Approaching work as a calling is associated with numerous positive outcomes; yet, we know little about how callings develop. Grounded in social cognitive career and meaning-making theory, this study explores whether leaders who are called to their roles foster the development of calling in their employees.
Using a three-wave longitudinal design with a sample of 285 US employees, we examined whether perceiving a leader as called – and trusting that leader – predicts growth in employees’ sense of calling over time.
Results indicate that perceiving calling in a leader fosters the development of calling when the leader is trusted and lowers calling in employees when the leader is not trusted. We found evidence that this process is incremental and slow compared to other modeling processes, requiring four to eight months to be detected.
This study contributes to the literature by identifying leaders as potential socialization agents in the development of calling. The practical and theoretical implications of these results are discussed.
In today’s work environment, in which employees seek more than just a paycheck, the role of leaders in shaping meaningful careers and preventing disengagement and “quiet quitting” has become increasingly important (Gallup, 2023). Work becomes especially meaningful when it is perceived as calling. Over the past decade, research has highlighted many benefits of experiencing work as a calling – ranging from greater well-being (e.g. Dobrow et al., 2023; Duffy et al., 2017) to lower turnover intention (Cardador et al., 2011), to the point that some scholars describe a calling as the ultimate subjective experience of career success (Hall and Chandler, 2005). However, the mechanisms by which employees develop a sense of calling remain unclear.
A calling reflects multiple facets of an individual’s relationship to a specific life role and has been defined in several ways (Dik and Shimizu, 2019; Schabram et al., 2023). We define calling as the perception of a passion and transcendent force toward involvement in a life role, which pervades thoughts, provides a sense of purpose and identity, motivates sacrifices and includes a prosocial contribution (Vianello et al., 2018; Gerdel et al., 2025b). This definition entails both the modern and neoclassical approaches to calling. The literature has made notable progress in identifying individual predictors of calling, such as vocational clarity or job satisfaction (Duffy et al., 2014a, b). However, in organizational contexts, the role of leaders as potential catalysts of employees’ calling remains underexplored. Leaders shape employees’ experiences and perceptions through support, role modeling and their own sense of calling, which may profoundly influence those they lead (Xie et al., 2019).
Further, research has shown that engagement in learning processes and social support are two environmental factors that foster calling (Dalla Rosa et al., 2019). Moreover, mentors can influence the development of calling in students either by embodying calling themselves (Dalla Rosa et al., 2018) or through a high-quality mentoring relationship (Ensher and Ehrhardt, 2022). Together, these findings highlight that influential figures can play a critical role in cultivating a sense of calling – yet the specific mechanisms through which this occurs in organizations remain insufficiently understood.
Despite these insights, an important gap persists regarding the boundary conditions that shape when and how leaders’ calling or leader behaviors translate into employees’ calling. On the one hand, Xie and colleagues (2019) demonstrated that supervisor status can explain the relation between leaders’ calling and employees’ calling, suggesting that trickle-down effects may depend on hierarchical position. On the other hand, the authors explicitly call for longitudinal research to unpack the dynamics of how calling orientations may flow from leaders to followers over time. Thus, it remains unclear under which conditions leaders effectively transmit a sense of calling to their employees – and when such effects may fail to materialize.
Building on this rationale, this study investigates how employees perceive their leaders’ calling and how this perception influences the development of their own sense of calling. Specifically, we examine how leader–follower calling transmission occurs and whether trust in leadership moderates this relationship.
This study makes several theoretical contributions. First, it extends existing research on leader–follower calling transmission by integrating it with social cognitive career theory (SCCT) to explain how social learning and contextual factors shape the development of employees’ calling. Through this integration, the study reframes calling development as a socially embedded process, highlighting the mechanisms by which leaders’ demonstrated sense of calling can influence followers’ perceptions. Second, it identifies trust in leaders as a critical boundary condition, showing that relational quality determines the extent to which leader calling is internalized by employees. Together, these contributions enrich theoretical understanding of how calling develops within leader–follower relationships and offer practical insights for organizations aiming to foster meaningful leadership.
The role of leadership in the development of career calling
Social cognitive career theory (SCCT; Lent et al., 1994) integrates core ideas from social cognitive theory (SCT; Bandura, 1986) and applies them specifically to the domain of career development. SCCT is specifically interested in how people develop career interests, goals and self-efficacy in particular career areas. The theory draws from SCT by proposing that these constructs grow through career-relevant learning experiences – such as personal mastery, vicarious learning, social persuasion and affective states – which shape perceptions of the link between one’s capabilities, effort and anticipated outcomes in pursuing meaningful careers (Lent and Brown, 2013; Duffy et al., 2011). Within this framework, a leader’s sense of calling can represent a social learning experience for employees, one that may promote a sense of efficacy and interest in one’s own career. Observing a leader who demonstrates passion, purpose and perseverance provides employees with opportunities for vicarious learning, enabling them to internalize both the behaviors and the affective significance associated with pursuing personally meaningful work (Xie et al., 2019). By demonstrating observable and mimicable behaviors, called leaders create environments that support career development and personal growth, reinforcing SCCT’s assertion that learning experiences are shaped by social and contextual factors (Ni et al., 2024).
SCCT also emphasizes the role of contextual signals in shaping individuals’ internal frameworks. When employees observe a leader who is guided by a sense of purpose and commitment to a cause greater than the self, they are exposed to a living example of someone whose work is a calling. This can support followers in constructing their own narrative of calling (Bloom et al., 2021). For instance, a leader who prioritizes the broader mission over personal needs will significantly influence meaning-making processes in employees (Rosso et al., 2010).
Accordingly, individuals may develop and sustain career-related attitudes in part through exposure to meaningful experiences within their environment. Leaders who are perceived as experiencing a strong sense of calling may act as powerful contextual influences, shaping the work environment in ways that enable employees to make sense of their own work purpose. In this perspective, the leader’s calling becomes a contextual affordance – a socially embedded signal that helps employees interpret their work as calling (Park, 2010). Hence, we hypothesize that:
Perceived leader’s calling increases employees’ sense of calling.
