenteeism Lowest counterproductive behaviours	Lowest work engagement Lowest work‐family conflict Lowest presenteeism	Lowest work engagement Highest turnover intentions Low presenteeism Highest counterproductive behaviours	Highest work engagement Work‐family conflict Highest turnover intentions Highest presenteeism Highest counterproductive behaviours

Abbreviations: HP, harmonious passion; OP, obsessive passion.

5.1. Work passion profiles

Although a four‐profile solution was retained in each sample, our results revealed that a total of five profiles best summarised the work passion configurations observed across the different samples and subsamples considered in this study: (1) High HP Dominant, (2) High OP Dominant, (3) Average HP Dominant, (4) Low HP Dominant, and (5) Moderately Low Passion. The High HP Dominant profile was systematically the smallest (2.1%–7.7%), whereas the High OP Dominant profile was systematically the largest (37.5%–54.1%). These results suggest that, unfortunately, much fewer employees approach work in a HP manner, whereas many of them display OP. Indeed, a third to half of the employees demonstrated an uncontrollable urge to engage in their work, experienced conflicts between their work passion and the other activities in their life, presented an increasing risk of experiencing failures or setbacks, and felt unable to fulfil their needs for autonomy, competence, and relatedness without working (Vallerand & Houlfort, 2019). Alternatively, the current work intensification phenomenon, implying increasing responsibilities and shorter deadlines, might also explain these higher levels of OP relative to those of HP (Huyghebaert‐Zouaghi et al., 2022).

The High HP Dominant and High OP Dominant profiles were similar to those identified by Gillet, Morin, et al. (2023), Li et al. (2020), and Morin et al. (2023). The only differences with Gillet, Morin, et al. (2023) and Morin et al. (2023) concerned the Moderately Low Passion profile, as both previous studies identified a Low Passion profile (low to very low HP‐OP), and our Average HP Dominant profile as Gillet, Morin, et al. (2023) identified a Mixed Passion‐Average profile (average HP‐OP), while Morin et al. (2023) and Li et al. (2020) found a Mixed Passion‐OP Dominant profile (high to very high HP‐OP).

The High HP Dominant and High OP Dominant profiles were identified across all subsamples, the Average HP Dominant profile was identified in 10 out of 11 subsamples, and the Low HP Dominant profile was identified in nine out of the 11 subsamples. These results led to us anticipate that the High HP Dominant and High OP Dominant profiles should emerge in most samples, together with at least two profiles displaying close to average levels of passion dominated, or not, by HP. In contrast, the Moderately Low Passion profile was only identified among three subsamples, and only as a replacement for the Low or Average HP Dominant profiles. This observation suggests that profiles with close to average passion are often dominated by HP but can also sometimes reflect a mixed passion (Gillet, Morin, et al., 2023). Although these results only partially supported our hypotheses (i.e., a Mixed Passion profile characterised by high HP‐OP was not identified), all profiles were expected based either on prior person‐centred studies (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b) and/or correspondence with the QMP (Schellenberg et al., 2019). In addition to providing evidence of replicability to the current samples of Western (i.e., French) employees, our results supported the generalisability of four of these five profiles across most subsamples considered in this study.

As noted in the introduction, person‐centred research evidence is cumulative and requires an accumulation of studies conducted on diversified samples to distinguish core profiles that emerge across all situations, additional profiles that emerge in specific situations, and idiosyncratic profiles likely to represent random sampling variations (Morin et al., 2016b). The generalisability of our findings across 11 subsamples of Western (i.e., French) employees and their consistency with those observed in previous research conducted among employees from different cultures (e.g., Li et al., 2020; Morin et al., 2023b) suggest that the identified profiles may reflect core psychological mechanisms involved in the experience of work passion, rather than sample‐specific ephemeral phenomena. More precisely, we can expect the following profiles to be observed in future person‐centred studies: (1) High HP Dominant profile characterised by high HP and moderately low to low OP (rather than high HP coupled with low OP; Schellenberg et al., 2019); (2) High OP Dominant profile characterised by average to moderately low HP and high OP (rather than low HP coupled with high OP; Schellenberg et al., 2019); (3) Average HP Dominant profile characterised by average to moderately high HP and moderately low to low OP (rather than high HP coupled with high OP; Li et al., 2020; Schellenberg et al., 2019); and (4) Low HP Dominant profile characterised by average to moderately low HP and low OP (rather than low HP coupled with low OP; Gillet, Morin, et al., 2023; Morin et al., 2023b; Schellenberg et al., 2019). Based on the present findings, this latter profile may not be as generalisable as the three others, and when not identified would likely be replaced by a profile presenting moderately low levels of HP and OP (Moderately Low Passion profile). However, it is important to investigate whether and how similar profiles will be identified in other occupations (e.g., teaching, industry), cultures (e.g., South America, Eastern Europe, Asia), or research designs. Likewise, the conditions under which profiles presenting close to average levels of passion will be dominated by low HP, moderate HP, or neither will need to be more thoroughly examined.

