“Let’s stick together”: toward a serial mediation model about the impact of relational leadership on workers’ employability

Healthcare organizations must respond to complex, worldwide megatrends, such as demographic and digital shifts (Baptista et al., 2020; International Labour Organization, 2019). Specifically, nurses must cope with patients who are more likely to have multiple chronic diseases (Dall et al., 2013), making their work demands more complex. Moreover, the COVID-19 pandemic accelerated the implementation of technologies in healthcare, which subsequently transformed nurses’ jobs drastically (Isidori et al., 2022), resulting in technological skill gaps (PricewaterhouseCoopers, 2018). When nurses fail to meet these changing work demands, they become dissatisfied and might leave their profession entirely (Buchan et al., 2013; Lu et al., 2005; World Health Organization, 2020), herewith worsening the severe nursing labor shortage (International Labour Organization, 2019).

These changing work demands require nurses to develop new knowledge, skills and attitudes (Isidori et al., 2022). Highly employable workers have acquired the necessary knowledge, skills and competences and are thus better able to cope with all requirements at the workplace (Stoffers et al., 2020a; Stoffers, 2023). Employability, or “the subjective perception held by an employee (or by their supervisor) about their possibilities, in terms of competences, to obtain and maintain work” (Van der Heijden et al., 2018, p. 237), is thus paramount to both nurses’ career success and healthcare organizations’ performance (Fugate et al., 2021).

Recently, Leclerc et al. (2022) found that relational leadership (RL) contributes to positive nursing outcomes, but, so far, the effects of RL on employability, particularly in healthcare, remain underexplored (cf. Hoedemakers et al., 2023). Stoffers et al. (2020b) and Van der Heijden and Spurk (2019) identified the relevance of Leader–Member eXchange (LMX), for workers’ employability enhancement. Building on the principles of social exchange (Blau, 1964) as our underlying theoretical framework, we argue that supervisor’s RL might affect nurses’ behavior, and, in turn, their perceptions of support (PSS) and opportunities for competence development (POCD) (Wang et al., 2022), which foster their employability (Froehlich et al., 2014; Stoffers et al., 2014; Van der Heijden and Spurk, 2019). Indeed, to establish employability, organizations should invest in employees’ competence development (Martini and Cavenago, 2017). In doing so, the current study aims to shed more light on the mechanisms underlying the impact of RL on employability and hereby addresses Fugate et al.’s (2021) and Van Harten et al.’s (2020) call to examine how employer–employee interdependence in terms of POCD, stimulates employability.

This study therefore examines a serial mediation model in which supervisors’ RL predicts employability, with PSS for development and POCD being sequential mediators (see Figure 1).

RL is “a social influence process through which emergent coordination (i.e. evolving social order) and change (e.g. new values, attitudes, approaches, behaviors, and ideologies) are constructed and produced” (Uhl-Bien, 2006, p. 655). Relational leaders recognize the importance of relationships and focus on the interactions they have with others (Clarke, 2018). During interactions, they focus on “the need to be respectful, establishing trust, and for people to be able to ‘express themselves’” (Cunliffe and Eriksen, 2011, p. 1433). RL is thus characterized by inclusion, empowerment, caring, ethicality and having vision and intuition (Carifio, 2010). Inclusion concerns the positive value of every individual, regardless of dissimilarities (Komives et al., 1998), and empowerment associates with leaders’ choices to create value for individuals (e.g. autonomy enhancement; Honold, 1997). Caring emphasizes the importance of empathy in relationships (Regan and Brooks, 1995), and ethicality is about positive energy and integrity (Komives et al., 1998). Having vision and intuition is concerned with being explicit about goal achievement at the team and individual levels and stimulating commitment to goals (Regan and Brooks, 1995).

Since RL is a new concept, empirical evidence of its predictive value for work-related outcomes is scarce. Building on extant literature that evidenced positive associations with, for instance, innovative work behaviors (Kim, 2022), we investigate its impact on employability.

