Mentorship has been described as a pedagogical tool. It is a process between a professionally active person (the mentor) and a protégé, which is democratic, involves deep reflection, and includes psychological and emotional support (Arnesson and Albinsson, 2017). Traditionally, mentoring involves a formally arranged one-on-one relationship between an experienced mentor and a less experienced mentee at an early career stage (Opengart and Bierema, 2015). However, informal mentoring relationships can also exist in workplace settings (Baugh and Fagenson-Eland, 2007).

Diversifying career and employment contexts have led research and practice to understand mentoring as “a multiple developmental relationship phenomenon” (Higgins and Kram, 2001, p. 264), occurring in new forms such as peer, virtual, topical, situational, and (inter-) cultural mentoring (Irby et al., 2017). Additionally, the stage in which mentoring occurs has extended from focusing on early career to later career stages, as well as specific purpose-related stages such as career transitions. The practices described in mentoring are “multifaceted, ambiguous […] contextually driven” and process-related (Gallucci et al., 2010; Mullen, 2012, p. 8). In recent years, we have also seen a rise in professional mentoring programs offering a criteria-based matching of mentor and mentee and a structured process. Mentoring has a developmental function for personal learning and growth, emphasizing the strong individualized learning type, including close support and feedback.

Mentoring research in educational contexts has been conducted with an academic or school-based focus. However, school-based research mainly focuses on teaching staff and their job satisfaction and retention (Renbarger and Davis, 2019). The rising complexity and systemic change of educational leaders’ roles call for structured mentoring opportunities to offer support and continuous development. Researchers have reported an effect of mentoring on educational leaders’ identity and self-efficacy through critical reflection (Rhodes and Fletcher, 2013; Muir, 2014). Gimbel and Kefor (2018) highlight the importance of deep thinking and joint reflection for decision-making on staff retention in a structured principal mentoring initiative. The ever-changing demands on educators call for measures to reduce the risk of “work-related stress, anxiety, burnout and increasing work-life imbalance,” and mentoring can be a helpful tool in this regard (Kutsyuruba and Godden, 2019, p. 229). Consequently, these issues call for a focus on well-being in education since leaders are concerned with the well-being of students, lecturers, and administrative staff and must provide learning environments conducive to these needs. However, as Kutsyuruba and Godden (2019) point out, educational leaders also need to be attuned to their own well-being as a basis for fostering the well-being of others.

Next, to basic human needs, well-being can be seen as one of the elementary topics in life. Behavioral scientists investigate “the factors that lead people to think and experience their lives in positive versus negative ways” (Diener et al., 2018, p. 253). Subjective well-being (SWB) describes people’s subjective evaluations of life focusing on a hedonic view of well-being as contrasted with a eudemonic view (Diener, 2009). The psychological concept of SWB is an umbrella term that comprises thoughts, judgments, and feelings (Diener and Ryan, 2009). Das et al. (2020) connect the feeling of well-being to the emotional or affective dimension of SWB, whereas they link thinking about it to an evaluative or cognitive dimension. The theoretical foundation for well-being is dispersed across various theories and can be argued to suffer from a lack of a coherent framework (Das et al., 2020).

Social support has repeatedly been found to positively affect SWB. The size of the social network, the quality of relationships, and interaction frequency all play a vital role (Chou, 1999; Olsson et al., 2014; Sandstrom and Dunn, 2014). High SWB also benefits supportive relationships, work performance, and resilience (Helliwell et al., 2020). In educational leadership, the research focused on SWB is based on quantitative methods and the individual perspectives of leaders. Amongst other themes, COVID-19 initiated an impetus to research the well-being of educational leaders, including principals, middle-managers, administrative staff, and students (Swapp, 2020; Chen et al., 2023). Factors typically impacting principals’ levels of SWB include workplace bullying, balancing work and family demands, and working conditions (Buonomo et al., 2020; DeMatthews et al., 2023). Working conditions influencing principals’ job performance and well-being include the institution’s organizational culture and work motivation (Ekosusilo, 2020). Beausaert et al. (2021) identified social capital as collegial support from inside and outside the school environment, collaboration, and trust in management to predict principals’ well-being positively. Chen et al. (2023, p. 15) highlight the recent research focusing on “interaction with leadership and on promoting well-being.”