The moderating role of trust
Although there are almost as many definitions of leadership as there are scholars who have attempted to define the concept (Stogdill, 1974), most include a process of influence (Yukl, 1989) that implies a change in employees’ behaviors. Given that trust mediates much of the variance across a wide range of leadership styles (Legood et al., 2021), it is one of the most important factors associated with positive employee outcomes. Since developing calling requires profound changes in personal identity and in the purpose and meaning of one’s job (Bloom et al., 2021), it seems unlikely that employees will emulate leaders they do not trust, because it is unlikely that employees will admire leaders they do not trust.
Trust is a complex psychological construct that is defined in many ways. Some authors describe it as a stable individual trait, while others depict it as either an emergent state or a process (Burke et al., 2007). In this study, we conceptualize trust as a dynamic, context-dependent attitude that can be expressed as an affective, motivational or cognitive state (Marks et al., 2001). Affective trust refers to an emotional connection and feelings toward the trustee, and it includes the belief that the trustee will act benevolently toward the trustor (Dirks and Ferrin, 2002). Given the relational and inspirational dynamics explored in this study, the focus here is on examining the affective component of trust.
Considering the intrinsically relational nature of affective trust, it is plausible to speculate that trust enhances employees’ sense of closeness and similarity to their leaders. For instance, trust is positively correlated with both leader–member exchange and satisfaction with the leader (Dirks and Ferrin, 2002), while trusting a leader promotes followership and open communication (Burke et al., 2007). Furthermore, affect-based trust fosters emotional openness with minimal concern about vulnerability. This social intimacy enables employees to develop shared values, perceptions and mental models, thereby enhancing the potential impact of professional role modeling (Chowdhury, 2005). In this regard, it has been shown that perceived competence, one of the key antecedents of trust, significantly enhances vicarious behavioral learning (Gioia and Manz, 1985).
Trust facilitates learning (Burke et al., 2007) because it increases knowledge sharing, dialog, collaborative inquiry and experimentation (Abrams et al., 2003; Levin et al., 2006; Mayer et al., 1995). In addition, it enhances the willingness to incorporate feedback from supervisors (Burke et al., 2007). SCCT emphasizes that learning experiences – particularly vicarious learning through observing others – are foundational to how individuals develop career-related beliefs and goals. Therefore, we expect trust to play a crucial role in enhancing the ability of called leaders to serve as effective inspirational figures for employees. In contrast, when employees distrust their supervisors, employees will be motivated to distance themselves from their supervisor. Accordingly, we hypothesize that.
Trust moderates the relationship between perceived leader’s calling and employees’ calling such that when trust is high, the positive longitudinal relation between the leader’s calling and the employees’ calling will be stronger after four months (H2a, lag 1 effect) and eight months (H2b, lag 2 effect).
We hypothesized both lag 1 and lag 2 moderation effects because trust is expected to increase over time (Dirks et al., 2022; Vanneste et al., 2014).
Method
Design
We employed a full longitudinal design (Little, 2013), collecting data at three distinct time points: May 2022 (T1), September 2022 (T2) and January 2023 (T3), with an interval of approximately four months between each point. This timeframe was chosen to balance the likelihood of capturing intraindividual changes in calling, which are necessary to investigate its dynamics, and the need to minimize participant attrition. Longer intervals in comparable longitudinal studies have often resulted in attrition rates exceeding 50% (e.g. Dalla Rosa et al., 2024; Praskova et al., 2014), which can compromise the validity of longitudinal inferences. At the same time, prior research has demonstrated that employees’ sense of calling can exhibit meaningful within-person changes over similar periods (e.g. Vianello et al., 2020; Dalla Rosa et al., 2024).
Participants and procedure
The study adhered to the ethical standards and guidelines outlined by the Institutional Review Board (IRB) at the University of Florida. Approval of the research was granted by the IRB under protocol number IRB202200465. Participants were recruited from the ResearchMatch participant pool, a database that facilitates data collection from adult participants across the United States (Harris et al., 2012). Specifically, Research Match is composed of over 150,000 volunteers living in the United States who are interested in taking part in health-related research. To date, thousands of studies have been conducted using Research Match as a data collection platform (including dozens in psychology), recognizing it as a valid tool of online data collection so long as researchers include validity check items to eliminate fraudulent responding (Faro et al., 2021; Pageau and Ling, 2025). For this study, an invitation containing a brief overview of the study was distributed to ResearchMatch members who met our participant criteria (aged 18 or older, N = 120,000). Those who expressed willingness to participate received an email link to the survey, which was hosted on Qualtrics. Informed consent was requested with the survey itself, indicating that they were willing to answer it. This took on average 10 min. Participants received no reimbursement.
A total of 358 volunteers from ResearchMatch participated in the first wave (T1). Of this initial group, 98 participants were excluded for failing attention check items (e.g. Please select item “strongly agree”) or failing quality check items (e.g. Please select “yes” if you took this survey seriously) (Aust et al., 2013; Kung et al., 2018). The final sample size for T1 was 260. Of the initial 358 participants, 202 replied to the second wave (T2). Of this group, 42 participants were excluded for failing attention and quality check items. The final sample size for T2 was 160. In the last wave (T3), of the 202 participants in the second wave, 110 responded to the survey. Due to failing attention and quality check items, seven participants were excluded, and the final sample size for T3 was 103. Analyses were performed on the complete dataset of 285 participants who participated in at least one wave and did not fail the attention and quality check items. This gave a partial nonresponse rate of 39% across all waves.
The participant pool was highly educated, with 42.7% holding a bachelor’s degree and 32% possessing a master’s degree. The majority were female (71.8%), while 26.2% identified as male, and 1.9% selected another gender category. At T1, participants had an average age of 44.66 years, with a mean tenure of 6.74 years at their organization and 3.40 years under their current supervisor. The sample was diverse, with individuals primarily working in administration (14.4%), research and development (10.6%), management (8.7%), customer service (7.2%), sales and service (5.8%) and Information Technology (IT) (5.8%). Working hours varied slightly over time, increasing at T2 (M = 39.53, standard deviation (SD) = 10.59) compared to T1 (M = 37.62, SD = 11.58) and T3 (M = 36.84, SD = 11.47).
Measures
The survey was conducted in English, and all responses were gathered using a Likert scale (1 = strongly disagree, 5 = strongly agree). Participants also responded to additional questions regarding their leader’s performance and their relationship with the leader; however, these aspects are not the focus of the current study and have been reported in Gerdel et al. (2025b). The raw dataset and a codebook listing all variables are publicly available on the Open Science Framework: https://osf.io/568jx/?view_only=e75b3004d91d4a17a80a09257608366f.