By relying on person‐centred analyses to account for the joint role of HP and OP, we did not have to rely on a rigid classification of participants into one of the QMP theoretical configurations (Schellenberg et al., 2019). Indeed, rather than assuming that HP and OP would necessarily be high or low (e.g., such as in the theoretical Mixed Passion profile), we identified the configurations truly present in the samples under consideration, allowing these configurations to be characterised by moderate levels of HP and OP as in Gillet, Morin, et al. (2023). Our findings thereby extend both the DMP (Vallerand, 2015; Vallerand et al., 2003) and the QMP (Schellenberg et al., 2019), especially in relation to the synergy between HP and OP which does not seem to occur at high levels. Indeed, the Average HP Dominant profile displayed average to moderately high HP and moderately low to low OP, rather than high HP and High OP (Mixed Passion profile) as proposed in the QMP (Schellenberg et al., 2019). Such findings allow us to refine the theoretical expectations of the QMP (Schellenberg et al., 2019) in showing that the combination of very high levels of HP and OP seemed quite rare. This result also informs the DMP (Vallerand, 2015; Vallerand et al., 2003) about the natural synergy between HP and OP, by revealing that, at high levels, HP and OP might be mutually exclusive or incompatible. From a theoretical point of view, our results suggest that employees could not be characterised by both high levels of HP and OP because such a work passion profile does not allow them to maintain a sufficient level of resources (Hobfoll, 2011). In other words, since HP is associated with an increase in resources while OP is associated with a decrease in resources (Vallerand et al., 2003), these two forms of passion cannot occur at high levels at the same time in the same employee. Further research is needed to determine whether the theoretical propositions of the DMP and QMP can be applied to work passion profiles, and in particular to the Mixed Passion profile. We should perhaps focus more on the fluctuations in HP and OP levels depending on the time of year or the tasks to be carried out, rather than considering that these two types of passion can be present in the same individual at the same time. To this end, relying on new methods such as dynamic structural equation modelling (Huyghebaert‐Zouaghi et al., 2023) could be an interesting avenue for future research.

Given the desirable outcomes associated with the High HP Dominant profile (contrary to the High OP Dominant one), it would seem important for organisations, supervisors, and the society more generally to consider implementing actions to help nurture employees' passion for their work in a way that is in harmony with the other spheres of their life, while maximally limiting their tendencies to become obsessive about their work. Consulting the few purely HP employees may make it possible to identify the most useful (or harmful) practices in this regard. Such interventions may subsequently be expanded to help employees displaying other work passion profiles (e.g., High OP Dominant or Low HP Dominant) increase their likelihood of developing a High HP Dominant profile at work. Interventions may also be offered to employees with the most adaptive work passion profile (i.e., High HP Dominant) to help them decrease their likelihood of eventually experiencing a rise in OP. Obviously, even if the content of these interventions could present common points, differences would also be observable insofar as the actionable levers aiming at reinforcing HP are not the same as those aiming at reducing OP (Vallerand, 2015). Beyond these hypothetical practical considerations, studies will be needed to examine the reasons underlying this problematically high prevalence of the High OP Dominant profile, and this concerningly low prevalence of the High HP Dominant profile. Fortunately, the theoretical predictors considered in this study provide some information in this regard, while also suggesting action levers that could be mobilised in interventions seeking to nurture High HP Dominant profiles and to limit the emergence of High OP Dominant profiles.

5.2. Predictors of work passion profiles

Although our predictive hypotheses were only partially supported in relation to the associations between POS and work‐home segmentation and employees' likelihood of profile membership, our results help us extend the nomological network of work passion and provide practical guidance regarding possible levers of intervention. Thus, perceptions of organisational support played the dual role of increasing HP and OP, being associated with a higher likelihood of membership into the High HP Dominant and High OP Dominant profiles. These results are consistent with previous evidence showcasing the benefits of POS (Kurtessis et al., 2017; Riggle et al., 2009). Indeed, employees who feel supported are more likely to internalize their work into their sense of identity (Gillet, Becker, et al., 2017), leading to HP (Vallerand & Houlfort, 2019).

