Leadership behavior, employee engagement, and retention: lessons from the COVID-19 pandemic in the hospitality industry

1 Introduction

1.1 Employee engagement and organizational performance

Employee engagement has been a persistent priority for organizational leaders and scholars for more than three decades. Defined as “a positive, fulfilling, work-related state of mind that is characterized by vigor, dedication, and absorption” (Schaufeli et al., 2002b, p. 74), employee engagement is consistently linked to outcomes that matter directly to organizational performance: productivity (Gallup, 2025; Tensay and Singh, 2020), profitability (Xanthopoulou et al., 2009), customer satisfaction (Schneider et al., 2018), safety performance (Nahrgang et al., 2011), and employee wellbeing (Shuck and Reio, 2014). Organizations worldwide have invested substantially in engagement initiatives, and for good reason, as the business case is well established. Yet despite decades of attention and investment, only 21% of employees globally report being engaged at work (Gallup, 2025), meaning nearly four in five workers are either not engaged or actively disengaged. That persistent gap between aspiration and reality suggests that organizations have not yet identified, or have not consistently implemented, the leadership conditions most likely to produce and sustain engagement. The COVID-19 pandemic made this challenge consequential in ways that prior organizational research had not fully anticipated.

Among the many potential drivers of employee engagement, including job demands, job resources, organizational climate, and individual characteristics, leadership behavior is one of the most consistently identified organizational influences. A meta-analysis across 7,939 business units found significant relationships between manager behavior and business-unit-level engagement outcomes including productivity, profitability, and turnover (Harter et al., 2002), with more recent estimates indicating that managers account for approximately 70% of variance in team-level engagement (Gallup, 2025). Unlike structural features of work, leadership behavior is directly modifiable through selection, training, and development, making it a practical target for organizational intervention. This study therefore focuses on leadership behavior as the primary antecedent of engagement and, in turn, retention, given its potential as an actionable organizational lever.

1.2 The COVID-19 pandemic and the retention crisis in hospitality

Beginning in early 2020, the pandemic created widespread simultaneous disruptions to workforce structures, supervisory relationships, and the nature of work across every sector of the global economy. Remote work, furloughs, layoffs, and rapidly changing health and safety protocols transformed the employment relationship in ways that existing leadership frameworks had not been designed to address. Prior to the pandemic, leadership models emphasized transformational, change-oriented, and visionary approaches as primary drivers of employee engagement and performance (Derue et al., 2011; Yukl et al., 2019). These frameworks were developed and validated in conditions of relative operational stability, and their applicability to crisis environments characterized by safety threats, operational volatility, and workforce instability had not been empirically established (Kniffin et al., 2021). The pandemic thus created an opportunity to examine whether existing leadership prescriptions would hold under conditions of sustained disruption. Research conducted during the pandemic found that effective crisis leadership required meaningful adaptation, including increased operational presence, direct communication, and task structuring, that differed from pre-pandemic approaches emphasizing vision and inspiration (Dale Oen et al., 2022).

The result was not only an engagement challenge but also a retention challenge. As restrictions eased and labor markets reopened, employees who had disengaged during the disruption may have also left their organizations in high numbers. This pattern of voluntary departures, documented in the literature as the Great Resignation, refers to the sustained increase in voluntary quit rates observed across industries beginning in 2021, presumably as workers reassessed their employment relationships in the wake of pandemic disruption (Liu-Lastres et al., 2023). In the hospitality and tourism industry specifically, the pandemic may have accelerated pre-existing workforce vulnerabilities, with the core driver being a sustained deterioration in employee experience that had been developing for years before 2020 (Liu-Lastres et al., 2023).

Among the industries affected by pandemic disruption, hospitality experienced particularly severe employment losses and persistent retention challenges. Workers in leisure and hospitality experienced an unemployment peak of 39.3% in April 2020, the highest of any industry sector (U.S. Bureau of Labor Statistics, 2020). While employment began recovering in early 2021 as pandemic restrictions eased, the recovery was turbulent and incomplete. A review of Bureau of Labor Statistics data from 2001 to 2022 found that hospitality turnover consistently ranks higher than other economic sectors and is disproportionately sensitive to macroeconomic disruptions, meaning that crisis events amplify the industry's retention challenges to a degree not observed in other sectors (Dogru et al., 2023). Organizations in the sector faced a practical question with significant strategic stakes: what can leaders do, specifically and concretely, to re-engage their employees and keep them?

1.3 Leadership behavior as a driver of engagement and retention

Research across industries and contexts has established that multiple categories of leadership behavior, including task-oriented, relations-oriented, and change-oriented, contribute positively to engagement (Popli and Rizvi, 2016; Wefald et al., 2011; Xu and Thomas, 2011). Prior research testing all three leadership behavior types simultaneously in a quantitative study of 117 employed professionals across 13 countries found that task, relations, and change-oriented leadership all positively predicted engagement, with change-oriented leadership, characterized by envisioning, innovation encouragement, and collective learning, emerging as the strongest predictor (Li et al., 2021). That finding was consistent with a broader literature linking transformational and change-oriented leadership to engagement in pre-pandemic organizational settings (Derue et al., 2011).

However, the pre-pandemic emphasis on change-oriented and transformational leadership as the primary engagement drivers may not translate intact into the crisis and recovery context of the hospitality industry during COVID-19. Consideration-based leadership behaviors (i.e., supportive, relational behaviors focused on employee wellbeing, trust, and interpersonal concern) served as a significant job resource for sustaining employee engagement during COVID-19, but effects were meaningfully differentiated by work context, with employees in higher-disruption settings responding differently than those in stable environments (Sedefoglu et al., 2024). This finding supports the argument that crisis context shapes which leadership behaviors matter most for engagement, and raises the question of whether the dominance of change-oriented leadership identified in pre-pandemic general samples holds in the acute operational crisis of the hospitality industry.

Leadership behavior has also been linked to whether employees remain with their organizations. Supervisory quality predicts intention to stay among IT professionals (Eom, 2015) and health care workers (Schopman et al., 2017), and leadership satisfaction has been associated with engagement, loyalty, and retention among hospitality employees specifically (Book et al., 2019). More recent work continues to report associations between leadership and retention across sectors, including among nurses (AbdELhay et al., 2025) and knowledge workers early in their careers (Zhang and Wang, 2025). The mechanisms proposed vary, but the recurring account is that supervisors shape the conditions employees weigh when deciding whether to stay, including how clearly work is structured, how their contributions are recognized, and how much support they receive in doing the job.

What is less settled is how leadership reaches retention. Retention is a behavioral intention rather than an attitude, and supervisory behavior is unlikely to influence it directly in the way it might influence satisfaction or perceived support. This suggests an intervening mechanism, and engagement is a plausible candidate given its established associations with both leadership and turnover intention. Whether engagement in fact carries the leadership-retention relationship, and whether it does so equally across settings, has received less attention than either link considered separately.