Employee’s career calling
Employee’s career calling was measured with the short version of the Unified Multidimensional Calling Scale-7 (UMCS-7; Gerdel et al., 2022). The scale consists of seven items and each item represents one facet of calling: passion, prosociality, purpose, pervasiveness, sacrifice, transcendent summons and identity. Example items are: “I am passionate about my work”, “I believe that I have been called to pursue my current line of work and “My work helps me live out my life’s purpose”. In this study, the reliability of the compound score was good (αT1 = 0.85; αT2 = 0.81; αT3. = 0.85). We performed a confirmatory factor analysis (CFA) showing that a one-factor model had a good fit across observations according to the following criteria (Weston and Gore, 2006): comparative fit index (CFI) ≥ 0.95, root mean square error of approximation (RMSEA) ≤ 0.06 and standardized root mean square residual (SRMR) ≤ 0.08. The fit at T1 was χ2 (14) = 38.96, CFI = 0.96, RMSEA = 0.09, 95% confidence interval (CI) [0.06, 0.13], SRMR = 0.04; at T2 χ2 (14) = 16.47, CFI = 0.99, RMSEA = 0.04, 95% CI [0, 0.11], SRMR = 0.04; and at T3 χ2 (14) = 21.48, CFI = 0.96, RMSEA = 0.08, 95% CI [0, 0.15], SRMR = 0.05.
Perceived Leader’s calling
Employees assessed their leader’s level of career calling by responding to all items of the UMCS-7, changing the subject from “I” to “My supervisor” (Gerdel et al., 2022). Sample items illustrating the perceived leader’s calling include: “My supervisor is passionate about his/her work” and “My supervisor's work gives meaning to his/her life.” In this study, the reliability of the compound score was good (αT1 = 0.90; αT2 = 0.91; αT3 = 0.92), and a one-factor model had a good fit across observations. The fit at T1 was χ2 (14) = 68.36, CFI = 0.93, RMSEA = 0.15, 95% CI [0.11, 0.18], SRMR = 0.05; at T2: χ2 (14) = 57.35, CFI = 0.92, RMSEA = 0.16, 95% CI [0.12, 0.21], SRMR = 0.05; and at T3: χ2 (14) = 30.45, CFI = 0.96, RMSEA = 0.12, 95% CI [0.06, 0.18], SRMR = 0.04.
Affect-based trust
Employees answered five items on the Affect-Based Trust Scale (McAllister, 1995). Example items are: “If I shared my problems with this person, I know (s)he would respond constructively and caringly” or “My supervisor and I have a sharing relationship. We can both freely share our ideas, feelings, and hopes”. In this study, the reliability of the compound score was good (αT1 = 0.93; αT2 = 0.91; αT3 = 0.91), and a one-factor model had a good fit across observations, after correlating items 3 and 5, which have the longest wording (Item 3: “My supervisor and I would both feel a sense of loss if one of us was transferred and we could no longer work together”; Item 5: “I would have to say that my supervisor and I have both made considerable emotional investments in our working relationship”). The correlation between the two residuals is theoretically justified by the method variance they share compared to the other, shorter items. The fit at T1 was χ2 (4) = 13.26, CFI = 0.99, RMSEA = 0.11, 95% CI [0.05, 0.18], SRMR = 0.02; at T2: χ2 (4) = 20.57, CFI = 0.96, RMSEA = 0.18, 95% CI [0.10, 0.22], SRMR = 0.03; and at T3: χ2 (4) = 12.85, CFI = 0.97, RMSEA = 0.17, 95% CI [0.07, 0.28], SRMR = 0.04.
Results
Attrition analysis
To examine whether participants who dropped out differed from those who completed the study, independent-samples t-tests were conducted on baseline (T1) measures of trust, perceived leader’s calling and employee’s calling (Table 1). At Time 2, there were no significant differences between completers (n = 160) and dropouts (n = 125) on T1 trust, t(258) = 0.73, p = 0.466, 95% CI [–0.18, 0.39]; employee’s calling, t(258) = 0.60, p = 0.548, 95% CI [–0.15, 0.28]; or perceived leader’s calling, t(258) = 0.43, p = 0.665, 95% CI [–0.17, 0.27]. Similarly, at Time 3, there were no significant differences between completers (n = 103) and those who dropped out by that wave (n = 182) on T1 trust, t(258) = 1.22, p = 0.222, 95% CI [–0.11, 0.48]; employee’s calling, t(258) = 0.20, p = 0.838, 95% CI [–0.21, 0.25]; or perceived leader’s calling, t(258) = 1.68, p = 0.095, 95% CI [–0.03, 0.43]. Hence, in the remainder of the analysis, we will use full information maximum likelihood as an estimator for missing data, which works accurately if there is no strong selective attrition (Little, 2013).
Comparison of mean differences between participants who completed the study and those who dropped out for employee’s calling, trust and perceived leader’s calling
| Variable | Attrition status | M (SD) | t(df) | p | d |
|---|---|---|---|---|---|
| Employee’s calling | T1 | 3.76 (0.85) | – | – | – |
| Dropped out at T2 | 3.69 (0.92) | 0.6 (258) | 0.55 | 0.08 | |
| Dropped out at T3 | 3.72 (0.88) | 0.2 (258) | 0.84 | 0.03 | |
| Trust | T1 | 3.74 (1.16) | – | – | – |
| Dropped out at T2 | 3.63 (1.14) | 0.73 (258) | 0.47 | 0.09 | |
| Dropped out at T3 | 3.62 (1.16 | 1.22 (258) | 0.22 | 0.16 | |
| Perceived leader’s calling | T1 | 3.8 (0.84) | – | – | – |
| Dropped out at T2 | 3.76 (0.98) | 0.43 (258) | 0.67 | 0.05 | |
| Dropped out at T3 | 3.71 (0.98) | 1.68 (258) | 0.1 | 0.22 |
Note(s): Total number of participants for each wave: T1 = 260, T2 = 160, T3 = 103
Analytical approach
To test our hypotheses, we applied a cross-lagged panel model with latent variables. All models were estimated using the lavaan package in R (Rosseel, 2012; version 0.6–7), and full-information maximum likelihood estimation was used for all models to account for missing data. Prior to testing the model, longitudinal measurement invariance was ascertained. We found support for scalar invariance, which is necessary to interpret differences among regression weights (Mackinnon et al., 2022). Results are reported in Table 2. Details about this analysis can be found in the online web supplement (https://osf.io/zat7n/?view_only=d402e22d122f4d83bffbac5b0a2e899f).
