DIFFERENT STYLES ON

In hospitality, certain management styles can play a crucial role in achieving positive employee outcomes. This study aims to investigate how different leadership styles can contribute to maximizing hospitality workers ’ potential. The proposed theoretical model draws on emerging approaches to leadership (paradoxical, empowering, servant) and is tested with structural equation modeling (SEM) using data from 340 employees in Spanish hotels. The findings may be explained by Self-Determination Theory: empowering and paradoxical leadership styles show positive relationships to psychological empowerment. Contrary to expectations, servant leadership style was not an antecedent of psychological empowerment. Furthermore, this study ascertains the positive relationship of empowering and servant leadership styles to engagement. The findings also demonstrate psychological empowerment to be a clear antecedent of job engagement, extending previous research. Implications for hospitality service managers, educators, and researchers are discussed.


INTRODUCTION
The hospitality industry has a unique culture as compared to other industries. It is a sector of frequent interaction between customers and employees, where frontline staff play a crucial role in service delivery (Terglav, Konečnik Ruzzier, & Kaše, 2016). The profitability of hospitality organizations depends on essential employee attitudes and behaviors (Úbeda-García, Claver Cortés, Marco-Lajara, & Zaragoza-Sáez, 2014). Paradoxically, hospitality workers frequently report emotional exhaustion, lack of appreciation, occupational stress, overwork, and low pay (Kim & Agrusa, 2011;Tongchaiprasit & Ariyabuddhiphongs, 2016). In fact, high absenteeism and turnover are also characteristics of these working environments.
Further, as Øgaard, Marnburg, and Larsen (2008) highlight, hospitality organizations are conventionally characterized as highly hierarchical, with a predominance of traditional management styles; most hotels follow the classical model of centralized decision making within strict pyramidal organization charts. Nevertheless, in this particular service sector, many unpredicted situations may occur, and if frontline employees are not given responsibility for decision making, they may find it hard to solve problems quickly and provide high-quality customer service (Jha & Nair, 2008). Since every customer and each service experience are different, hospitality employees should have some degree of autonomy and discretion in service delivery -and its recovery, when necessary -to meet customers' differing needs, demands, and expectations (Ro & Chen, 2011). For this and other reasons, managing people in this special sector involves unique challenges for managers (Bowen & Ford, 2002).
On the other hand, several economic and financial variables have pushed companies to new standards, forcing them to redefine business processes in the past decade. Delayering of management structures (in many cases due to downsizing) has been accompanied by a changing balance of power between employer and employee. As a result, new flattened organizational models seem to have emerged in conjunction with changes in the management philosophy. In fact, many organizations have been compelled to shift their traditional pyramidal, top down concept of control towards more flexible and participatory managerial formulas. This change in managers' roles and responsibilities appears to have required a corresponding regeneration of the types of leadership behavior they employ. Leaders are now required to be more adaptable and people-oriented. New leadership strategies are therefore needed to motivate the 21 st -century workforce and to increase their positive psychological capital (Deloitte, 2014).
In line with this perspective, certain empowering behaviors by managers (related to their leadership style) can foster workers' positive feelings of proficiency and autonomy by enhancing their self-perception and sense of achievement, in turn generating psychological empowerment and engagement. For example, certain enabling managerial behaviors can satisfy two of our three essential human needs, those of competence and self-determination.
In sum, SDT is useful for understanding how various social forces and managerial behaviors affect individuals at work.
Drawing on theory and empirical evidence, this study thus argues that paradoxical leadership, empowering leadership, and servant leadership are associated with employee psychological empowerment and workers' engagement. The research model proposes that employees' perceptions of empowering, paradoxical, and servant leadership styles in their managers are positively related to both psychological empowerment and job engagement levels in employees, as depicted in Figure 1.

Theoretical framework on engagement
Employee engagement is commonly described as a psychological state in the workplace in which feeling competent and secure motivates employees' proactive dedication through physical, cognitive, and emotional commitment to their roles (Schaufeli et al., 2002). Although engagement is often confused with or seen as part of different constructs (e.g., commitment or involvement) in practitioner-oriented research, it differs in associating these cognitive, emotional, and behavioral elements closely with the ways individuals perform their job roles (Saks, 2006). For Schaufeli et al. (2002), engagement is characterized as workers' experience of vigor (desire and commitment to perform their jobs energetically and to do their very best), dedication (being devoted to and inspired by their job), and absorption (being completely focused on and immersed in the task), thus conceptualizing engagement as a three-dimensional construct. Engaged employees are self-assured about their capacity to accomplish the demands of their jobs.
Job engagement is not a transient and situational state of mind, but rather an "employee's persistent and pervasive affective-cognitive" connection with work activities (Schaufeli et al., 2002: p. 74). Key antecedents of workplace engagement include realistic workloads, employees' belief that they have autonomy and are in control, a supportive work environment, appropriate recognition, and meaningful work. Maslach, Schaufeli, and Leiter (2001) argue that one should think of job engagement as the opposite of burnout. Kahn (1990: p. 700) views engagement as "the simultaneous employment and expression of a person's preferred self in task behaviors that promote connections to work and to others, personal presence (physical, cognitive, and emotional), and active, full role performances." These three psychological conditions are security, significance, and availability, as both personal matters and the workplace strongly determine whether one is engaged or disengaged in one's job (Kahn, 1990). In this line, Rich, Lepine, and Crawford (2010) identify value congruence (belief that one's values match the organization's), perceived organizational support (feeling safe in organizational contexts perceived as having clear behavioral consequences), and core self-evaluations (assessing personally whether one is valuable, effective, and capable) as key antecedents of job engagement.