The hospitality industry represents an important and structurally distinct context for examining these questions. In hospitality, supervisors translate organizational expectations into daily operational practices and have a direct influence on employee engagement and wellbeing through continuous frontline contact (Abdou and Albohnayh, 2026). Unlike many non-hospitality sectors where workers have access to remote work arrangements, flexible scheduling, and professional identities that buffer engagement independently of supervisory behavior, hospitality frontline workers are physically present, operationally dependent on their supervisors for task guidance and resource allocation, and more directly exposed to the effects of supervisory quality on their daily work experience (Bakker and Demerouti, 2007; Dogru et al., 2023). This structural difference suggests that the relationship between leadership behavior and engagement, and by extension retention, may be stronger in hospitality than in other industries, and that the specific type of leadership behavior that matters most may differ as well.

When operational survival is the immediate organizational priority, when protocols change daily, staffing is volatile, and the physical safety of employees and customers is uncertain, the supervisory behaviors that matter most may not be visionary or inspirational ones. They may be the ones that bring order, clarity, and stability to daily work: planning and organizing, clarifying roles and expectations, monitoring performance, and solving operational problems. Task-oriented leadership may therefore assume primacy in precisely the conditions that characterized the hospitality pandemic experience. To date, no study has directly examined whether the relative importance of these three leadership behavior categories shifts as a function of crisis context and industry demands, a gap this study addresses.

1.4 Gaps in the literature

While the positive effects of leadership on engagement are well established, and the positive relationship between engagement and retention is similarly well replicated (Chanana and Sangeeta, 2021; Eltaybani et al., 2018), several important gaps remain in the literature. Research on leadership behavior and employee outcomes in hospitality has grown substantially but has focused predominantly on transformational and servant leadership styles, leaving task-oriented, relations-oriented, and change-oriented behavioral approaches, as defined by the Yukl et al. (2002) taxonomy, underexamined in this context (Elkhwesky et al., 2022; Pittaway et al., 1998). The conditions under which different leadership behavior types are most effective, and whether these conditions are moderated by industry context or crisis circumstances, remain important research opportunities.

Our previous research made an important contribution by demonstrating that task, relations, and change-oriented leadership behaviors all positively predict engagement in a general, multi-industry sample of 117 employees across 13 countries, with change-oriented leadership emerging as the strongest predictor (Li et al., 2021). However, that study did not include employee retention as an outcome, did not focus on a specific industry context, and did not examine the mediating role of engagement in linking leadership to retention. Whether the findings generalize to the hospitality industry, whether the dominant leadership behavior shifts under crisis conditions, and whether leadership affects retention indirectly through engagement remained unanswered.

The specific pathway through which leadership behavior translates into retention has received limited empirical attention more broadly. A systematic review of 79 leadership studies in the hospitality industry spanning 2008 to 2020 explicitly identified the leadership-to-employee outcome pathway as underspecified in the literature, noting that several leadership styles have not been rigorously examined in relation to retention and engagement outcomes (Elkhwesky et al., 2022). It is plausible and theoretically coherent that leadership behavior affects retention not only directly but indirectly, by first shaping the engagement state of employees, which in turn shapes their intention to remain. Employee engagement, in this framework, is the psychological mechanism through which supervisory behavior becomes organizational stability (Albrecht et al., 2023). Testing this mediated pathway, with appropriate attention to industry context and employee characteristics, represents a meaningful contribution to both the leadership and retention literatures.

1.5 Overview of the present study

This study addresses these gaps through three specific extensions of our previous research (Li et al., 2021). First, it incorporates employee retention as a downstream outcome, testing a mediation model in which leadership behavior affects retention both directly and indirectly through employee engagement. Second, it examines the hospitality industry as the primary context, allowing a direct test of whether the dominant leadership behavior driving engagement shifts from change-oriented, as found in the general sample, to task-oriented in a crisis and recovery context, and whether the mediation of leadership on retention through engagement is stronger in hospitality than in other industries. Third, it examines two boundary conditions, industry type and employee age, as moderators of the direct and indirect effects, providing a more complete account of when and for whom these relationships are strongest. The inclusion of age as a moderator is theoretically motivated by evidence that career stage, reciprocity norms, and supervisory expectations differ across age groups in ways that shape both the direct and indirect pathways from leadership to retention (Cappelli, 2014; James et al., 2011).

Hypotheses are tested using a two-step analytic approach. Leadership behavior is first examined as a composite construct reflecting the overall supervisory behavioral profile defined by the Yukl et al. (2002) leadership behavior taxonomy, providing the primary structural test of the leadership-engagement-retention model. The composite is then decomposed to identify which specific behavior type, task-oriented, relations-oriented, or change-oriented, drives the observed effects in the hospitality context. This approach directly addresses the theoretical question of whether task-oriented leadership behavior assumes primacy under crisis and operational conditions characteristic of the hospitality industry.

The practical stakes of this work are direct. Hospitality organizations rebuilding their workforces in the post-pandemic era need evidence-based guidance on which leadership behaviors to develop, which employee groups to prioritize, and which mechanisms to cultivate. This study offers that guidance by integrating lessons from the pandemic experience with a rigorous moderated mediation framework grounded in established theory and empirically tested against a diverse, multinational sample.

2 Theoretical framework and hypotheses

2.1 Leadership behavior model

The theoretical foundation for this study is the hierarchical taxonomy of leadership behaviors developed by Yukl et al. (2002) and Yukl et al. (2019), operationalized through the managerial practices survey (MPS). This framework classifies leadership behaviors into three primary meta-categories: task-oriented, relations-oriented, and change-oriented, each with distinct behavioral sub-categories.

Task-oriented behaviors focus on creating efficiencies in how people and resources are used to ensure operational reliability. Sub-categories include planning and organizing, problem-solving, clarifying roles, and monitoring operations. In the hospitality context, task-oriented behaviors are especially relevant because the industry depends on consistent service delivery, rapid operational adaptation, and real-time supervisory responsiveness (Pittaway et al., 1998).

Relations-oriented behaviors focus on building commitment and psychological safety within the workgroup through mutual trust, support, and cooperation. Sub-categories include supporting, developing, recognizing, and empowering team members. Relations-oriented leadership has been linked to engagement through satisfaction of employees' psychological needs for support, autonomy, and belonging (Albrecht and Andreetta, 2011; Xu and Thomas, 2011).

Change-oriented behaviors emphasize collective learning, innovation, and adaptability in response to external environmental demands. Prior research identified change-oriented leadership as the strongest engagement predictor in a general sample, consistent with pre-pandemic research linking transformational leadership to engagement (Derue et al., 2011; Li et al., 2021). However, the COVID-19 pandemic created a context of operational crisis in which immediate stability may have displaced aspirational change leadership as the primary supervisory need.