Measurement invariance tests for employee’s calling, perceived leader’s calling and trust
| χ2 | df | χ2/df | CFI | RMSEA | 95% CI | SRMR | Δ χ2 | Δdf | p | Δ CFI | ΔRMSEA | ΔSRMR | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LL | UL | |||||||||||||
| Employee’s calling | ||||||||||||||
| Configural | 286.76 | 165 | 1.74 | 0.93 | 0.05 | 0.04 | 0.06 | 0.07 | ||||||
| Metric | 300.24 | 177 | 1.69 | 0.93 | 0.05 | 0.04 | 0.06 | 0.08 | 13.48 | 12 | 0.34 | 0 | 0 | −0.01 |
| Scalar | 314.57 | 189 | 1.66 | 0.93 | 0.05 | 0.04 | 0.06 | 0.08 | 14.33 | 12 | 0.28 | 0 | 0 | 0 |
| Perceived leader’s calling | ||||||||||||||
| Configural | 345.69 | 165 | 2.1 | 0.93 | 0.06 | 0.05 | 0.07 | 0.06 | ||||||
| Metric | 352.68 | 177 | 1.99 | 0.93 | 0.06 | 0.05 | 0.07 | 0.07 | 6.99 | 12 | 0.86 | 0 | 0 | −0.01 |
| Scalar | 372.53 | 189 | 1.97 | 0.93 | 0.06 | 0.05 | 0.07 | 0.07 | 19.85 | 12 | 0.07 | 0 | 0 | −0.01 |
| Trust | ||||||||||||||
| Configural | 230.11 | 72 | 3.2 | 0.93 | 0.09 | 0.08 | 0.10 | 0.06 | ||||||
| Metric | 236.16 | 80 | 2.95 | 0.93 | 0.08 | 0.07 | 0.10 | 0.06 | 6.49 | 8 | 0.64 | 0 | −0.01 | 0 |
| Scalar | 243.32 | 86 | 2.83 | 0.93 | 0.08 | 0.07 | 0.09 | 0.06 | 7.16 | 6 | 0.31 | 0 | 0 | 0 |
Note(s): N = 280. Thresholds for accepting non-invariance: ≤ΔCFI 0.010, supplemented by ≥ ΔRMSEA 0.015 or ≥ ΔSRMR 0.030. CFI = Comparative fit index; RMSEA = root-mean-square error of approximation; SRMR = standardized root-mean-square residual
We used item parceling to build indicators in the models (Little et al., 2002). Parcels were created using the item-to-construct balance approach, in which items were ordered by their average factor loadings across time from an initial single-factor CFA and then paired such that high-loading and low-loading items were combined within the same parcel (Little et al., 2002). Each construct at each time point was therefore represented by two or three parcels, which served as indicators of the corresponding latent variables in all subsequent structural models. Model fit was assessed using the χ2, the CFI, the RMSEA and unbiased SRMR (uSRMR). The traditional SRMR statistic is known to be downwardly biased in models with latent variables, complex structures, or non-normally distributed indicators (Ximénez et al., 2022). The uSRMR corrects for this tendency by adjusting the discrepancy between observed and model-implied correlations, thereby providing a more accurate and conservative estimate of global model fit. CFI values between 0.95 and 0.97 suggest a good fit, whereas values > 0.97 suggest an excellent fit. RMSEA values between 0.05 and 0.08 suggest good fit, and values < 0.05 suggest excellent fit (Schermelleh-Engel et al., 2003). uSRMR values < 0.1 suggest adequate fit, whereas values < 0.5 suggest close fit (Ximénez et al., 2022).
To test for H1, we first specified and estimated a baseline model for the longitudinal effects of a leader’s calling on trust and of trust on employees’ calling, as well as model-fit information prior to the inclusion of moderators. This model included autoregressive paths for perceived leader’s calling, trust and employees’ calling (accounting for employee’s prior standings on each construct) and reciprocal effects from perceived leader’s calling, trust and employee’s calling, four (lag 1) and eight (lag 2) months later. Following the guidelines proposed by Orth and colleagues (2024), we interpret cross-lagged effects as small (β = 0.03), medium (β = 0.07) and large (β = 0.12) and removed all cross-lagged effects that were nonsignificant and smaller than 0.07 from subsequent specifications.
After specifying a well-fitting baseline model, we test for longitudinal moderation effects (H2a), by adding the latent interaction term using Marsh’s product-indicator approach (Marsh et al., 2004). In this approach, each item of the first latent construct is multiplied with each item of the second latent construct to form a set of product indicators, which serve as manifest indicators of the latent interaction factor. These product indicators are then included in the model with appropriate constraints to ensure model identification. Thus, employees’ calling was regressed on trust, perceived leader’s calling and the latent interaction between trust and perceived leader’s calling for both lag 1 and lag 2 effects.
Descriptive statistics and correlations
Table 3 reports the means, standard deviations, Cronbach’s alpha and correlations between the studied variables at T1, T2 and T3. All variables were correlated in the expected direction. Correlations between perceived leader’s calling and employees’ calling ranged from r = 0.20 to r = 0.43 and between perceived leader’s calling and trust ranged from r = 0.33 to r = 0.56.