In contrast, these employees may feel internally pressured to repay their organisation for its support by increasing their work investment beyond what might be reasonable, making it harder for them to withdraw from work when they should be recovering from it (Gillet et al., 2017b) and paving the way for OP (Vallerand, 2015; Vallerand & Houlfort, 2019). Our results confirm the idea that high levels of POS may sometimes be harmful in a “too much of a good thing” manner (Harris & Kacmar, 2006). What remains to be identified, however, are the conditions under which POS will support the emergence of profiles dominated by high levels of HP (High HP Dominant) or of OP (High OP Dominant). Arguably, either one of those pathways could depend on the extent to which employees' work motives are primarily driven by autonomous reasons and pleasure, internal or external contingencies, or both (Gillet, Becker, et al., 2017). These possibilities need to be more thoroughly examined in future research.

Work‐home segmentation was associated with a lower likelihood of membership into the High OP Dominant profile relative to the other profiles, suggesting that its main benefits involve limiting the emergence of a strong sense of OP. By enforcing a healthy separation between work and personal life (Kreiner, 2006), work‐home segmentation is incompatible with the emergence of OP (Vallerand & Houlfort, 2019). However, by making easier for employees to achieve their work outcomes while fulfiling their personal roles (Althammer et al., 2021), it represents a fertile ground for the emergence of profiles dominated by HP, as illustrated by the fact that work‐home segmentation was associated with a higher likelihood of membership into the Average HP Dominant profile relative to the Low HP Dominant or Moderately Low Passion profiles. These observations are consistent with links between work‐home segmentation and experiencing positive affect when involved in family roles, allowing employees to benefit from these roles to rebuild their psychological resources (Huyghebaert‐Zouaghi et al., 2022), upon which they can then rely to support their HP for work (Vallerand & Houlfort, 2019). However, it is important to keep in mind that our results primarily suggest that work‐home segmentation decreases the risk of experiencing a High OP Dominant profile, but without necessarily increasing the chance of becoming passionate about their work. As such, work‐home segmentation is equally likely to predict profiles characterised by low to average levels of passion (e.g., Moderately Low Passion profile), as profiles dominated by HP (e.g., High HP Dominant profile). Future research will be needed to better understand the mechanisms underlying these associations.

5.3. Outcomes of profile membership

Some profiles appeared more desirable than others in terms of outcomes, although these associations differed across outcomes. Consistent with the core assumptions of the DMP (Vallerand & Houlfort, 2019), the High HP Dominant profile consistently displayed the lowest levels of work‐family conflict (together with the Average HP Dominant profile), presenteeism, turnover intentions, and CWB, whereas the highest levels on all of these outcomes were systematically associated with the High OP Dominant profile (together with the Low HP Dominant profile for turnover intentions and CWB). The other profiles fell in between these two extremes with few additional noteworthy differences. These findings clearly support the positive effects of HP and the detrimental effects of OP identified in previous research (e.g., Gillet, Morin, et al., 2023; Houlfort et al., 2018; Schellenberg et al., 2019).

This pattern of results was not replicated for work engagement, which reflects a strong form of work investment without any consideration of balance (Gillet et al., 2018c; Schaufeli et al., 2019). Indeed, levels of work engagement were equally highest among the High HP Dominant and High OP Dominant profiles, and lowest in the Low HP Dominant and Average HP Dominant profiles. As expected, passionate employees (HP or OP) invest a lot of time, effort, and energy in a work that they enjoy, which is consistent with work engagement (Birkeland & Buch, 2015; Tóth‐Király et al., 2021; Vallerand & Houlfort, 2019). In contrast, employees with average to low levels of HP and OP (Low HP Dominant and Average HP Dominant profiles) may not enjoy their work as much, making them less likely to display high levels of work engagement (Vallerand, 2010).

It is important to acknowledge that the High HP Dominant profile did not differ from the Average HP Dominant profile (or from the Moderately Low Passion profile in Sample 1) in terms of presenteeism, while these profiles differed from one another in relation to most other outcomes. These results suggest that average HP may be sufficient to protect employees from presenteeism and that higher levels of HP do have additional benefits for this outcome. In contrast, the higher levels of HP displayed by High HP Dominant employees seemed particularly important in the prediction of higher levels of work engagement, and of lower levels of work‐family conflict, turnover intentions, and CWB in most of the samples relative to those observed in the Average HP Dominant profile. Past research has also shown that HP was the most important predictor of work engagement (Birkeland & Buch, 2015) and turnover intentions (Gong et al., 2020), possibly because of its known associations with positive affect (Vallerand, 2010). Similarly, these benefits of HP are in line with the numerous studies (Gillet et al., 2016, 2017a) showing a positive effect of autonomous motivation (i.e., actions are driven by pleasure and choice) and of experiencing a greater sense of control related to when and how to engage into one's passion (e.g., van Steenberger et al., 2021). Future research will be needed to better unpack these mechanisms, and to achieve a clearer differentiation between the High HP Dominant and Average HP Dominant profiles.