This study employs a two-step analytic approach consistent with the hierarchical structure of the Yukl et al. (2002) leadership behavior taxonomy. First, leadership behavior is tested as a composite construct, reflecting the overall supervisory behavioral profile defined by the combination of task, relations, and change-oriented behaviors, providing the primary structural test of the leadership-engagement-retention model. This composite construct is used in testing H1, H3, and H4. H2 tests the independent effect of employee engagement on retention and does not involve the leadership behavior composite. Second, the composite is decomposed to identify which specific behavior type drives the observed effects in the hospitality context, providing confirmatory evidence that task-oriented leadership behavior is the specific driver underlying H1 and H3. While H1, H3, and H4 are framed in terms of task-oriented leadership behavior based on the theoretical arguments developed below, the initial tests use the composite model, and the decomposition analysis supports task-oriented leadership behavior as the specific driver of the composite effects. H5 and H6 test whether the effects of task-oriented leadership behavior on engagement and the indirect effect on retention are stronger in the hospitality industry than in non-hospitality contexts. H7 examines whether age moderates the direct and indirect pathways from task-oriented leadership behavior to retention. This approach addresses the theoretical question of which leadership behavior type may assume primacy in a crisis and recovery context.

2.2 Leadership behavior and employee retention (H1)

Prior research establishes leadership behavior as an important antecedent of employee retention across industries (AbdELhay et al., 2025; Eom, 2015; Sang, 2025; Schneider and George, 2011; Schopman et al., 2017; Zhang and Wang, 2025). Evidence from hospitality similarly suggests that supervisory behavior matters for employee outcomes related to retention. Task-oriented leadership has been positively associated with employee service performance (Zhang et al., 2025), while supportive supervisory behavior has been linked to lower turnover intentions among hotel employees (Gordon et al., 2019). In the operationally demanding environment of hospitality, supervisors who provide task clarity, structure, and consistent guidance may reduce uncertainty and provide stability, particularly during periods of disruption such as the COVID-19 pandemic. Accordingly, task-oriented leadership behavior is expected to contribute positively to employee retention.

• H1: Task-oriented leadership behavior has a positive impact on employee retention.

2.3 Employee engagement and retention (H2)

The positive relationship between engagement and retention is consistently reported (Chanana and Sangeeta, 2021; Eltaybani et al., 2018). Engaged employees, those experiencing vigor, dedication, and absorption in their work, develop stronger organizational identification and lower turnover intentions. In hospitality, where employee turnover remains a persistent challenge (Dogru et al., 2023), maintaining employee engagement may therefore contribute to greater retention.

• H2: Employee engagement has a positive impact on employee retention.

2.4 Leadership behavior and employee engagement (H3)

Leadership behavior is consistently identified as a primary organizational driver of employee engagement (Book et al., 2019; Gallup, 2025; Xu and Thomas, 2011). In the hospitality context during COVID-19, supervisors who could define tasks clearly, organize resources, and monitor operations consistently may have provided greater structural certainty, thereby supporting employee engagement despite a volatile external environment. Among the three behavior types in the MPS taxonomy, task-oriented leadership behavior is particularly relevant to the hospitality context during COVID-19, given its direct relevance to operational clarity, service delivery consistency, and the kind of structured supervisory presence that frontline workers depend on most during periods of operational uncertainty. Accordingly, the following hypothesis is proposed:

• H3: Task-oriented leadership behavior has a positive impact on employee engagement.

2.5 Mediation of task-oriented leadership behavior on employee retention through engagement (H4)

While leadership and engagement are each independently related to retention, their interrelationship, and specifically whether engagement serves as the psychological mechanism linking supervisory behavior to retention, is less thoroughly examined. Theoretically, task-oriented leadership provides the structural clarity and operational support that activate employee engagement, and engaged employees are more likely to remain. Employee engagement, in this model, is the mediating process through which supervisory behavior translates into workforce stability.

• H4: Employee engagement mediates the relationship between task-oriented leadership behavior and employee retention.

2.6 Industry context as a moderator of task-oriented leadership behavior effects on employee engagement (H5)

The operational intensity and direct supervisory dependence that characterize hospitality employment suggest that the engagement effect of task-oriented leadership may be stronger in hospitality than in non-hospitality contexts. Task-oriented leadership behavior appears to be particularly beneficial for employee engagement in hospitality because the close supervisory interaction, performance feedback, and support for skill transfer characteristic of hospitality work can foster vigor, dedication, and absorption (El-Said et al., 2020; Zhang et al., 2025). During the pandemic, when health and safety protocols demanded new levels of operational precision, task-oriented supervision may have become even more important for maintaining frontline worker engagement.

• H5: The positive effect of task-oriented leadership behavior on employee engagement is stronger in the hospitality industry than in non-hospitality contexts

2.7 Industry context as a moderator of the indirect effect of task-oriented leadership behavior on employee retention (H6)

If the engagement effect of task-oriented leadership behavior is stronger in hospitality (H5), the indirect path from leadership through engagement to retention may be stronger there as well. The reasoning parallels H5 but extends one step further, from the supervisor's effect on how employees experience their work to the effect of that experience on whether they remain. Non-hospitality workers may have access to alternative engagement drivers, including remote work autonomy, professional identity, and flexible scheduling, that reduce their dependence on supervisory task behavior for engagement maintenance and, consequently, for retention.

• H6: The indirect effect of task-oriented leadership behavior on retention through employee engagement is stronger in the hospitality industry than in non-hospitality contexts.

2.8 Age as a moderator of task-oriented leadership behavior effects on employee engagement and retention (H7)

Employee age shapes the psychological relationship between leadership and work outcomes through differences in career stage, reciprocity norms, and expectations of the employment relationship (Guerrero and Marín, 2024; Sang, 2025; Zhang and Wang, 2025). Leadership behavior was found to have a stronger direct effect on engagement for younger employees in a general sample (Li et al., 2021). In the hospitality context, however, older workers, who have more organizational investment and stronger reciprocity expectations, may be more attuned to supervisor quality as a direct signal of organizational commitment to them (Cappelli, 2014; James et al., 2011). Younger hospitality workers cycle through roles at higher rates and may be retained not so much by the leader directly but by engagement itself: when they are engaged (experiencing meaning and pride in their work), they stay; when they are not, they leave. This suggests a divergent age pattern:

• H7: Age moderates the relationships between task-oriented leadership behavior, employee engagement, and retention: (H7a) the direct effect of task-oriented leadership behavior on engagement is stronger for older employees; (H7b) the direct effect of task-oriented leadership behavior on retention is stronger for older employees; and (H7c) the indirect effect of task-oriented leadership behavior through engagement is significant across age groups and strongest for younger employees.

3 Methods

3.1 Study sample

This study employed a quantitative, cross-sectional survey design. Participants were recruited via social media platforms (LinkedIn and Facebook Groups) during the COVID-19 pandemic recovery period and were required to be currently employed with an immediate supervisor. All study procedures were conducted in accordance with ethical standards for research involving human subjects. The project was approved by the Lawrence Technological University Institutional Review Board. Data were collected via Alchemer online survey.