The construct of competence-based employability comprises “knowledge, skills and attitudes, or more general competencies to assess employability” (Van der Heijden et al., 2018, p. 237), and it contains five dimensions—occupational expertise, anticipation and optimization, personal flexibility, corporate sense and balance (Van der Heijde and Van der Heijden, 2006). Occupational expertise means possessing the right professional skills and knowledge (e.g. meta-cognitive knowledge) and being acknowledged socially by focal stakeholders. Anticipation and optimization refers to competence with self-initiated anticipation and optimization with regard to changes in an individual’s job or career in the future, focusing on realizing the best job or career outcomes. Personal flexibility encompasses reactive, passive adaptation to changes within and between jobs and organizations. Corporate sense relates to social capital, associating with the capacity to participate and perform through collaboration on and in teams, groups, departments and networks. Balance is paradoxical because it refers to balance between opposing interests, such as from employers and employees, regarding jobs, careers and private interests.

Although, to the best of our knowledge, up until now, employability has not been studied empirically as an outcome of RL, extant research suggests a link between the two. For instance, Hoedemakers et al.’s (2023) systematic literature review indicated that other leadership types, especially transformational leadership (e.g. Vermeeren and Van der Heijden, 2022) and LMX (e.g. Epitropaki et al., 2021), influence employability. Such leadership styles are characterized by inspirational motivation, intellectual stimulation, idealized influence and individualized consideration (Avolio et al., 1991), representing high-quality, dyadic relationships between a leader and followers (Graen and Uhl-Bien, 1995), much like RL (Uhl-Bien, 2006). Tu et al. (2022) demonstrated the importance of leaders’ secure-base support for employees’ career-related outcomes (i.e. job crafting). Furthermore, in education contexts, leadership, in terms of social relationships between lecturers and students, affects students’ perceptions of obtaining a job, and thus their employability (Alawamleh and Mahadin, 2022; Jackson and Tomlinson, 2020). Since a supervisor with a RL style engages in social exchanges with their employees, particularly caring about, supporting and empowering them, and Richards (2024) denoted RL as an important job resource for supporting nurses toward work-related outcomes, we argue that RL contributes to their employability.

Relational leaders may elicit positive employee outcomes because they signal greater support for development. They focus on dyadic supportive relationships (Cleary et al., 2018), including behaviors of inclusiveness, empowerment and caring (Clarke, 2018). Recently, Leclerc et al. (2022) argued that a relational leader motivates nurses and advocates their professional growth by focusing on providing support for development. Furthermore, a systematic review by Aleo et al. (2024) demonstrated that supervisors who adopt a RL style play a substantial role in the effectiveness of professional development of their subordinates, which may be because they signal the need for support for their development.

PSS for development, in turn, leads to POCD. PSS is associated with employees’ perceptions concerning how supervisors value their contributions, support them and care about their well-being (Eisenberger et al., 2002). As such, supportive supervisors operate as a job resource that “provides guidance, assistance, and feedback to subordinates” (Yang et al., 2018, p. 498) and facilitate training tailored to their employees’ needs (Komives et al., 1998), herewith raising awareness for POCD both inside and outside of the organization (Cunliffe and Eriksen, 2011). Consequently, PSS for development, for instance through delegating tasks (Van Vianen et al., 2019) positively affects employees’ perceptions of available development opportunities for them (Kuvaas and Dysvik, 2010), such as POCD.

Finally, POCD by practices such as training (Bozionelos et al., 2020) and functional rotation (De Vos et al., 2011) are likely to lead to perceptions of increased employability. Previous research already suggested that POCD, both informal (Froehlich et al., 2023) and formal (Van den Elsen et al., 2022), contribute to workers’ employability (Van der Klink et al., 2014). Since such opportunities are also evidenced to be of critical importance for career progression in healthcare (e.g. Aleo et al., 2024), we argue that they stimulate employability in that context as well.

In sum, based on the above-mentioned hypotheses we argue that nurses’ PSS for development and POCD mediate serially in explaining the association between RL style and nurses’ employability. Specifically, RL motivates nurses to continue their professional growth as they experience support for development (Leclerc et al., 2022). Supervisors foster nurses’ competence development through, for instance, guidance and feedback (Yang et al., 2018, p. 498), hereby increasing their perceptions of available competence development opportunities (Kuvaas and Dysvik, 2010). The latter, in turn, stimulates their employability (e.g. Bozionelos et al., 2020; Froehlich et al., 2023).