Mentoring has been discussed widely as a tool to support well-being (Hobson and van Nieuwerburgh, 2022). Longitudinal data from Boeder et al. (2021) suggest that career mentoring in emerging adulthood is linked to later flourishing and SWB. Firzly et al. (2021) showed that the mentor’s need-supportive interpersonal behaviors were linked to greater autonomous motivation and, thus, to greater well-being. In a comparative study in Portugal, Simões and Alarcão (2014) identified no significant differences between mentored and non-mentored students in how a mentoring program satisfied their basic psychological needs and, in turn, improved their well-being. The early focus of mentoring and well-being research and practice was on students and teachers. However, in recent years, this focus has shifted to educational leadership and its development (Kutsyuruba and Godden, 2019; Hobson and van Nieuwerburgh, 2022), suggesting a need for mentoring structures for educational leaders. According to Howley et al. (2002), more than two-thirds of new school leaders in a leadership program ranked mentoring as a key component of the program. In a sample of professors and Master’s students in higher education, Woloshyn et al. (2021) focused on students’ well-being, support, and mentorship. They found that personal and professional support (including mentoring) predicted well-being for professors, whereas, for students, the professors’ mentorship had little effect on their well-being. Recently, Connery and Frick (2021) reported that leaders perceived a development in their communication, time management, leadership competencies, and situational problem-solving through participating in a formal mentoring program. Furthermore, Huffman (2018) reports that mentored principals indicated that they developed interpersonal collaboration, school scheduling, delegation, and designing professional learning. A recent study on the effect of mentoring VET leaders revealed that leadership presented a driving mechanism, whereas well-being was the ultimate outcome (Prummer et al., 2024).

Based on the recent findings, we identify two foci in the latest research: well-being and developing necessary leadership competencies. This raises the question of whether mentoring programs that lead to a rise in well-being also result in (well-being) competencies that leaders can apply in their work contexts. Our understanding of competence is based on a dispositional construct that has to be theoretically and cognitively understood to be executed in a real-life setting (Pittich, 2014). Well-being research has previously linked high well-being with EI or emotional competence (Schutte et al., 2002; Diener and Ryan, 2009).

In recent years, there has been a trend to focus on emotions in the workplace and understand how they influence how we engage, make decisions, and communicate. The concept of EI subsumes this. However, as Roberts et al. (2008, p. 710) put it, there are “knowns and unknowns” when it comes to EI. Epistemologies describe EI as “a competence, a skill, an adaptive outcome, a set of cultural beliefs, or some other construct” (Roberts et al., 2008, p. 711). Most common definitions of EI across various models refer to “the abilities to accurately perceive emotions, to access and generate emotions to assist thoughts, to understand emotions and emotional knowledge, and to reflectively regulate emotions in order to promote emotional and intellectual growth.” (Mayer et al., 2004, p. 197). Thus, EI can be described as a “co-operative combination of intelligence and emotion”; in other words, it is something people can feel and express because they are competent to do so and apply this competence in different contexts (Jonker and Vosloo, 2008, p. 21).

Three prominent streams of EI research conceptualize EI as either a trait, an ability, or a combination of the two (Ashkanasy and Daus, 2005; Pérez et al., 2005). Trait EI is measured using self-report questionnaires, whereas ability EI—also known as cognitive-emotional ability—is measured using maximal performance tests. Because emotional experiences are subjective, they are not accessible to objective grading methods (Spain et al., 2000; Furnham and Petrides, 2003). Being emotionally intelligent does not imply that people can simultaneously apply and regulate this knowledge in real life. This understanding has led some researchers and practitioners to recalibrate from speaking about and assessing EI to focusing on (social) emotional competence. Thus, as Bar-On et al. (2007, p. Xiii) put it, this raises the question of whether we can “educat[e] people to be emotionally intelligent.” Grant (2007) and Tschannen-Moran and Carter (2016) provide evidence that training fosters EI in coaching. In educational leadership development, Núñez et al. (2023) evaluated leaders’ EI competencies and found significant positive differences in social awareness and relationship management. However, development programs aiming at or integrating EI competencies are still scarce.