The final sample included N = 223 employed participants with an immediate supervisor. The sample was predominantly male (65.9%, n = 147), with 70.9% (n = 158) in the 18–39 age range. The age distribution was as follows: 18–29 (n = 63, 28.3%), 30–39 (n = 95, 42.6%), 40–49 (n = 32, 14.4%), 50–59 (n = 25, 11.2%), and 60 and above (n = 8, 3.6%). Participants were geographically diverse: 58% from the United States and 42% from Asia, Europe, and South America. Industry representation was approximately balanced, with 101 participants (45.3%) employed in the hospitality sector (restaurants, hotel/motel, travel) and 122 participants (54.7%) in non-hospitality industries. According to a power analysis using G^*Power 3.1 (Faul et al., 2009), this sample size was sufficient to obtain a power of 0.80, an alpha of 0.05, and an effect size of f² = 0.08, exceeding standard thresholds for the regression analyses conducted in this study. For the mediation and moderated mediation analyses, the sample size of N = 223 exceeds the minimum recommended threshold of N = 200 for stable indirect effect estimation in structural equation modeling contexts (Kline, 2016), providing reasonable confidence in the reliability of the bootstrapped confidence intervals reported.

3.2 Statistical analyses

Mplus v8.11 statistical modeling software (Muthén and Muthén, 2017) was used for all structural analyses. JASP 0.97 (JASP Team, 2026) was used for descriptive statistics and group comparison analyses. A competing measurement modeling strategy was employed in Mplus using confirmatory factor analysis (CFA) to investigate the construct validity of the three study constructs. Models were estimated using the robust maximum likelihood (MLR) estimator, which addresses slight to moderate departures from normality in Likert-type rating scale data (Li, 2016). For all CFA models, indicators were estimated to load onto their a priori theoretical factors and cross-loadings were constrained to zero (Wang and Wang, 2020). Evidence for the best-fitting model was established by robust (Satorra-Bentler) chi-square difference tests (Δχ²) calculated from loglikelihood values and scaling correction factors (Satorra and Bentler, 2010), evaluated at p < 0.05. Model fit was assessed against the following criteria: χ²/df < 3, CFI > 0.90, TLI > 0.90, RMSEA <0.08, and SRMR <0.08 (Chen, 2007; Wang and Wang, 2020).

Following the two-step approach recommended by Anderson and Gerbing (1988), construct validity was first established via CFA, after which validated composite scores were used as observed variables in subsequent path analyses. This approach was adopted because full SEM specifications with all indicator variables produced inadequate model fit when leadership behavior types were examined as separate predictors, likely due to high intercorrelations among the MPS subscales.

Mediation analyses were conducted using the MODEL INDIRECT command in Mplus with bias-corrected bootstrapped 95% confidence intervals (CIs) based on 5,000 draws. For moderated mediation analyses, the MODEL CONSTRAINT command was used to define conditional indirect and direct effects at specified values of the moderator, with bias-corrected bootstrapped 95% CIs based on 5,000 draws. For moderation analyses involving interaction terms, continuous variables were standardized prior to computing interaction terms to reduce multicollinearity (Aiken and West, 1991).

Standardized coefficients are reported for models without interaction terms. For models with interaction terms, unstandardized coefficients are reported instead, since a product term cannot be standardized in a way that leaves its coefficient interpretable (Aiken and West, 1991). Standardizing the constituent variables before forming the product places these coefficients on a common metric. Full information maximum likelihood (FIML) estimation was used to handle missing data.

3.3 Measurement

Leadership behavior was measured using the 48-item Managerial Practices Survey (MPS), a validated instrument grounded in a hierarchical taxonomy of leadership behaviors (Yukl et al., 2002). Participants rated their immediate supervisor's behaviors across 12 sub-categories organized into three meta-categories: task-oriented (planning and organizing, problem-solving, clarifying, and monitoring operations), relations-oriented (supporting, developing, recognizing, and empowering), and change-oriented (envisioning, encouraging innovation, facilitating collective learning, and implementing change). All items used a 5-point Likert scale (1 = Not at all to 5 = To a very great extent).

Employee engagement was measured using the 17-item UWES-17 (Schaufeli et al., 2002a), which assesses vigor (six items), dedication (five items), and absorption (six items) on a 7-point frequency scale (1 = Never to 7 = Always). The UWES-17 has been validated extensively in cross-national samples (Schaufeli et al., 2006). The dimensionality of the UWES has been debated in the literature, with some studies supporting a unidimensional structure and others supporting the original three-factor model (Kulikowski, 2017). In the present study, engagement was modeled as a single latent factor indicated by the vigor, dedication, and absorption subscale scores. This specification retains the three-dimensional structure at the indicator level while treating overall engagement as the construct of interest, and dimension-specific effects are examined separately in Section 4.2.

Employee retention was measured using the three-item intention-to-remain scale (Armstrong-Stassen and Ursel, 2009), which assesses the degree to which employees intend to continue working in their current organization. The scale demonstrated acceptable reliability in the original study (α = 0.84) and has been used in subsequent research on older worker retention (Armstrong-Stassen and Stassen, 2013). Items were scored on a 5-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree).

Table 1 presents means, standard deviations, Cronbach's alphas, and Pearson correlations among all study variables. Internal consistency was strong across all measures, with alphas ranging from 0.844 to 0.982. All bivariate correlations among study variables were statistically significant at p < 0.01. Task-oriented leadership behavior was positively correlated with employee engagement (r = 0.549) and retention (r = 0.491), and all three leadership behavior types were strongly intercorrelated (r = 0.826 to 0.925). Employee engagement was positively correlated with retention (r = 0.646).

Table 1. Means, standard deviations, Cronbach's alphas, and Pearson correlations among study variables.

N = 223. TO, task-oriented leadership behavior; RO, relations-oriented leadership behavior; CO, change-oriented leadership behavior; LB, leadership behavior composite (mean of TO, RO, and CO); VIG, vigor; DED, dedication; ABS, absorption; EE, employee engagement composite (mean of VIG, DED, and ABS); RET, employee retention; M = mean; SD = standard deviation. Cronbach's alpha coefficients are presented on the diagonal in italics. Leadership behavior variables are scored on a 5-point scale (1–5). Employee engagement variables are scored on a 7-point scale (1–7). Retention is scored on a 5-point scale (1–5). All correlations are significant at p < 0.01.

4 Data analysis and results

4.1 Construct validity

Construct validity was evaluated via CFA in Mplus, testing all three constructs simultaneously. Leadership behavior was modeled as a second-order factor indicated by three first-order factors (task-oriented, relations-oriented, and change-oriented), each of which was indicated by its constituent subscale scores. Employee engagement was modeled as a first-order factor indicated by the vigor, dedication, and absorption subscale scores. Retention was modeled as a first-order factor indicated by its three items. Subscale scores were used as indicators rather than individual items, an approach whose tradeoffs have been discussed at length in the measurement literature and which reduces the indicator-to-sample-size ratio (Little et al., 2002). To demonstrate that the proposed three-factor measurement model provides the best fit relative to more parsimonious alternatives, three competing models were also estimated: a one-factor model in which all indicators loaded onto a single common factor, a two-factor model combining leadership behavior and engagement indicators (Model A), and a two-factor model combining engagement and retention indicators (Model B). Acceptable model fit was determined using the following goodness of fit criteria: χ²/df < 3, CFI > 0.90, TLI > 0.90, RMSEA <0.08, SRMR <0.08 (Chen, 2007; Wang and Wang, 2020).