To test our serial mediation model, an empirical study was conducted in a homecare organization in the Netherlands. Our sample included direct nursing staff members (Backhaus et al., 2018) (n = 496) who held a qualification level of 1 through 6 on the European Qualification Framework (EQF). Due to COVID-19, nurses had to cope with challenging work demands at the time of data collection.

We collected two-wave, self-reported online survey data from this sample during October 2021 (T0) and April 2022 (T1) (i.e. adopting a six-month interval). The online questionnaires contained validated scales for all our study’s variables. Complying with principles of scientific integrity (NWO, 2023), participants received an informed consent letter about the objective of the study, their voluntary participation, the time needed to complete the questionnaire, and the anonymity and confidentiality of data processing. To maximize participation, we informed potential respondents about the response rate during the study. The research was approved by The Research Ethics Committee of the Open Universiteit (number: U202108211).

To enhance the reliability and validity of our study and to mitigate common-method bias (among others), we utilized procedural controls (Kock et al., 2021). For this, we established temporal separation when collecting data from the same respondents, using two waves (see Figure 1). Furthermore, to prevent response bias (Dolnicar, 2020), we executed a pretest of the questionnaire using a sample of four nurses.

The sample of nurses who provided full data at T0 and T1 consisted of 109 nurses (response rate was 21.98%; 95.40% female and 4.60% male), with a mean age of 49.49 years (SD = 11.24). Respondents’ education level varied from EQF 2 (9.2%) to EQF 6 (18.3%), with the majority having EQF level 3 or 4 (69.7%). 54.1% held a certified nurse assistant position, followed by baccalaureate-educated registered nurses (23.8%), vocationally trained registered nurses (15.6%), nurse assistants (3.7%) and nurse aides (2.8%). Job tenure was, on average, 15.5 years (SD = 12.20), and nurses worked, on average, 11.6 years (SD = 13.80) in their function.

We measured RL, PSS for development, POCD, and employability at two times using the same scales, which have been validated in extant research. Items for RL, perceived supervisory support for development and perceived opportunities for professional competence development were translated into Dutch using the translation-back-translation methodology (Brislin, 1970). The items for employability have not been translated as the original items were in Dutch.

Supervisors’ RL style was assessed using Carifio’s (2010) 25-item, 5-dimensional scale of each 5 items, which consists of inclusion (e.g. “My supervisor creates opportunity for professional and personal growth for supervisors and others”), empowerment (e.g. “My supervisor acknowledges the abilities and skills of others”), caring (e.g. “My supervisor steps out of his/her personal frame of reference into that of others”), ethicality (e.g. “My supervisor influences others by mutual liking and respect”) and having vision and intuition (e.g. “My supervisor provides inspiring and strategic goals”). The response format was a 7-point Likert scale ranging from “strongly disagree” to “strongly agree”.

Perceived supervisory support for development was assessed using Salminen and Miettinen’s (2019) 8-item scale. The response format was a 5-point Likert scale ranging from “strongly disagree” to “strongly agree”. A sample item was: “My supervisor supports me in learning new things.”

Perceived opportunities for competence development was assessed using Salminen and Miettinen’s (2019) 5-item instrument. The response format was a 5-point Likert scale ranging from “strongly disagree” to “strongly agree”. A sample item was: “My employer creates opportunity to receive training that promotes professional competencies.”

Employability was measured using the 22-item instrument from Van der Heijden et al. (2018), consisting of five dimensions—occupational expertise (5 items, e.g. “During the past year, I was, in general, competent to perform my work accurately and with few mistakes”), anticipation and optimization (4 items, e.g. “I consciously devote attention to applying my newly acquired knowledge and skills”), personal flexibility (5 items, e.g. “I adapt to developments within my organization”), corporate sense (4 items, e.g. “I share my experience and knowledge with others”) and balance (4 items, e.g. My work and private life are evenly balanced”). The response format was a 6-point Likert scale ranging from “strongly disagree” to “strongly agree”.

Control variables participants’ age, gender, education level, job tenure and years in function were controlled for in our data analysis. Since the inclusion of these controls did not affect the results, we present our results without them.