Means, standard deviations, correlations and Cronbach’s alpha between study variables at three different time points
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Employees’ calling T1 | 3.73 | 0.89 | 0.85 | ||||||||
| 2. Employees’ calling T2 | 3.7 | 0.81 | 0.66** | 0.81 | |||||||
| 3. Employees’ calling T3 | 3.68 | 0.87 | 0.74** | 0.71** | 0.85 | ||||||
| 4. Trust T1 | 3.69 | 1.15 | 0.20** | 0.13 | 0.10 | 0.93 | |||||
| 5. Trust T2 | 3.76 | 1.06 | 0.13 | 0.19* | 0.10 | 0.63** | 0.91 | ||||
| 6. Trust T3 | 3.77 | 1.06 | 0.13 | 0.07 | 0.21* | 0.74** | 0.68** | 0.91 | |||
| 7. Perceived leader’s calling T1 | 3.78 | 0.91 | 0.28** | 0.21* | 0.23* | 0.53** | 0.32** | 0.48** | 0.90 | ||
| 8. Perceived leader’s calling T2 | 3.86 | 0.88 | 0.21** | 0.21** | 0.29** | 0.33** | 0.49** | 0.44** | 0.58** | 0.91 | |
| 9. Perceived leader’s Calling T3 | 3.89 | 0.90 | 0.30** | 0.20* | 0.43** | 0.42** | 0.47** | 0.56** | 0.67** | 0.67** | 0.92 |
Note(s): Cronbach’s alpha values are displayed in the main diagonal. T1 = time 1; T2 = time 2; T3 = time 3
**p < 0.01, *p < 0.05
The relation between perceived Leader’s calling and employees’ calling
The baseline model (including all cross-lagged paths) showed a good fit to the data (χ2(217) = 439.021, CFI = 0.93, RMSEA = 0.09, uSRMR = 0.00, 95% CI [−0.05; 0.05]). The paths from leader’s calling to employees’ calling are nonsignificant at lag 1 (β = 0.04, p = 0.55) and at the threshold of statistical significance from T2 to T3 (β = 0.15, p = 0.05). The direct effect is weak and potentially unstable and may be conditional on unknown moderators. Accordingly, H1 is only partially supported.
We then optimized model parsimony, following established Structural Equation Model (SEM) guidelines (Little, 2013; Kline, 2023; Byrne, 2012). Regression paths that were both nonsignificant and small in magnitude (|β| < 0.07) were constrained to zero to enhance model parsimony and avoid unnecessary model complexity. The final baseline model is an excellent fit to the data (χ2(224) = 436.972, CFI = 0.93, RMSEA = 0.08, uSRMR = 0.00, 95% CI [−0.050.05]). In this model, we observe non-trivial and non-significant longitudinal effects of employees’ calling on leaders’ calling across both time lags (T1–T2: β = 0.13, p = 0.13; T2–T3: β = 0.10, p = 0.34) and of employees’ calling at T1 on trust at T2 (β = 0.11, p = 0.21). We also observed a strong longitudinal effect of trust at T2 on leaders’ calling at T3 (β = 0.23, p = 0.04). Trust did not mediate the effect of followers’ calling on leaders’ calling (indirect effect = 0.02, 95% CI [−0.03, 0.11]).
Moderation model
The main effect of perceived leader’s calling at T2 on employees’ calling T3 increases slightly from β = 0.11 (p = 0.11) to β = 0.12 (p = 0.09) after we incorporated lag 1 moderation effects using Marsh’s product-indicator approach. The interaction between trust and perceived leader’s calling on employees’ calling was not significant at lag 1 (T1-T2: β = −0.06, p = 0.23; T2-T3: β = 0.07, p = 0.33). Thus, Hypothesis 2a was not confirmed.
To test lag 2 effects (H2b), we estimated a model in which the latent interaction term predicted employees’ calling at T3 only (Figure 1). The cross-lagged effects were similar in size and nonsignificant. The effect of a leader’s calling at T2 on employees’ calling T3 was β = 0.14 (p = 0.07). The latent interaction between trust and perceived leader’s calling at T1 on employees’ calling at T3 was positive and at the threshold of statistical significance (β = 0.15, p = 0.05), confirming H2b.
The diagram shows three groups of constructs labeled “Leader Calling T 1”, “Trust T 1”, and “Employee Calling T 1” on the left, “Leader Calling T 2”, “Trust T 2”, and “Employee Calling T 2” in the center, and “Leader Calling T 3”, “Trust T 3”, and “Employee Calling T 3” on the right. Each construct is represented by an oval connected to rectangular indicator boxes by arrows. On the left side, “Leader Calling T 1” connects to three rectangles labeled “L C 1”, “L C 2”, and “L C 3” with path values “0.86”, “0.82”, and “0.86”. “Trust T 1” connects to “T R 1” and “T R 2” with path values “0.96” and “0.89”. “Employee Calling T 1” connects to “E C 1”, “E C 2”, and “E C 3” with path values “0.84”, “0.75”, and “0.84”. Below these constructs appears another oval labeled “L C x T R T 1” connected to rectangles labeled “L C x T R 1”, “L C x T R 2”, “L C x T R 3”, “L C x T R 4”, “L C x T R 5”, and “L C x T R 6” with values “0.92”, “0.91”, “0.85”, “0.84”, “0.88”, and “0.88”. Arrows extend from the left constructs to the center constructs with path values “0.62” from “Leader Calling T 1” to “Leader Calling T 2”, “0.66” from “Trust T 1” to “Trust T 2”, “0.83” from “Employee Calling T 1” to “Employee Calling T 2”, and “0.14” from “Employee Calling T 1” to “Leader Calling T 2” and another “0.11” from “Employee Calling T 1” to “Trust T 2”. In the center section, “Leader Calling T 2” connects to “L C 1”, “L C 2”, and “L C 3” with values “0.89”, “0.84”, and “0.91”. “Trust T 2” connects to “T R 1” and “T R 2” with values “0.90” and “0.88”. “Employee Calling T 2” connects to “E C 1”, “E C 2”, and “E C 3” with values “0.82”, “.83”, and “0.84”. Arrows extend from the center constructs to the right constructs with path values “0.57” from “Leader Calling T 2” to “Leader Calling T 3”, “0.76” from “Trust T 2” to “Trust T 3”, and “0.82” from “Employee Calling T 2” to “Employee Calling T 3”. Additional arrows appear between the constructs with values “0.14”, “0.22”, and “0.11” from “Leader Calling T 2” to “Employee Calling T 3”, “Trust T 2” to “Leader Calling T 3”, and “Employee Calling T 2” to “Leader Calling T 3”, respectively. On the right side, “Leader Calling T 3” connects