Results of the model comparison are presented in Table 2. Models were estimated using the robust maximum likelihood (MLR) estimator, and model comparisons used robust (Satorra–Bentler) chi-square difference tests. The proposed three-factor model provided significantly better fit than the one-factor model [Δχ(3)2 = 562.43, p < 0.001], Two-factor Model A [Δχ(2)2 = 232.19, p < 0.001], and Two-factor Model B [Δχ(2)2 = 90.79, p < 0.001], and met all goodness-of-fit criteria: χ² = 179.24, df = 129, χ²/df = 1.39; RMSEA = 0.042 [90% CI:0.026, 0.056]; CFI = 0.982; TLI = 0.979; SRMR = 0.031. Constraining the three leadership behavior factors to load on a second-order leadership behavior factor did not significantly degrade fit relative to a model allowing them to covary freely, Δχ(4)2 = 6.55, p = 0.162, supporting the hierarchical specification and the use of a leadership behavior composite in the structural models. Standardized factor loadings for leadership behavior subscale indicators ranged from 0.817 to 0.911, with second-order loadings of task, relations, and change-oriented factors onto leadership behavior ranging from 0.914 to 0.999. Standardized factor loadings for engagement subscale indicators ranged from 0.910 to 0.924, and for retention items from 0.798 to 0.809. These results provide strong evidence of construct validity across all three measures. Figure 1 presents the full measurement model, including the second-order structure of leadership behavior and the standardized loadings for all indicators.

Table 2. Confirmatory factor analysis model comparison.

LB, leadership behavior; EE, employee engagement; RET, retention. Two-factor Model A combines LB and EE indicators on a single factor with RET separate. Two-factor Model B combines EE and RET indicators on a single factor with LB separate. Models were estimated using the robust maximum likelihood (MLR) estimator, and model comparisons used robust (Satorra–Bentler) chi-square difference tests. The three-factor model was compared to the one-factor model [Δχ(3)2 = 562.43, p < 0.001], two-factor Model A [Δχ(2)2 = 232.19, p < 0.001], and two-factor Model B [Δχ(2)2 = 90.79, p < 0.001], indicating the three-factor model provided the best fit among the models tested. RMSEA, root mean square error of approximation; CI, confidence interval; CFI, comparative fit index; TLI, Tucker–Lewis index; SRMR, standardized root mean square residual; AIC, Akaike information criterion.

4.2 Direct effects: task-oriented leadership behavior, employee engagement, and retention (H1, H2, and H3)

To test H1, H2, and H3, a path analysis mediation model was estimated in Mplus in which leadership behavior (composite of task, relations, and change-oriented behaviors) predicted employee engagement, employee engagement predicted retention, and a direct path from leadership to retention was simultaneously estimated. Following the two-step approach recommended by Anderson and Gerbing (1988), construct validity was first established via CFA in Section 4.1, after which validated composite scores were used as observed variables in the path analysis. Analyses were conducted on the full sample (N = 223). Unstandardized and standardized structural path coefficients are presented in Table 3, Panel A.

Table 3. Direct and indirect effects of leadership behavior on retention through employee engagement.

N = 223. LB, leadership behavior composite; TO, task-oriented leadership behavior; RO, relations-oriented leadership behavior; CO, change-oriented leadership behavior; EE, employee engagement; VIG, vigor; DED, dedication; ABS, absorption; RET, retention; b, unstandardized coefficient; β, standardized coefficient (STDYX); SE, standard error of the unstandardized coefficient; CI, bias-corrected bootstrapped 95% confidence interval for the standardized coefficient, based on 5,000 draws; LL, lower limit; UL, upper limit. Panel A reports the composite leadership behavior model. Panel B reports the model testing the three behavior types as separate predictors. Panel C reports the model testing the three engagement dimensions as separate mediators. ^*p < 0.05. ^**p < 0.01.

Leadership behavior had a significant positive direct effect on retention [H1; β = 0.205, 95% CI (0.046, 0.384)]. Employee engagement had a significant positive effect on retention [H2; β = 0.531, 95% CI [0.349, 0.677)]. Leadership behavior significantly predicted employee engagement (H3; β = 0.562, 95% CI (0.432, 0.676)]. H1, H2, and H3 are all supported.

To identify which specific leadership behavior type drives the composite effects underlying H1 and H3, a supplementary path analysis was estimated testing task-oriented (TO), relations-oriented (RO), and change-oriented (CO) behavior composite scores as separate predictors of engagement, with coefficients presented in Table 3, Panel B. This approach was adopted because the full SEM specification with all indicator variables produced inadequate model fit when the three behavior types were entered as separate predictors, likely due to high intercorrelations among the MPS subscales. Task-oriented leadership behavior was the only significant predictor of employee engagement [β = 0.335, 95% CI (0.100, 0.572)]. Relations-oriented leadership showed a non-significant path [β = 0.282, 95% CI (−0.084, 0.592)], and change-oriented leadership was also non-significant [β = −0.022, 95% CI (−0.372, 0.294)]. None of the three behavior types had a significant direct effect on retention when entered simultaneously [TO: β = 0.104, 95% CI (−0.107, 0.319); RO: β = −0.039, 95% CI (−0.404, 0.332); CO: β = 0.150, 95% CI (−0.184, 0.486)]. It should be noted that VIF values for RO (7.232) and CO (8.249) indicate substantial multicollinearity among the three predictors, which is expected given the high intercorrelations among the MPS behavior types. These results suggest task-oriented leadership behavior is the specific driver of the composite leadership effects on engagement and retention, representing a notable departure from our previous research, which found change-oriented leadership to be the dominant predictor in a general sample (Li et al., 2021), and is discussed further below.

4.3 Mediation of task-oriented leadership behavior on employee retention through engagement (H4)

To test H4, the indirect effect of leadership behavior on retention through employee engagement was examined from the composite path analysis model estimated in Section 4.2. As shown in Table 3, Panel A, the composite indirect effect was statistically significant [IE = 0.298, 95% CI (0.204, 0.407)], suggesting that employee engagement mediates the relationship between leadership behavior and employee retention, supporting H4. Furthermore, the supplementary decomposition analysis in Section 4.2 established that task-oriented leadership behavior was the only significant predictor of engagement among the three behavior types, and only the indirect effect via task-oriented leadership behavior was significant (IE = 0.178, 95% CI [0.059, 0.329]), while the indirect effects for relations-oriented [IE = 0.150, 95% CI (−0.031, 0.361)] and change-oriented [IE = −0.012, 95% CI (−0.223, 0.145)] behaviors were not significant (see Table 3, Panel B).