We used R (R Core Team, 2023) to conduct PLS-SEM, which allowed us to estimate a complex model that includes two multidimensional latent variables, and which accommodates the small sample size (Hair et al., 2019). We used the bootstrap method (10,000 iterations; Streukens and Leroi-Werelds, 2016), and measurement and structural models were evaluated using Hair et al.’s (2019) procedure. We assessed two structural models for hypotheses’ testing. Figure 2, and Model 1 exclude the mediators PSS and POCD, and thus examined the total association between RL and employability. Figure 2 and Model 2 include the mediators, herewith assessing both direct and indirect relationships among variables.

We evaluated the quality of the measurement model by examining the internal consistency and the convergent and discriminant validity of the latent constructs. We first assessed the reflective indicator loadings on corresponding constructs (Kock, 2014). Results demonstrated insufficient loadings, for one indicator in each case, for the dimensions inclusion, and vision and intuition of the RL style construct, and for the dimensions occupational expertise, anticipation and optimization, personal flexibility, corporate sense and balance for the employability construct; they did not exceed the threshold value of 0.708, as Hair et al. (2019) recommend. We thus excluded those seven indicators during our subsequent analyses. We then reassessed indicator loadings, which suggested that, for all remaining indicators, loadings exceeded the threshold value, except one indicator for the inclusion dimension of the RL style construct. We thus also excluded this indicator, resulting in loadings that all exceeded the threshold value. We used Cronbach’s alpha (α) and rho_A to examine the constructs’ internal consistency. Our results suggested an acceptable reliability for all measurement scales (see  Supplementary File 1), since for all constructs reliabilities exceeded both the threshold values of 0.65 for α (i.e. ranging from 0.706 to 0.911) (Vaske, 2008) and the cut-off of 0.70 for rho_A (i.e. ranging from 0.766 to 0.974) (Fornell and Larcker, 1981)

Subsequently, we examined each construct’s convergent validity using average variance extracted (AVE). Results suggested acceptable convergent validity for all measures because AVEs values ranging from 0.605 to 0.768 and thus exceeded the 0.5 threshold value (Hair et al., 2021). Following Henseler et al. (2015), we assessed the discriminant validity for all measures using bootstrapping (10,000 iterations; Streukens and Leroi-Werelds, 2016) to investigate whether the HeteroTrait-MonoTrait (HTMT) correlations for the inter-construct correlations differed from 1. Our results suggested acceptable HTMT correlations, since the upper bounds of the 95% confidence intervals (CI) were lower than 1 (Henseler et al., 2015), except for those that between caring and ethicality (both dimensions of RL). Hence, the inter-construct correlations between RL dimensions’ caring and ethicality exceeded the upper value of the 95% CI [0.914, 1.023] suggesting discriminant validity has not been established (Henseler et al., 2015). We therefore aggregated the supervisor’s RL style dimensions of caring and ethicality into one dimension, so that the different meanings of all the dimensions are consequently reflected by the corresponding four different dimensions (Van der Heijden and Verhelst, 2002) and reassessed the measurement model (see  Supplementary File 2).

Reflective indicator loadings (Kock, 2014) demonstrated insufficient loadings, for one indicator in each case, for the dimensions of inclusion, caring/ethicality, vision and intuition, for the RL style construct, and for the dimensions of occupational expertise, anticipation and optimization, personal flexibility, corporate sense and balance, for the employability construct (i.e. they did not exceed 0.708, as Hair et al. (2019) recommend). We thus excluded these indicators during subsequent analyses. A reassessment of the measurement model suggested acceptable reliabilities for all measurement scales (i.e. α ranged from 0.650 to 0.929; and rho_A from 0.654 to 0.958); and acceptable convergent validity for all measures, since the AVEs for each construct exceeded the 0.5 threshold value, because they ranged from 0.576 to 0.760 (Hair et al., 2019) (see  Supplementary File 3).

We reassessed discriminant validity between the measures using the method of bootstrapping HTMT correlations (10,000 iterations; Henseler et al., 2015; Streukens and Leroi-Werelds, 2016). Our results suggested a lack of discriminant validity between the measures for anticipation and optimization, on the one hand, and corporate sense, on the other hand. However, since the upper value of the 95% CI was close to one [0.914, 1.023], and since the employability measurement instrument (Van der Heijden et al., 2018) has been thoroughly validated in extant studies (e.g. Crans et al., 2021), we proceeded without further alterations to the measurement model (see  Supplementary File 4).