to rectangles labeled “L C 1”, “L C 2”, and “L C 3” with values “0.88”, “0.85”, and “0.90”. “Trust T 3” connects to “T R 1” and “T R 2” with values “0.97” and “0.89”. “Employee Calling T 3” connects to “E C 1”, “E C 2”, and “E C 3” with values “0.90”, “0.79”, and “0.87”. Curved arrows appear between the constructs within each time section. A long arrow labeled “0.15” extends from “L C x T R T 1” to “Employee Calling T 3”.Cross-lagged panel model with latent variables and lag 2 interaction. Note: Correlations between residuals of the same indicators at different measurement occasions have been specified but are omitted in the path diagram for clarity. Observed variables represent item parcels. Parcels were constructed using the item-to-construct balance approach, in which items were first ordered by their average factor loadings across time based on an initial single-factor CFA. Items were then paired so that higher-loading and lower-loading items were combined within each parcel (Little et al., 2002). The observed indicators of the latent interaction factor were derived using Marsh’s product-indicator approach. In this procedure, each item of the first latent construct is multiplied by each item of the second latent construct to generate a set of product indicators (Marsh et al., 2004). Model fit: χ²(373) = 1,329.383, CFI = 0.86, RMSEA = 0.11, 95% CI [0.10; 0.12], uSRMR = 0.00, 95% CI [−0.05; 0.05]. All estimates in the diagram are standardized. Values higher than β = 0.22 are significant (p < 0.05). Source(s): Authors’ own work
The diagram shows three groups of constructs labeled “Leader Calling T 1”, “Trust T 1”, and “Employee Calling T 1” on the left, “Leader Calling T 2”, “Trust T 2”, and “Employee Calling T 2” in the center, and “Leader Calling T 3”, “Trust T 3”, and “Employee Calling T 3” on the right. Each construct is represented by an oval connected to rectangular indicator boxes by arrows. On the left side, “Leader Calling T 1” connects to three rectangles labeled “L C 1”, “L C 2”, and “L C 3” with path values “0.86”, “0.82”, and “0.86”. “Trust T 1” connects to “T R 1” and “T R 2” with path values “0.96” and “0.89”. “Employee Calling T 1” connects to “E C 1”, “E C 2”, and “E C 3” with path values “0.84”, “0.75”, and “0.84”. Below these constructs appears another oval labeled “L C x T R T 1” connected to rectangles labeled “L C x T R 1”, “L C x T R 2”, “L C x T R 3”, “L C x T R 4”, “L C x T R 5”, and “L C x T R 6” with values “0.92”, “0.91”, “0.85”, “0.84”, “0.88”, and “0.88”. Arrows extend from the left constructs to the center constructs with path values “0.62” from “Leader Calling T 1” to “Leader Calling T 2”, “0.66” from “Trust T 1” to “Trust T 2”, “0.83” from “Employee Calling T 1” to “Employee Calling T 2”, and “0.14” from “Employee Calling T 1” to “Leader Calling T 2” and another “0.11” from “Employee Calling T 1” to “Trust T 2”. In the center section, “Leader Calling T 2” connects to “L C 1”, “L C 2”, and “L C 3” with values “0.89”, “0.84”, and “0.91”. “Trust T 2” connects to “T R 1” and “T R 2” with values “0.90” and “0.88”. “Employee Calling T 2” connects to “E C 1”, “E C 2”, and “E C 3” with values “0.82”, “.83”, and “0.84”. Arrows extend from the center constructs to the right constructs with path values “0.57” from “Leader Calling T 2” to “Leader Calling T 3”, “0.76” from “Trust T 2” to “Trust T 3”, and “0.82” from “Employee Calling T 2” to “Employee Calling T 3”. Additional arrows appear between the constructs with values “0.14”, “0.22”, and “0.11” from “Leader Calling T 2” to “Employee Calling T 3”, “Trust T 2” to “Leader Calling T 3”, and “Employee Calling T 2” to “Leader Calling T 3”, respectively. On the right side, “Leader Calling T 3” connects to rectangles labeled “L C 1”, “L C 2”, and “L C 3” with values “0.88”, “0.85”, and “0.90”. “Trust T 3” connects to “T R 1” and “T R 2” with values “0.97” and “0.89”. “Employee Calling T 3” connects to “E C 1”, “E C 2”, and “E C 3” with values “0.90”, “0.79”, and “0.87”. Curved arrows appear between the constructs within each time section. A long arrow labeled “0.15” extends from “L C x T R T 1” to “Employee Calling T 3”.Cross-lagged panel model with latent variables and lag 2 interaction. Note: Correlations between residuals of the same indicators at different measurement occasions have been specified but are omitted in the path diagram for clarity. Observed variables represent item parcels. Parcels were constructed using the item-to-construct balance approach, in which items were first ordered by their average factor loadings across time based on an initial single-factor CFA. Items were then paired so that higher-loading and lower-loading items were combined within each parcel (Little et al., 2002). The observed indicators of the latent interaction factor were derived using Marsh’s product-indicator approach. In this procedure, each item of the first latent construct is multiplied by each item of the second latent construct to generate a set of product indicators (Marsh et al., 2004). Model fit: χ²(373) = 1,329.383, CFI = 0.86, RMSEA = 0.11, 95% CI [0.10; 0.12], uSRMR = 0.00, 95% CI [−0.05; 0.05]. All estimates in the diagram are standardized. Values higher than β = 0.22 are significant (p < 0.05). Source(s): Authors’ own work
To further investigate the shape of the interaction effect, we conducted a multigroup analysis on subsets of participants who were either high (1 SD above the mean) or low (1 SD below the mean) in trust. We observed the effect of leader’s calling on employees’ calling to be β = 0.51 [0.33, 0.69] in the high trust group and β = 0.00 [−0.14, 0.15] in the low trust group. Figure 2 depicts the interactions at five different levels of trust.