Based on the theoretical background provided above, we can now integrate the conceptual and empirical findings on mentoring, SWB, and EI and how they interrelate. In addition to SWB, discussed above, EI relates to other constructs in psychology and social sciences, such as empathy, physical health, social interaction, and performance at school and the workplace (Bar-On et al., 2007). Trait EI is conceptually and empirically related to happiness and well-being (Palmer et al., 2002). Evidence suggests that EI correlates with various outcomes that signal social-emotional well-being (Zeidner et al., 2012) and improves well-being and social relationships (Nelis et al., 2009). More precisely, EI might help to manage emotionally challenging encounters (Lenaghan et al., 2007) and to work through positive and negative emotions to impact job satisfaction (Kafetsios and Zampetakis, 2008). However, other studies may fail to find a relationship between EI and SWB (Zeidner and Olnick-Shemesh, 2010).

In education, Lucas-Mangas et al. (2022) showed that adjusting SWB predicts the mismatch in teachers’ work through regulating relationships, the school environment, and having a purpose in life. In a systematic review of EI and school leadership, Gómez-Leal et al. (2021) identified competencies such as self-awareness, self-management, empathy, communication, and conflict management. This implies the critical nexus of EI and SWB for effective leadership “that influences the school climate, teachers’ commitment, and well-being; family and community partnerships, and student outcomes” (Mahfouz et al., 2019, p. 2). The changing school environment highlights the need for development programs for (middle space) leaders since school quality improvement relates to educational leadership preparation and development (Eacott and Asuga, 2014; Armstrong, 2015). In the school environment, also related to the challenges of the 21st century, amplified emotional reactions based on conflict, change, and ambiguity are increasing. Thus, Schmidt (2010) advocates combining cognitive and emotional factors in school leadership development based on the above-described contexts.

In the era of lifelong learning, professional development must occur in different career stages and management levels. In recent years, professional development has shifted from a one-sided focus on teachers and students to designing programs for educational leaders. With the theoretical understanding of SWB and EI as a basis, the question arises of how professional development can support leaders’ EI and SWB. Most leadership programs provide a one-size-fits-all perspective on development, but there have also been context-and competence-specific initiatives in different countries (Núñez et al., 2023). Additional support structures are necessary alongside professional development, and mentoring can bridge this gap (Faizuddin et al., 2022).

There is a need for mentoring programs to support and develop leaders’ EI. These might be based on experience from providing mentoring opportunities to address SWB of mentees and evidence of how EI is related to SWB (Middlewood and Bush, 2013). Researchers point out that EI competencies can be developed through coaching and mentoring (Boyatzis, 2007; Corrie, 2015). Research on the connection between mentoring and EI highlights the perceived contribution to mentees’ EI by influential mentors (Shapira-Lishchinsky and Levy-Gazenfrantz, 2016) and an indirect association between mentees’ EI and trust in mentors (Chun et al., 2010). Further research is scarce or investigates additional constructs such as intimacy (Bennetts, 2002), salary (Rode et al., 2017), and leadership styles (Yong, 2013). The career-enhancing potential of EI and mentoring—and the research gap in this area—call for a more detailed look at connecting the two.

To address RQ1—to validate the SREIT used to test for EI in the sample—we conducted exploratory and confirmatory factor analyses to explore and confirm the underlying factor structure for the South African VET context because previous studies highlighted inconsistencies in the factor structure especially across different cultures (Davies et al., 2010; Pisnar et al., 2022).

To address RQ2, we focused on identifying the differences between the groups. As a rule for testing our research hypotheses (below), we adhere to the following principles. For direct effects, t-test statistics and p-values are produced; if the p-value is less than 0.05, the difference is statistically significant. A bootstrapped 95% confidence interval (CI) is provided for indirect effects.

Since the mentoring framework comprises three mentoring types, we were interested in determining which type is most suitable for developing EI. We used the leaders who participated in the mentoring framework as the sample.

Taking the VET system in South Africa as a broader research context, we additionally wanted to understand if the perceived importance of EI (regarding job position, VET school, and VET sector in South Africa) influences the identified factors. The occupational role, the organization, and the sector in which individuals work need to be considered.