Table 1 revealed that correlations among the variables under study were aligned with our expectations.

To mitigate bias during the regression analyses, we tested for multicollinearity using Variance Inflation Factors (VIF; Hair et al., 2019) (see  Supplementary File 5). Our results indicated VIF values that did not exceed the cut-off of 5 (Becker et al., 2015), because values ranged from 1.264 to 4.892 herewith demonstrating an absence of multicollinearity. We bootstrapped the estimated structural model using 10,000 iterations to assess the significance of the structural paths with bias-corrected 95% CI (Streukens and Leroi-Werelds, 2016; see Table 2 for the significant structural paths or  Supplementary File 6 for all structural paths).

Contrary to our expectations, no path coefficient for the relationship between RL style (T0) and employability (T1) differed from zero. Hence, H1 was not supported with our data. Our results suggest a significant coefficient for the relationship between the empowerment dimension of RL (T0) and PSS for development (T1), but not for the other dimensions. Thus, H2 was partially supported with our data. Our results revealed a significant coefficient for the relationship between PSS for development (T1) and POCD (T1). Hence, H3 was supported with our data. We found that POCD (T1) was associated positively with the employability dimensions of anticipation and optimization, corporate sense and balance (T1), but not with occupational expertise and personal flexibility. H4 was thus partially supported with our data. We found significant serial mediation effects of the RL style dimension empowerment (T0) on both the employability dimensions of anticipation and optimization, and balance (T1), through PSS for development, and through POCD. Other indirect effects were non-significant. H5 was thus partially supported with our data.

We assessed the model’s in-sample predictive power (Schmueli and Koppius, 2011) by calculating the coefficient of determination (R²) of the endogenous constructs. RL (T0) explained 20.9% of the variance in PSS for development (T1). RL (T0) and PSS for development (T1) explained 62.1% of the variance in nurses’ POCD (T1). RL (T0), PSS for development (T1) and nurses’ POCD (T1) explained 21.9% of the variance in anticipation and optimization (T1), 16% of the variance in corporate sense (T1) and 11.7% of the variance in balance (T1). Hence, we conclude that a supervisor’s RL style (particularly empowerment) stimulates nurses’ PSS for development, and their POCD, which subsequently enhance their employability regarding anticipation and optimization, and balance.

Our findings have both theoretical contributions and practical implications.

First, this study contributes to the employability literature, specifically to research that emphasizes the importance of leadership to employability. Although in previous research, a significant effect of leadership styles, such as servant leadership (Wang et al., 2019), on employability was shown, to the best of our knowledge, this is the first empirical work that focuses on RL as a possible predictor for workers’ employability. Hoedemakers et al. (2023) argued for a positive association between leadership and employees’ employability, herewith laying the ground for this empirical work. As such, the current study adds the concept of RL to the employability literature, with its focus on respect, trust and self-expression in dyadic relationships (e.g. Clarke, 2018).

Whereas in earlier research servant leadership was found to influence employability directly (Wang et al., 2019), we conclude RL effects employability indirectly. This finding may occur because RL’s conceptualization is considerably different from the one of servant leadership. Contrary to RL (e.g. Uhl-Bien, 2006), servant leadership (Greenleaf, 1970) focuses on one angle, namely, the angle of the leader holding a formal leader position, and thus largely neglects the role of, in terms of interaction with, their followers. These traditional leadership theories utilize leaders’ characteristics behavior and personality, to explain how leadership affects followers toward goal achievement (e.g. enhancing their employability). RL, in contrast, denotes leadership as the constitution of reciprocal relationships between leaders and followers, hence stresses the role of followers in addition to the role of their leaders, when trying to understand the impact of leadership style on effects under study (Clarke, 2018). Hence, indicated by our empirical work, RL appears to influence (desired) outcomes, through the empowerment that is created by means of employees’ perceptions about the nature and characteristics of this particular leadership style (cf. Carifio, 2010).

As such, RL style does lead to PSS that influence POCD, which subsequently enhance followers’ employability.