The panel shows five side-by-side scatter plots. The vertical axis on the left is labeled “Employee’s Calling (T 3)” and ranges from negative 3 to 1 in increments of 1 unit. The horizontal axis at the bottom is labeled “Perceived Leader’s Calling (T 1)” and ranges from negative 3 to 1 in increments of 1 unit. A dashed horizontal line appears near the top of the panel, and another dashed horizontal line appears near the bottom. Each section contains scattered circular points and a slanted line with a shaded band and a diamond marker. Above the first section, the text reads “negative 1.5 S D (P T C L equals 10.18) b equals negative 0.13 95 percent C I equals [negative 0.3, 0.05]”. The line in this section extends from about negative 0.2 at negative 3 on the horizontal axis to about negative 0.7 at 1 on the horizontal axis. Above the second section, the text reads “negative 1 S D (P T C L equals 19.3) b equals 0 95 percent C I equals [negative 0.14, 0.15]”. The line in this section extends from about negative 0.4 at negative 3 on the horizontal axis to about negative 0.4 at 1 on the horizontal axis. Above the third section, the text reads “0 S D (P T C L equals 43.16) b equals 0.26 95 percent C I equals [0.13, 0.38]”. The line in this section extends from about negative 1.1 at negative 3 on the horizontal axis to about 0.2 at 1 on the horizontal axis. Above the fourth section, the text reads “1 S D (P T C L equals 84.56) b equals 0.51 95 percent C I equals [0.33, 0.69]”. The line in this section extends from about negative 1.6 at negative 3 on the horizontal axis to about 0.9 at 1 on the horizontal axis. Above the fifth section, the text reads “1.5 S D (P T C L equals 98.95) b equals 0.64 95 percent C I equals [0.42, 0.86]”. The line in this section extends from about negative 1.9 at negative 3 on the horizontal axis to about 1.1 at 1 on the horizontal axis. Note: All numerical data values are approximated.Moderating effect of trust on the relation between perceived leader’s calling and employees’ calling at five different levels. Note: Simple slopes are provided for levels of moderator at 1.5 SD and 1 SD below the mean, at the mean and at 1 SD and 1.5 SD above the mean. Each graphic shows the computed 95% confidence region (shaded area), the observed data (gray circles), the maximum and minimum values of the outcome (dashed horizontal lines) and the crossover point (diamond). The x-axis represents the full range of the focal predictor. CI = Confidence interval; PTCL = percentile (McCabe et al., 2018). Source(s): Authors’ own work
The panel shows five side-by-side scatter plots. The vertical axis on the left is labeled “Employee’s Calling (T 3)” and ranges from negative 3 to 1 in increments of 1 unit. The horizontal axis at the bottom is labeled “Perceived Leader’s Calling (T 1)” and ranges from negative 3 to 1 in increments of 1 unit. A dashed horizontal line appears near the top of the panel, and another dashed horizontal line appears near the bottom. Each section contains scattered circular points and a slanted line with a shaded band and a diamond marker. Above the first section, the text reads “negative 1.5 S D (P T C L equals 10.18) b equals negative 0.13 95 percent C I equals [negative 0.3, 0.05]”. The line in this section extends from about negative 0.2 at negative 3 on the horizontal axis to about negative 0.7 at 1 on the horizontal axis. Above the second section, the text reads “negative 1 S D (P T C L equals 19.3) b equals 0 95 percent C I equals [negative 0.14, 0.15]”. The line in this section extends from about negative 0.4 at negative 3 on the horizontal axis to about negative 0.4 at 1 on the horizontal axis. Above the third section, the text reads “0 S D (P T C L equals 43.16) b equals 0.26 95 percent C I equals [0.13, 0.38]”. The line in this section extends from about negative 1.1 at negative 3 on the horizontal axis to about 0.2 at 1 on the horizontal axis. Above the fourth section, the text reads “1 S D (P T C L equals 84.56) b equals 0.51 95 percent C I equals [0.33, 0.69]”. The line in this section extends from about negative 1.6 at negative 3 on the horizontal axis to about 0.9 at 1 on the horizontal axis. Above the fifth section, the text reads “1.5 S D (P T C L equals 98.95) b equals 0.64 95 percent C I equals [0.42, 0.86]”. The line in this section extends from about negative 1.9 at negative 3 on the horizontal axis to about 1.1 at 1 on the horizontal axis. Note: All numerical data values are approximated.Moderating effect of trust on the relation between perceived leader’s calling and employees’ calling at five different levels. Note: Simple slopes are provided for levels of moderator at 1.5 SD and 1 SD below the mean, at the mean and at 1 SD and 1.5 SD above the mean. Each graphic shows the computed 95% confidence region (shaded area), the observed data (gray circles), the maximum and minimum values of the outcome (dashed horizontal lines) and the crossover point (diamond). The x-axis represents the full range of the focal predictor. CI = Confidence interval; PTCL = percentile (McCabe et al., 2018). Source(s): Authors’ own work
Discussion
To address the lack of knowledge about how calling is developed in organizations, we conducted a longitudinal study to test a modeling process occurring between leaders and employees that fosters the development of calling. In line with SCCT, this study theorized that leaders function as contextual affordances by serving as role models who shape the social environment in ways that support the development of employees’ sense of calling. However, we did not find support for a direct effect of perceived leader’s calling on employees’ calling. Drawing on SCCT and meaning-making theory, we further predicted that trust would moderate the relation between perceived leader’s calling and employees’ calling. Our results support the notion that employees’ calling increases as a result of their leader’s calling when trust is high. When trust in the leader is low, employees’ calling decreases as a result of the leader’s calling.
Theoretical implications
This study offers several theoretical contributions to the calling literature by enhancing the understanding of how calling develops in workplace settings. It demonstrates that calling can emerge through relational processes, particularly within leader-follower interactions. Previous studies indicate that a supportive work environment fosters the development of a sense of calling among employees (e.g. Dalla Rosa et al., 2019). Supervisor-related factors – including attitude, personality, leadership style and behavior – have all been shown to significantly shape employees’ job attitudes and actions (e.g. Lyubykh et al., 2022), with individuals in higher positions within the hierarchy exerting notable influence on those in lower positions (Lu et al., 2018). However, this study did not find a direct, unconditioned, effect of leaders’ sense of calling on the development of employees’ calling.
One key social dynamic is trust, which appears to be a necessary condition that facilitates the transfer of calling-related beliefs and behaviors from leaders to employees. While prior research has acknowledged the importance of mentors in shaping calling (Dalla Rosa et al., 2018), our study expands this understanding by showing that a leader-follower relation that is based on trust can serve as a significant mechanism for calling development. This is in line with previous literature showing that a leader’s calling increases employees’ calling through support and a high-quality relationship (Gerdel et al., 2025a).
When employees trust their leader, they are more likely to view the leader’s expressions of purpose, mission and commitment as authentic and worth emulating. In this way, trust becomes a boundary condition for employees to construct meaning from the leader’s behaviors and narratives. This interpretation aligns with evidence showing that relational factors such as communication (r = 0.37), interaction frequency (r = 0.28), shared mental models (r = 0.31) and team cohesion (r = 0.44) are positively related to perceptions of supervisor trustworthiness (Hancock et al., 2023). Trust creates the psychological safety and openness necessary for employees to view the leader’s purpose-driven behaviors as authentic and meaningful, thereby facilitating the sensemaking processes through which calling can be transmitted. Future studies should further elaborate on the influence of contextual factors on the relation between leader’s and employees’ calling.