Finally, in line with previous research investigating gender differences in EI among students, we wanted to investigate the effect a leader’s gender plays in the South African VET system. Previous research in the South African context showed that teachers perceived their female school leaders to have significantly higher inter- and intrapersonal emotional competence than male respondents (Grobler, 2014).

The study is based on data collected in the VET sector in South Africa. We e-mailed VET leaders who participated in the mentoring program of the first and second cohorts, while the third cohort received information about the study during an online synchronous session in which the data collection occurred. To compose the control group, the Department of Higher Education and Training facilitated access to VET leaders who did not participate in the mentoring program, using official communication channels. The treatment and control group composite reflected the whole VET sector in South Africa. All participants were asked to participate voluntarily, and information on data handling, security, and participants’ rights was given verbally and/or in written form. The data collection with those participating in the program (N = 48) took place from October to November 2022. The data collection with the control group (N = 95) occurred from April to May 2023. The data collection was administered through Unipark, an online survey tool ensuring participants’ anonymity and confidentiality to adhere to ethical data handling. We removed four participants from the sample because they did not meet the inclusion criteria. The remainder completed all the questions of the survey. Table 1 gives an overview of the socio-demographic variables of the sample.

We employed G*Power software (version 3.1.9.4) (Faul et al., 2007) to calculate minimum sample sizes, carefully considering the effect size as a critical factor in this calculation. Small effect sizes were disregarded, as detecting them may lack real-world and practical significance (Baicus and Caraiola, 2009; Peeters, 2016). Accordingly, we input medium effect sizes in G*Power, as their minimum sample size requirement is stricter than that of large effect sizes. We set alpha at 0.05 and use two-tailed tests. Minimum sample sizes were determined: 84 for correlations, 55 for mediation and moderated mediation analysis, and 43 per group for the Mann–Whitney test. Considering all the minimum sample size requirements, the total sample size of 139 is sufficient for all statistical tests to have a power of at least 0.8. However, when considering the sample sizes of the separate groups (see Table 1), if only managers and heads of department (HoDs) of the mentored group were considered, the sample size would be 40, short of the 43 per group requirement for the Mann–Whitney test to have sufficient power. Accordingly, we included the eight (vice) principals in the mentored group for our statistical analyses. For the group that was not mentored, separate analyses were conducted for managers (47) and lecturers (44) who met the criteria. For factor analysis, the minimum sample size recommendations have been disputed in the literature for decades (Taherdoost et al., 2022), with recommendations based on a constant value or a minimum number of observations per variable. In recent research, it is argued that, when using a constant value, the absolute minimum is 50 observations, and when using the ratio of observations to variables, the ratio should be at least larger than 3:1. When considering both scenarios, we judged our sample size to be sufficient.

We used the SREIT, which was developed and validated by Schutte et al. (1998) and based on the trait EI model by Salovey and Mayer (1990). Schutte et al. (1998, p. 174) describe EI “as a somewhat enduring, trait-like characteristic”, highlighting the potential for individuals with low EI scale scores to benefit from “special guidance, training or support” (p. 176). This understanding reflects the developmental nature of EI and emphasizes the need to understand more about educational tools for this purpose. Previous studies investigating the SREIT showed good internal consistency, test-rest reliability, and construct validity (Pisnar et al., 2022). It is one of the most extensively used EI tests (O’Connor et al., 2019). A further advantage over other instruments is its brevity and free availability for research. The SREIT comprises 33 statements, much shorter than the leading commercial trait EI instrument, the Bar-on, with 133 items. Three statements are reverse-scored, and participants use a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). It tests self-reported trait EI and consists of four factors: optimism/mood regulation, appraisal of emotions, social skills, and utilization of emotions.

Jonker and Vosloo (2008) culturally validated the SREIT in the South African higher education context. They identified six factors (positive affect, emotion-others, happy emotions, emotions-own, non-verbal emotions, and emotional management). Results indicated significant differences between gender and language groups. However, further research points to inconsistencies in the factor structure. Pisnar et al. (2022) found four factors (although they only used 21 of the 33 items). Davies et al. (2010) stated that an expert panel found 17 items unsuitable for analysis and thus used only 10 items, finding a 5-factor solution.