Altogether, we found support for a fully mediated serial model that explains the mechanisms (Kazdin, 2007) behind the association between RL and employability. RL appears to influence employability as relational leaders signal support for development and provide opportunities for competence development, which in turn enhance workers’ employability. This finding accords with previous research that suggested that workers should make use of learning opportunities to build up skills and competences that make them more employable (e.g. Froehlich et al., 2014). Principles of social exchange (Blau, 1964) serve as our underlying theoretical framework—nurses seem to reciprocate supervisors’ rewards (e.g. empowerment, PSS for development and POCD) by developing knowledge and skills, resulting in an increase in their employability in terms of anticipation and optimization, and balance.

Second, this study contributes to the RL literature by evidencing additional outcomes of this recent leadership concept. Extant studies identified RL as a predictor of positive employee outcomes, including innovative work behaviors (Kim, 2022), and improvements to nurses’ work environments (Leclerc et al., 2022). By focusing on employability enhancement as our outcome variable, we enrich the RL literature by identifying that RL, especially the empowerment dimension, contributes to new outcomes such as PSS for development, POCD, and employability in terms of anticipation and optimization, and balance. The association between RL and employability appears to be indirect, through the mediators of PSS for development and POCD.

Besides, we add to the RL literature by testing the concept in a new occupational context. We, therefore, expand our knowledge on Carifio’s (2010) RL construct that originated from a sample of high school teachers in the United States to a sample of direct nursing staff members within the Dutch nursing homecare.

Another contribution to the RL literature lies in testing the impact of separate dimensions of the construct. Most earlier work treated Carifio’s (2010) RL construct as one-dimensional (e.g. Ansong et al., 2024), while the construct was originally launched as a five-dimensional one (Carifio, 2010). The current study is one of the very few that assesses the five dimensions separately, herewith further validating the concept’s multidimensional nature. We reveal that it is difficult to distinguish caring and ethicality statistically, and we thus aggregated caring and ethicality into one dimension, which demonstrated acceptable internal consistency, and both convergent and discriminant validity. However, these findings might depend on characteristics of our specific sample, and we therefore suggest further validation of the instrument.

This study has implications for HRM and supervisors in homecare organizations. In particular, we suggest that focusing on policies and practices that stimulate supervisors’ RL, particularly empowering subordinates, is a valuable evidence-based HR strategy for working organizations.

To induce the necessary conditions for RL development, earlier work (Hong et al., 2019) revealed the critical role of collaborative HRM (e.g. job rotation) that stimulate social exchange (Liao et al., 2007), and through this experiences of RL. Moreover, Ansong et al. (2024) stressed the importance of open communication to foster employee’s participation in decision-making, which is associated with empowerment and inclusion (Carifio, 2010), when aiming to implement a RL style. Ansong et al. (2024) proposed policies and practices that are linked to both leader and leadership development that focus on interpersonal relationships, and that effectuate intra- and interpersonal competences (e.g. self-awareness and social skills). Homecare organizations should thus provide development practices that subsequently enhance mutual trust, respect and emotional intelligence, which relate to RL (Clarke, 2018).

We collected self-reported data, which might be prone to some common-method bias, and constructs were measured using nurses’ survey responses, and thus our results might have been affected by a single-source bias as well (Podsakoff and Organ, 1986). The study’s two-wave panel design mitigated some of these effects, but future research is recommended wherein multi-source data from leader–follower pairs are collected.

To enhance our scholarly knowledge about the impact of RL on employability, future studies should investigate possible mediation effects by other variables, such as mutual trust, respect and emotional intelligence (Clarke, 2018), which would advance the RL and employability literatures. We also call for more research wherein the role of possible moderators, such as age and gender are considered.

Furthermore, this empirical work was conducted using a survey research design that was based on deductive reasoning, and it was thus limited by theoretically proposed relationships among four constructs in a conceptual model. To gain greater insights into our model’s relationships, future studies should use inductive, qualitative research designs (Neuman, 2014).

Finally, Carifio’s (2010) RL construct still lacks empirical work aimed at cross-validation of its predictive value across contextual and occupational settings. Therefore, we invite scholars to continue research in this field, ideally by further examining its multidimensional character.

This study evidences the role RL plays in enhancing nurses’ employability. Both perceived supervisory support for development and perceived opportunities for competence development explain the positive association between RL and competence-based employability.

We are grateful to Dr Anne Richter for sharing her experiences with studying leadership within a healthcare context.

The Supplementary material for this article can be found online.