A supplementary path analysis further examined which dimension of engagement may serve as the primary indirect pathway from task-oriented leadership behavior to retention (see Table 3, Panel C). Task-oriented leadership behavior was a significant positive predictor of all three engagement dimensions: vigor [β = 0.549, 95% CI (0.430, 0.644)], dedication [β = 0.545, 95% CI (0.418, 0.650)], and absorption [β = 0.471, 95% CI (0.342, 0.583)], suggesting that task-oriented supervision broadly activates employee engagement. However, only dedication served as a significant indirect pathway from task-oriented leadership to retention [IE = 0.198, 95% CI (0.079, 0.320)]. The indirect effects through vigor [IE = 0.064, 95% CI (−0.093, 0.211)] and absorption [IE = 0.063, 95% CI (−0.068, 0.216)] were not significant. The sum of all three indirect effects was significant [IE = 0.325, 95% CI (0.217, 0.451)], and the direct effect of task-oriented leadership on retention remained significant [β = 0.195, 95% CI (0.046, 0.355)].

Taken together, these results suggest employee engagement functions as the mechanism through which task-oriented supervision reaches retention, and that the mechanism is narrower than a composite account of engagement would imply. Task-oriented leadership behavior was associated with all three engagement dimensions, but only dedication carried the effect forward to retention. The engagement most consequential for whether employees remain may therefore be the experience of meaning, enthusiasm, and pride in one's work rather than energy or immersion, consistent with job demands-resources theory (Bakker and Demerouti, 2007).

4.4 Industry context as a moderator of task-oriented leadership behavior effects on employee engagement (H5)

To test H5, a moderation model was estimated in Mplus with task-oriented leadership behavior (TO), industry context (HOSP, coded 0 = non-hospitality, 1 = hospitality), and their interaction (TO × HOSP) as predictors of employee engagement. All continuous variables were standardized prior to analysis (Aiken and West, 1991). The interaction term was significant (b = 0.196, SE = 0.064, t = 3.085, p = 0.002), supporting H5. The main effect of TO on EE was also significant (b = 0.575, SE = 0.064, p < 0.001), while the main effect of HOSP was not significant (b = 0.069, SE = 0.055, p = 0.208). As shown in Figure 2, the positive slope of task-oriented leadership behavior on engagement was steeper for hospitality workers (predicted EE range: 4.35 to 6.17) than for non-hospitality workers (predicted EE range: 4.65 to 5.54), suggesting that hospitality employees are more sensitive to supervisory task behaviors as a driver of engagement.

Table 4 presents descriptive statistics and group comparisons by industry context. Brown-Forsythe tests indicated unequal variances for task-oriented, relations-oriented, and change-oriented leadership behavior and for the leadership behavior composite; Welch's t-tests are reported for all variables. Compared to non-hospitality workers, hospitality workers reported significantly higher task-oriented leadership behavior [t_(220.99) = 2.178, p = 0.030], overall leadership behavior composite [t_(220.87) = 2.013, p = 0.045], vigor [t_(220.25) = 2.467, p = 0.014], absorption [t_(217.78) = 2.728, p = 0.007], and overall employee engagement [t_(219.23) = 2.368, p = 0.019]. Relations-oriented leadership, change-oriented leadership, and dedication did not differ significantly between groups. Notably, retention did not differ significantly between groups [t_(220.86) = 1.624, p = 0.106], suggesting the industry differences observed in the moderation analyses may reflect the differential effects of leadership behavior rather than differences in retention between the two groups.

Table 4. Descriptive statistics and group comparisons by industry context.

N = 223 (hospitality, n = 101; non-hospitality, n = 122). TO, task-oriented leadership behavior; RO, relations-oriented leadership behavior; CO, change-oriented leadership behavior; LB, leadership behavior composite; VIG, vigor; DED, dedication; ABS, absorption; EE, employee engagement composite; RET, employee retention; M, mean; SD, standard deviation. Welch's t-tests are reported for all variables, as Brown–Forsythe tests indicated unequal variances for TO, RO, CO, and LB. ^*p < 0.05. ^**p < 0.01.

4.5 Moderated mediation by industry context (H6)

To test H6, a moderated mediation model was estimated in Mplus using bias-corrected bootstrapped confidence intervals (5,000 draws). Task-oriented leadership behavior, industry context (HOSP), and their interaction (TO × HOSP) were included as predictors of both employee engagement and retention, with employee engagement as the mediator. All continuous variables were standardized prior to analysis (Aiken and West, 1991). As shown in Table 5, the indirect effect of task-oriented leadership behavior on retention through engagement was significant in both non-hospitality [IE = 0.211, 95% CI (0.119, 0.344)] and hospitality contexts (IE = 0.421, 95% CI (0.269, 0.605)]. The index of moderated mediation for the indirect effect was significant [IMM_IE = 0.209, 95% CI (0.076, 0.392)], suggesting the indirect effect was stronger in the hospitality industry than in non-hospitality contexts (Hayes, 2015), supporting H6. The direct effect of task-oriented leadership behavior on retention did not differ by industry [IMM_DE = 0.054, 95% CI (−0.232, 0.345)]. Conditional direct effects were of similar magnitude in both contexts, and the difference between them was not distinguishable from zero. What varied by industry was therefore the mediated route rather than the association between supervision and retention overall. Figure 3 illustrates the indirect and direct effects by industry context.

Table 5. Conditional indirect and direct effects of task-oriented leadership behavior on retention by industry context and age group.

IE, indirect effect of task-oriented leadership behavior on retention through employee engagement. DE, direct effect of task-oriented leadership behavior on retention. Both are unstandardized coefficients estimated with all continuous variables standardized prior to analysis. CI, bias-corrected bootstrapped 95% confidence interval based on 5,000 draws; LL, lower limit; UL, upper limit; IMM, index of moderated mediation (Hayes, 2015); a CI excluding zero indicates the effect differs across levels of the moderator. Industry context was coded 0 = non-hospitality, 1 = hospitality; the IMM for the indirect effect was significant, while the IMM for the direct effect was not. Age was recoded into four groups (1 = 18–29, 2 = 30–39, 3 = 40–49, 4 = 50 and above) and standardized prior to analysis; the IMM for the indirect effect was not significant, while the IMM for the direct effect was significant. ^*p < 0.05. ^**p < 0.01.

4.6 Moderated mediation by age (H7)

To test H7, age was examined as a moderator of three pathways: (a) task-oriented leadership behavior → employee engagement (H7a), (b) task-oriented leadership behavior → retention direct effect (H7b), and (c) the indirect effect of task-oriented leadership behavior on retention through engagement (H7c). Age was recoded into four groups (1 = 18–29, 2 = 30–39, 3 = 40–49, 4 = 50 and above) to improve stability of estimates, particularly for the oldest age category. A moderated mediation model was estimated in Mplus with bias-corrected bootstrapped confidence intervals (5,000 draws). All continuous variables were standardized prior to analysis (Aiken and West, 1991). Results are presented in Table 5.