In contrast, when trust is low, employees may discount or even misinterpret the leader’s expressions of calling, such that leaders’ behaviors may be perceived as inauthentic or irrelevant to employees’ own experience. As a result, highly called leaders negatively impact their employees’ calling. This counterproductive effect of leader’s calling adds to the literature on the dark side of calling, extending the possible negative outcomes that some people may experience beyond workaholism and exploitation (Takagi et al., 2025). For some people, in some conditions, calling can not only be detrimental for the individual but also for others under their influence. Analogously to results by Gazica and Spector (2015), who showed that unanswered callings lead to poorer physical and psychological health compared to not having a calling at all. In this study, we found that when a leader is not trusted, it is better for them to have low-to-moderate levels of calling, since the combination of low trust and high calling decreases employees’ calling, probably as a protective mechanism in defense of employees’ identity and integrity.
From a SCCT perspective, trust thus acts as a proximal contextual support that enables employees to engage with, process and internalize the leader’s calling. Trust creates a psychologically safe and receptive environment in which employees feel comfortable drawing meaning from the leader’s actions and narratives (Maximo et al., 2019). This might be due to the fact that the most common antecedents of trust are perceived competence, integrity and benevolence (Schoorman et al., 2007). We are therefore likely to trust leaders when we assume them to be competent and benevolent and when their actions align with both the organization’s values and their own stated principles (Harshman and Harshman, 1999). Only under these conditions do leaders’ calling become a catalyst for employees’ own calling development.
We also observed that the longitudinal moderation effect was weak and unreliable after four months but stronger and reliable after eight months. There may be multiple interpretations of this effect. The first is that employees may require time to perceive their leader’s calling. Calling is not immediately visible, and employees may need repeated exposure to the leader’s behaviors and decisions before they can interpret them as expressions of deeper purpose. As employees observe leaders articulating meaningful goals over time, they form a clearer sense of their leader’s calling. This process aligns with sensemaking frameworks, which suggest that employees construct their understanding of their leader gradually (Kraft et al., 2018). Therefore, perceiving a leader’s calling likely requires a sustained accumulation of behavioral evidence before it can shape employees’ own sense of calling (Park, 2010).
A second possible mechanism in line with a slow change of calling after a called and trustful leader relates to the reflective process in which calling is constructed through activities and relations, such as re-evaluating personal values and goals, and acting on performance feedback, thus shaping a clearer purpose over time (Reed et al., 2022). Calling may develop through sensemaking job activities (Sturges et al., 2019), which involve lengthy processes.
A third possible explanation is that calling is a deeply intimate and self-defining construct. Variation in calling intensity, together with the presence of highly personal conditions that shape how, where and with whom individuals are willing to enact their callings, help explain why calling may sometimes shift rapidly following a personal insight, as individuals reassess whether particular activities or roles align with the strength and boundaries of their calling (Clinton et al., 2017; Robertson et al., 2025). At the same time, these features help explain why employees may be cautious in internalizing a leader’s calling through a modeling process: because callings vary in intensity and are constrained by individualized conditions, a leader’s calling may not readily translate into employees’ own experiences. This gradual adoption suggests that calling development may require sustained exposure to role models and continuous reinforcement through meaningful workplace interactions and trust.
Practical implications
The findings from this study may offer some implications regarding organizational leadership. If organizations hire leaders who have a calling and then teach them to share meaningful job activities, it’s possible that across time this could spill over to other employees (Gerdel et al., 2025a). Leaders might benefit from formal training in the effective communication of their own sensemaking activities. For example, leaders can help employees see their prosocial impact or their purpose in their work. Workshops on reflective leadership, for instance, could teach leaders how to share meaningful aspects of their work and communicate the significance of their roles to other employees (Castelli, 2016). When leaders openly discuss what drives them and share insights into the meaningful aspects of their work, employees are likely to adopt similar reflective practices. Employees may come to recognize a sense of calling in their own work as they understand how leaders connect to theirs.
Leaders can also help employees recognize meaning in their work by providing regular feedback that emphasizes the purpose and impact of specific tasks. For example, during performance reviews, leaders might highlight how an employee’s contributions align with broader organizational goals or positively affect clients and colleagues. This type of feedback can help employees identify their strengths, which can in turn deepen their understanding of their calling (Gerdel et al., 2025b). It is crucial that employees work with leaders whom they perceive to be credible. Trust can be fostered throughout an organization when positive, supportive interactions occur with direct supervisors (Fulmer and Ostroff, 2017).
Limitations and future directions
Despite the theoretical and practical relevance of this study, some limitations should be noted. First, focusing exclusively on the affective component of trust may limit our understanding of the overall process, potentially overlooking the influence of other aspects within the composite construct of trust. Nevertheless, given that affective trust has been relatively understudied (Dirks and Ferrin, 2002), our focus on this aspect enhances the contribution of this article to the existing literature. Future researchers could measure trust in a more multifaceted manner, for example, by including the cognitive dimension of trust. Role modeling is a cognitive process, and analyzing the cognitive component of trust could help clarify the boundary conditions of trust as a moderating factor in the enhancement of employees’ calling.
Second, this study relied on a single method and source to assess calling and trust, potentially introducing common method bias (Podsakoff et al., 2003). Future studies are encouraged to utilize different sources of information (i.e. leaders, team members) within the same study. For example, it would be valuable to explore whether similar dynamics apply when managers perceive their employees as having a calling. For instance, do managers show greater support, mentorship, or engagement with employees they view as “called”? Additionally, dyadic studies can help clarify whether trust between leaders and employees develops reciprocally, shedding light on whether a leader’s trust in an employee strengthens the employee’s trust in return and potentially accelerates mutual trust building.
Finally, examining specific behaviors that foster employees’ perceptions of a leader’s calling and their own trust in that leader could yield practical insights. For example, behaviors such as open communication, transparent decision-making and consistent embodiment of organizational values may reinforce a leader’s credibility (Chng et al., 2019). Identifying these behaviors could help organizations create targeted training programs that enhance both leaders’ and employees’ sense of calling and mutual trust.