5.4. Generalisability across samples

Our results supported the generalisability of the profiles and of their associations with predictors and outcomes across most subsamples considered in this multi‐sample investigation. Although some differences in profile size were observed in Samples 1 (nurses vs. other) and 2 (full‐time vs. part‐time), and some differences in outcomes were observed between full‐time and part‐time employees in Sample 2, these differences were restricted to a limited number of comparisons, and did not change the overall picture and generalisability of our findings. In this regard, these results contrast with previous studies suggesting that work passion may vary as a function of job settings (Fernet et al., 2014; Vallerand & Houlfort, 2019). However, our results remain aligned with prior research demonstrating the adaptive effects of a profile characterised by high HP and low OP on work‐related outcomes in very distinct work settings (Gillet, Morin, et al., 2023; Li et al., 2020; Morin et al., 2023b). By providing evidence of generalisability, our results are an important step forward in work passion research, supporting the desirability of generic interventions likely to be relevant to many employees.

5.5. Limitations and future directions

The present research has some limitations, which open the way to new research avenues. First, the fact that this study relied solely on self‐report measures increases the risk of social desirability and self‐report biases. To alleviate these concerns, it would be useful for future studies to consider incorporating objective measures (e.g., organisational data on work performance and absenteeism) and informant ratings of employees' functioning (e.g., colleagues, supervisors, spouse). Second, the present study was conducted among five mixed samples of employees working in France. Moreover, although the results revealed that demographic characteristics (including gender) were unrelated to profile membership and that our goal was never to recruit representative samples, it is worth noting that women were over‐represented in the five current samples. Further research is thus needed to generalize the current results among more balanced samples in different countries, languages, and cultures. Third, we used a cross‐sectional design, which does not allow establishing the temporal stability of the profiles or determining causal relations between the variables. Although studies provide support for some of the proposed associations (e.g., Gillet et al., 2023c; Houlfort et al., 2018), we cannot exclude the possibility of reciprocal or inverse relations between certain variables (Cole & Maxwell, 2003). Future studies should examine the nature of these relations using longitudinal methodologies better able to substantiate the temporal ordering of the observed associations. Finally, organisational support and work‐home segmentation were the only predictors of interest in our research. Yet, personal characteristics (e.g., personality traits: Vergauwe et al., 2022; job crafting: Slemp et al., 2021) can also play a major role in predicting employees' passion for their work. It would thus be interesting to examine how other personal characteristics (e.g., psychological capital, self‐efficacy) as well as hindrance (e.g., role conflict) and challenge (e.g., role responsibility and complexity) demands relate to employees' work passion profiles.

5.6. Practical implications

From an intervention perspective, our findings suggest that practitioners should be particularly attentive to workers exposed to an environment that does not promote clear physical, temporal, and behavioural boundaries between work role and personal life. Our results show that these workers were most likely to belong to the High OP Dominant profile, associated with the worst outcomes. Therefore, changes designed to increase work‐home segmentation could be leveraged to help employees stay away from this least desirable profile, and more generally to support a better functioning. For instance, it may be interesting to state clear segmentation norms and encourage balanced and healthier lifestyles (Kreiner, 2006), to create well‐being‐oriented workplaces, and to offer enabling versus enclosing work‐life policies (Bourdeau et al., 2019). Our findings suggest that interventions seeking to improve organisational support may also be useful. To nurture perceptions of organisational support, it may be helpful to promote a supportive culture to help managers break down the walls between themselves and employees. In such environments, managers and employees come to share power and to be more attuned to one another identity and culture, resulting in higher opportunities for the co‐creation of learning experiences and knowledge. Promoting procedural justice is also a meaningful way to increase organisational support (Kurtessis et al., 2017). Furthermore, OP could also be decreased at the individual level through coaching or counselling (Van Gordon et al., 2017). Finally, possible interventions include changes designed to reduce workload sustainably, which might help decrease the likelihood of membership into the High OP Dominant profile in the long run (Vallerand & Houlfort, 2019).

CONFLICT OF INTEREST STATEMENT

We have no conflicts of interest to disclose. All procedures performed in the current study involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments. Informed consent was obtained from all individual participants involved in the study.

Supporting information

Supporting Information S1

Gillet, N. , Morin, A. J. S. , Brault, S. , Becker, M. , & Verbeke, I. (2024). On the nature, predictors, and outcomes of work passion profiles: A generalisability study across distinct types of employees. Stress and Health, 40(6), e3495. 10.1002/smi.3495

DATA AVAILABILITY STATEMENT

This dataset has never been used in any previous or currently upcoming publication.