• H7a: Age as Moderator of Task-Oriented Leadership Behavior→Employee Engagement

The interaction between task-oriented leadership behavior and age on employee engagement did not reach conventional significance (b = −0.112, SE = 0.065, p = 0.082), providing partial support for H7a. The negative sign indicates the TO → EE path weakens as age increases, consistent with the hypothesized direction. The main effect of age on engagement was significant (b = 0.109, SE = 0.054, p = 0.043), indicating older employees reported higher engagement overall, independent of leadership behavior.

• H7b: Age as Moderator of Task-Oriented Leadership Behavior→Retention (Direct Effect)

The interaction between task-oriented leadership behavior and age on the direct effect of retention was significant (b = 0.112, SE = 0.047, p = 0.017), supporting H7b. The index of moderated mediation for the direct effect was also significant [IMM_DE = 0.112, 95% CI (0.023, 0.211)], suggesting the direct effect of task-oriented leadership on retention strengthens with age. As shown in Table 5, the direct effect was non-significant for the youngest group [18–29: DE = 0.067, 95% CI (−0.127, 0.279)] and increased, becoming significant from 30 to 39 [DE = 0.179, 95% CI (0.013, 0.342)] through 50 and above [DE = 0.401, 95% CI (0.193, 0.637)]. Older employees thus appear to respond to supervisory task behavior more directly in their retention decisions, independent of the engagement pathway.

• H7c: Age as Moderator of the Indirect Effect

The indirect effect of task-oriented leadership behavior on retention through engagement was significant across all four age groups, providing partial support for H7c. The indirect effect was largest for the youngest employees [18–29: IE = 0.383, 95% CI (0.256, 0.554)] and declined with age [50+: IE = 0.199, 95% CI (0.062, 0.357)], with all four indirect effects reaching statistical significance. The index of moderated mediation for the indirect effect was not significant [IMM_IE = −0.062, 95% CI (−0.139, 0.001)], indicating that while the pattern is consistent with H7c, the formal moderation test did not support it. Engagement therefore appears to operate as a retention pathway across age groups, with the effect somewhat stronger among younger employees.

5 Discussion

This study examined leadership behavior, employee engagement, and retention in the hospitality industry during and after COVID-19. Results supported H1 through H6 and H7b, with partial support for H7a and H7c. The findings indicate a moderated mediation structure in which task-oriented leadership behavior is positively associated with both engagement and retention, engagement mediates the leadership–retention relationship, and the direct and indirect pathways are conditioned differently by industry context and employee age. These findings extend our previous research (Li et al., 2021) and offer several contributions to theory and practice.

5.1 Theoretical contribution

Our findings carry four implications for how leadership behavior is theorized in relation to employee engagement and retention. First, decomposing the leadership composite into its three constituent behavior types showed task-oriented behavior alone retained a significant association with engagement, while relations-oriented and change-oriented behaviors did not. This suggests the relative primacy of leadership behavior types may be a function of the operating environment rather than a stable property of the leadership–engagement relationship. Prior work associating change-oriented and transformational approaches with engagement was developed largely under conditions of operational stability (Derue et al., 2011), and our earlier study of a general multi-industry sample found change-oriented behavior to be the strongest predictor of engagement (Li et al., 2021). If task-oriented behavior assumes primacy when operational conditions deteriorate, claims about which leadership behaviors matter most may need to be stated with reference to the conditions under which they were observed. Contingency accounts of leadership propose that no single behavioral profile is universally effective and that situational demands determine which behaviors are most useful (Yukl and Mahsud, 2010). What such accounts generally do not specify is which leadership behaviors matter under which conditions. Our finding that task-oriented behavior predicted engagement, alongside recent evidence that leadership effects on engagement differ by work context (Sedefoglu et al., 2024), suggests that operational disruption may be one such condition, increasing the importance of task-oriented behaviors such as planning, clarifying, and monitoring, though the present design cannot separate crisis conditions from other features of this sample.

Second, estimating the indirect effect separately for hospitality and non-hospitality employees showed the pathway from leadership behavior through engagement to retention operated in both groups but was stronger among hospitality employees. The direct association between leadership behavior and retention did not differ by industry, suggesting the strength of the mediating mechanism may itself depend on context while the direct association does not. How much of an employee's daily work experience the supervisor shapes may bear on how much of the effect that pathway carries. Employees with access to autonomy, scheduling flexibility, or professional identities that sustain engagement independently of supervision may rely less on the supervisory pathway. For employees without those alternatives, the pathway may carry more of the load. A systematic review of hospitality leadership research identified the leadership-to-employee outcome pathway as underspecified (Elkhwesky et al., 2022). Specification may require attention not only to which behaviors matter but to why the intervening mechanism varies in strength across settings.

Third, modeling the three engagement dimensions as separate indirect pathways showed only dedication carried the effect from leadership to retention, even though task-oriented behavior was positively associated with all three dimensions. This raises a question about treating engagement as a unitary outcome, since the dimensions may not be interchangeable with respect to what activates them or what they predict. A composite account of engagement may obscure which aspect of the construct a given supervisory behavior reaches. Job demands-resources theory offers one account for why supervisory clarity and structure might activate meaning and identification more than energy or immersion (Bakker and Demerouti, 2007), consistent with meta-analytic evidence linking job resources to engagement in operationally demanding environments (Nahrgang et al., 2011). Theoretical claims about engagement as a mediator may therefore need to specify which dimension is doing the work, particularly when the outcome is behavioral rather than affective.

Fourth, estimating the direct and indirect pathways separately across age groups showed the direct association between leadership and retention strengthened significantly with age, while the pathway through engagement remained significant in every age group without differing significantly across them. The age gradient in the direct route is consistent with accounts emphasizing accumulated organizational investment and reciprocity expectations among longer-tenured workers (Cappelli, 2014; James et al., 2011). That the two routes behaved differently raises the possibility that age shapes which pathway carries the effect rather than simply how large that effect is, though only the direct route showed statistically distinguishable variation by age. This account is therefore provisional and requires replication before it can be advanced with confidence.

5.2 Practical implications

The study design was cross-sectional and correlational. The implications below identify where organizations might direct attention given the associations observed, not interventions this study demonstrated to be effective. Causal claims about any of them would require experimental or longitudinal evidence.

Concentrate supervisory development where supervisory contact is highest. Employees reporting higher task-oriented supervision also reported higher engagement, and this association was stronger among hospitality employees than among employees in other sectors. Organizations allocating limited development resources may find frontline service settings a reasonable place to concentrate them, given the degree of supervisory control over daily work in those settings. Supervisor support has been linked meta-analytically to engagement in operationally demanding environments (Nahrgang et al., 2011), and a substantial share of managers report having received no formal management training (Gallup, 2025), which suggests development need is unlikely to be evenly distributed. What the present findings add is a basis for prioritizing among settings rather than evidence that such development raises engagement.