In addition to the SREIT, we included questions concerning socio-demographic characteristics (see Table 1), the perceived development of EI through each of the three mentoring types employed in the program, and the perceived importance of EI for the leaders’ position, VET-school, and VET sector in South Africa (How important would you rate EI for your current job position/your VET school/the VET sector in South Africa?).

For all statistical analyses, the Statistical Package for Social Sciences (SPSS, version 29) was used. An exploratory factor analysis (EFA) was conducted to address RQ1, exploring the underlying factor structure using Promax rotation (which allows for factors to exhibit some degree of correlation between them). Factors with eigenvalues greater than one were retained. An iterative process was followed where items with communalities less than 0.3 and factor loadings of less than 0.6 were systematically removed, and subsequent EFAs were performed. After several iterations, the final factor structure was obtained (see Results section).

Following the EFA, a confirmatory factor analysis (CFA) was conducted to confirm the underlying factor structure. Once the factors were confirmed, the reliability and validity of the instrument were established. For the former, each factor’s composite reliability (CR) values, representing each factor’s internal consistency, were calculated. A CR of 0.6 or greater is deemed acceptable in exploratory research, and a value of 0.7 in general (Hair et al., 2022). Since this instrument was used for the first time in our research context, we deemed a value of 0.6 or more to be sufficient. Construct validity, which consists of convergent and discriminant validity, was performed to establish the instrument’s validity. Convergent validity reflects the extent to which all items of a construct converge, thereby explaining the variance of the items. It is established when the construct’s average variance extracted (AVE) is higher than 0.5 (Hair et al., 2022). Discriminant validity is the extent to which a construct is truly distinct from other constructs and is verified when the value of correlations between constructs is less than the square root of the AVEs of their constructs (Hair et al., 2022).

To address the remaining research questions, it is important to note that the factors were created by averaging over the items that loaded onto them; thus, continuous variables were created. We tested the normality of our distribution to decide whether parametric tests could be used to analyze the data. The Shapiro–Wilk test indicated that the underlying distributions for all six factors were non-normal; accordingly, non-parametric tests were used.

For RQ2, the non-parametric Mann–Whitney U test (Z_U) was used to test for differences between two independent groups. For RQ3, descriptive analysis was used. For RQs 4 and 5, mediation effects between variables were tested through PROCESS Model 4 and the moderated mediation effects using PROCESS Model 59 (Hayes, 2022). For direct effects, t-test statistics and p-values were produced. A bootstrap sample with a 95% confidence interval (CI) is provided for indirect effects. When interpreting the effects of the mediation and moderated mediation analyses, categories of male were coded “1” and female “0,” or lecturer “1” and manager “0.”

After several iterations that involved removing items, the final EFA indicated an acceptable Kaiser-Meyer-Olkin (KMO) value of 0.549, and Bartlett’s test of sphericity (p < 0.001) suggested the feasibility of dimension reduction. Six factors were extracted, collectively explaining 73.0% of the total variance (Table 2).

Factor 1 (items 13 and 14) was named social engagement. Factor 2 (items 33 and 5, both reverse-scored) was named empathy difficulty. Factor 3 (items 1 and 2) was named resilience. Factor 4 (items 29 and 32) was named emotional perceptiveness. Factor 5 (items 21 and 9) was named emotional regulation. Factor 6 (items 4 and 7) was named trustworthy visionary.

A CFA confirmed the underlying factor structure (RMSEA = 0.059; CFI = 0.906; GFI = 0.935), all indicating a good model fit (RMSEA less than 0.08), comparative fit index, and goodness-of-fit index (greater than 0.9) (Table 3).

Regarding measurement invariance, configural invariance was established for the non-mentored leaders and lecturers, and the mentored leaders, meaning the 6-factor structure held for both groups. While there were some indications of potential challenges in achieving metric, scalar, and strict invariance, it is noteworthy that imposing the strictest levels of invariance has been regarded as excessively stringent (Flake et al., 2022).

The CR value for each factor was calculated using the factor loadings of the EFA. Table 4 shows that the CR values for all factors were above 0.6, establishing the instrument’s reliability. Table 4 also shows that all AVE values were above 0.5, establishing convergent validity.

Table 5 shows the correlations between the constructs with the square root of the AVE in brackets.