Include operational competencies in how supervisors are assessed. When the three leadership behavior types were examined simultaneously, only task-oriented leadership behavior retained a significant association with engagement. Given the high intercorrelation among the leadership behavior types, this identifies task-oriented leadership behavior as the strongest correlate rather than an isolated cause. Assessment frameworks emphasizing vision and inspiration may nonetheless under-weight the clarity, planning, and monitoring behaviors most closely associated with engagement in operationally intensive settings. This is consistent with the argument that leadership effectiveness depends on matching behavior to situational demand rather than on a fixed behavioral profile (Yukl and Mahsud, 2010), and with evidence that leadership effects on engagement differ by work context (Sedefoglu et al., 2024). Whether changing what is assessed changes what supervisors do, and whether that in turn changes employee outcomes, are questions this study cannot answer.

Differentiate retention efforts by employee age. The direct association between supervision and retention strengthened with age, while the association running through engagement was significant in every age group and largest among the youngest employees. For organizations with age-diverse frontline workforces, this suggests retention efforts aimed at older employees may need to attend to supervisory quality directly, consistent with evidence that longer-tenured workers interpret supervisory treatment as a signal of organizational commitment to them (Cappelli, 2014; James et al., 2011). Efforts aimed at younger employees may depend more on whether the work is experienced as meaningful, since dedication was the engagement dimension most closely tied to retention in these data. The age difference in the indirect route did not reach statistical significance, so this distinction should be treated as provisional.

5.3 Limitations and future research

Several limitations should be considered. First, the cross-sectional design precludes causal inference. Longitudinal data collected across crisis onset, peak disruption, and recovery phases would provide stronger evidence for the directional effects proposed here. Second, all data were collected during a single period of sustained disruption. Operational conditions did not vary within the sample, so the interpretation that task-oriented leadership behavior assumed primacy under crisis conditions could not be tested directly. Comparative designs sampling the same industry across stable and disrupted periods, or measuring operational volatility as a variable rather than treating it as a fixed feature of the study context, would allow that boundary to be examined. Third, all measures were self-reported from the same source, creating potential for common-method bias. Future research should employ multi-source designs, for example, supervisor-rated leadership behavior paired with subordinate engagement reports, or 360-degree assessment. Fourth, while the sample is geographically diverse across 13 countries, the hospitality subsample may reflect idiosyncratic conditions in specific regional markets. Cross-national comparative studies would clarify whether the task-oriented leadership advantage in hospitality generalizes across cultural orientations toward leadership and work.

Several constraints concern what was measured and how it was modeled. Industry served as a proxy for the structural features of work that plausibly account for the difference observed here, including supervisory contact frequency, task interdependence, and access to schedule autonomy, none of which were measured directly. Measuring them would permit a more precise test of why the supervisory pathway varies in strength, and would avoid attributing to industry what may be attributable to how work is organized within it. External-oriented leadership, encompassing networking, external monitoring, and organizational representation (Yukl et al., 2019), was not examined in this study. Given the pandemic's disruptions to external business environments, external-oriented behaviors may have played an important unmeasured role in organizational resilience and should be incorporated in future research.

, The decomposition analysis in Section 4.2 also revealed substantial multicollinearity among the three leadership behavior types (VIF values for RO = 7.232 and CO = 8.249). This is expected given the intercorrelations inherent in the MPS taxonomy but warrants cautious interpretation of the individual path coefficients. Future research should explore alternative analytic approaches, such as latent profile analysis (Spurk et al., 2020) or dominance analysis (Azen and Budescu, 2003), to more cleanly parse the relative contributions of each behavior type. Relatedly, dedication was the only engagement dimension carrying a significant indirect effect, which suggests studies modeling engagement solely as a composite may not detect which aspect of the construct a given predictor reaches. Research examining engagement as a mediator of behavioral outcomes may benefit from estimating dimension-level indirect effects alongside composite-level effects.

Two further directions warrant attention. Age moderation findings for H7a and H7c were partially supported. The interaction effects were in the hypothesized direction but did not reach conventional significance thresholds. These findings should be interpreted with caution and replicated in larger samples with greater statistical power for detecting moderation effects. Finally, future studies should investigate these relationships separately in for-profit vs. non-profit hospitality organizations, and across hospitality sub-sectors, as the motivational structures and retention challenges may differ meaningfully across these contexts.

6 Conclusion

The COVID-19 pandemic did not simply disrupt the hospitality industry. It created conditions under which existing leadership prescriptions, developed largely in stable environments, had not been tested. Accounts of the period describe rapidly changing protocols, volatile staffing, and immediate concern for the physical safety of employees and customers (Dale Oen et al., 2022; Kniffin et al., 2021). In this sample, drawn from that period, the supervisory behaviors most closely associated with engagement and retention were not visionary ones. They were the behaviors that brought order, clarity, and stability to daily work. Task-oriented leadership behavior was the behavior type most strongly associated with both outcomes.

What distinguishes these findings from prior work is not simply the confirmation that leadership matters, but the precision with which the conditions of that influence are specified. The mediation of leadership on retention through engagement operates more strongly in hospitality than in other sectors, with the indirect effect nearly twice as large in the hospitality context. The engagement dimension most closely tied to retention was dedication, the experience of meaning and pride in one's work, not engagement broadly defined. The age of the employee shapes whether leadership works directly on retention or indirectly through engagement, with the direct effect association strengthening significantly with age and the association through engagement remaining significant in every age group, though the latter difference did not reach conventional significance. These boundary conditions are not minor qualifications. They are the contribution, because they tell organizations not just that leadership matters but exactly how, for whom, and in which context it matters most.

The post-pandemic era has not resolved the hospitality workforce crisis. Quit rates remain among the highest of any industry sector (Dogru et al., 2023; U.S. Bureau of Labor Statistics, 2024), and the structural vulnerabilities that COVID-19 amplified have not disappeared. What has changed is our understanding of the supervisory behaviors and psychological mechanisms through which those vulnerabilities can be addressed. Organizations that invest in task-competent supervisors, cultivate dedicated engagement among their frontline workers, and attend to how retention pathways differ across age groups will be better positioned not only for the next crisis but for the ongoing challenge of sustaining a committed workforce in one of the world's most demanding industries. What this study offers is a clearer account of where to look. Establishing what changes when organizations act on it remains work for future research.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Anthropic Claude Sonnet 4.6 was used to assist with manuscript drafting, editing, and structural organization. All content, analytical decisions, interpretations of results, and conclusions were reviewed, verified, and approved by the author(s). The author(s) take full responsibility for the accuracy and integrity of the work, including all citations, statistical reporting, and theoretical claims. No generative AI was used in data collection, statistical analysis, or the generation of results.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/forgp.2026.1851922/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Ron Landis, Clemson University, United States

Reviewed by: Hyoun Sook Lim, Central Connecticut State University, United States

Nicholas Moon, California State University, San Bernardino, United States