The absolute values of the correlations were less than the square root of the AVE, thereby establishing discriminant validity. Thus, by establishing both convergent and discriminant validity, the construct validity of the instrument was established. In summary, the EFA and CFA provided good results and indicated a six-factor structure including 12 items for the VET context in which we apply the SREIT.

Mann–Whitney U tests were used to determine whether there were any significant differences in the six identified factors between occupations. When comparing the factor scores between the non-mentored leaders and lecturers (N = 91) and the mentored leaders (N = 48), only the score for Factor 2 Empathy difficulty differed significantly (Z_U = −2.146, p = 0.032), with the score for the non-mentored group being significantly higher (Mdn = 4.00 (IQR = 1.00)) than those of the mentored group (Mdn = 3.50 (IQR = 1.00)). When comparing the factor scores between the leaders (N = 47) and the lecturers (N = 44) in the non-mentored group, no significant differences were found (Z_U ranged from 0.061 to 1.152, p-values ranged from 0.249 to 0.951).

To answer this RQ, we used descriptive statistics to show the differences (M, Mdn) between the responses of the managers, the HoDs, and the principals using a 5-point Likert scale ranging from 1 (not at all) to 5 (extremely) (Table 6). Table 6 shows that leaders across all occupational roles (managers, HoDs, and principals) perceived peer group mentoring to best support them in developing EI with an overall mean of 4.17 (SD = 0.75; range of M = 4.00–4.23). When comparing the overall score, the key performance area mentoring was perceived to support developing EI the least (M = 3.63), followed by the individual professional mentoring (M = 3.71).

For the mediation analyses, various models were considered, with the independent variable (IV) being the different occupations (e.g., manager or lecturer), the dependent variable (DV) being one of the six factors, and the mediator (MED) being the various items measuring views on the importance of EI. The data composite reflects the whole VET sector in South Africa, allowing for an individual perspective on school level and a broader perspective of the sector. Since these different IV, DV, and MED combinations produced several models, only those where significant results were obtained are reported. We start with a discussion on the mediation analyses, followed by a discussion of the moderated mediation analyses.

The significant effects of the mediating analyses are reported in Table 7.

For Figure 1, only the indirect effect is significant, indicating that the view of the importance of EI for one’s current job position has a mediating effect between the occupational role and empathy difficulty, resilience, and emotional regulation. The negative effects suggest that, on average, being a lecturer is associated with a decrease in each of the three factors compared to being a manager.

For Figure 2, only the indirect effect is significant, indicating that the view of the importance of EI for the VET school has a mediating effect between the occupation role and emotional perceptiveness and trustworthy visionary. The negative effect suggests that, on average, being a lecturer is associated with a decrease in each of the two factors compared to being a manager.

For Figure 3, only the indirect effect is significant, indicating that the view of the importance of EI for the VET sector has a mediating effect between the occupation role and resilience and trustworthy visionary. The negative effect suggests that, on average, being a lecturer is associated with a decrease in each of the two factors compared to being a manager.

For the moderated mediation analyses, gender was added as the moderator (MOD). To determine whether gender moderated the perceived importance of EI in the identified factors, we conducted a moderated mediation analysis. Only significant results are reported.

For Figure 4, only path b is significantly moderated (t = 2.11, p = 0.038), meaning that the effect of views on the importance of EI for one’s VET school on trustworthy visionary varies based on gender. The positive effect of 0.362 (SE = 0.172) suggests that being male is associated with higher trustworthy visionary than being female.

For Figure 5, paths b and c are significantly moderated. The fact that the b path is significantly moderated (t = −2.33 p = 0.022) indicates that the effect of views on the importance of EI for the VET sector on empathy difficulty varies based on gender. The negative effect of −0.449 (SE = 0.193) suggests that, for this effect between views on the importance of EI for the VET sector and empathy difficulty, being male is associated with lower empathy difficulty than being female.

The fact that the c path is significantly moderated (t = −2.19, p = 0.032) indicates that the effect of occupation on empathy difficulty varies based on gender. The negative effect (−0.783 (SE = 0.358)) suggests that, for this effect between occupation and empathy difficulty, being male is associated with lower empathy difficulty than being female.