Psychological empowerment
Research on how to understand and foster the experience of empowerment by subordinates describes empowerment as a motivational process (Conger and Kanungo, 1988; Thomas and Velthouse, 1990). Drawing on prior research, Spreitzer (1995) also classifies the construct of psychological empowerment as motivational, conceptualizing it as a psychological work-related state reflecting the employee's experience of being actively oriented to his/her role. According to this author, psychological empowerment is a second-order construct signaled through four cognitive perceptions: meaning (employees value their jobs), competence (employees believe they can perform jobrelated tasks with skill), self-determination (perception that one has some choice), and impact (belief that fulfilling the goal of a task will have positive results).
The self-efficacy theory developed by Bandura (1977) has been traditionally described as one of the best frameworks for understanding employees experiencing psychological empowerment. Individuals with self-efficacy believe inwardly that they are capable and competent to carry out the required courses of actions successfully to achieve certain goals. Given its positive impact on the employee's intrinsic self-efficacy, theoretical arguments suggest that psychological empowerment can be critical for employee engagement to thrive. According to Murari and Gupta (2012), different types of leadership impact employees' psychological empowerment differently, since the management's style of leadership determines how empowered the organization's employees feel. Empowering, servant, and paradoxical leadership styles involve managerial practices that in one way or another enhance SDT basic needs, such as competence and autonomy. It is thus expected that these leadership approaches result in employees' experiencing more psychological empowerment and being more engaged.

Empowering leadership
The empowering leadership style involves managers giving their staff more power and freedom of choice in decisions. Following the multidimensional conceptualization of Ahearne, Mathieu, and Rapp, (2005), an empowering leadership approach creates conditions that foster employee participation by promoting decision making and reducing bureaucratic constraints. Empowering leaders make their belief in employees' competences and capabilities visible, granting employees more opportunity and responsibility in their jobs. Such leaders normally set participatory goals, foster autonomy, and promote self-development of the workforce by fostering autonomy through, for example, by delegating authority and permitting employees to assume a leadership role or regulate themselves (Ahearne et al., 2005;Bester, Stander, & Van Zyl, 2015).
The literature demonstrates across work contexts and national differences that managerial leadership to empower employees correlates highly with psychological empowerment of employees (see Bester et al., 2015;Klerk & Stander, 2014;Lorinkova & Perry, 2014;Zhang & Bartol, 2010). Namasivayam et al. (2014) also found evidence associating empowering leadership with psychological empowerment in their sample of restaurant frontline employees in the US. Hence, prior studies illustrate how empowering leadership exerts a direct influence on psychological empowerment.
Management that plays a participatory role is critical to the employee's psychological state (Conger & Kanungo, 1988). Tuckey et al. (2012) state that empowering leadership creates the right work environment to boost employee self-management by teaching employees self-leadership skills. Moreover, the actions oriented to empowering employees seem to improve their belief in their self-efficacy. The foregoing arguments thus lead us to hypothesize that empowering leadership has a positive effect on subordinates' psychological empowerment:

Hypothesis 1. Managers' empowering leadership is positively related to employees' psychological empowerment.
Leaders strongly influence employees' work experience and psychological well-being. Klerk and Stander (2014) demonstrate that empowering leadership, psychological empowerment, and engagement are closely related. Employees feel empowered when they perceive that their managers seek to empower them through their leadership. And these feelings will lead employees to feel that they are connected and belong to their organization and their jobs, resulting in employees being more engaged (Albrecht & Andreetta, 2011).
Specifically, since empowering leadership has the effect of optimizing working conditions for engagement, this style directly inspires work engagement in followers (Tuckey et al., 2012). This enabling leadership style thus facilitates the motivational processes that underpin employees' engagement at work. Perceiving that their manager is giving them power, autonomy, and authority makes employees feel trusted and more capable, and consequently will be more enthusiastic when performing work-related tasks and will be more engrossed in their work. It is thus hypothesized that managerial practices' emphasis on employees' autonomy and self-determinationas is the case in empowering leadershipwill give priority to employee care and dedication, generating a feeling of engagement. If empowering leadership practices that foster autonomy and discretion among the employees lead to increased level of engagement, we can expect a positive association between such leadership and employee engagement, as formulated in Hypothesis 3: Hypothesis 2. Managers' empowering leadership is positively related to employee engagement.

Paradoxical leadership
Paradox theory illuminates how managers understand and manage opposing requests in organizations. Paradoxical leader behavior responds to new and seemingly contradictory inquiries that managers must currently face. Outside the conventional managerial duties, a paradoxical leader establishes a balance between mutually exclusive assumptions or conditions in the workplace which are nevertheless deeply interrelated and may even coexist (Lewis, Andriopoulos, & Smith, 2014). The literature defines paradoxical leadership as apparently competing but in fact interrelated behaviors displayed by the leaders "to meet structural and follower demands simultaneously and over time" (Zhang et al., 2015: p. 538). They integrate both organizational and individual solutions harmoniously and simultaneously.
Only through paradoxical leadership can conflicting poles be united successfully. Managers following the paradoxical leadership approach maintain decision control while allowing autonomy (Zhang et al., 2015). For example, they normally control important work issues and make final important decisions but, at the same time, allow subordinates to handle other smaller issues and work details. They do not micromanage, but allow employees full control of specific work processes. While maintaining overall control, these leaders give subordinates appropriate autonomy and latitude. Paradoxical leaders thus allow others to share the leadership role in some aspects of the daily work.
Further, paradoxical leaders empower their followers while keeping control, combining both autonomy and discipline in the way they manage their teams. In fact, paradoxical leaders might "maintain long-term control precisely by continuously granting employees the discretion to bend the rules" (Lewis et al., 2014;Zhang et al., 2015: p. 541). As paradoxical leadership provides some degree of discretion and freedom of choice to employees, it is likely to influence workers' feeling of empowerment: Hypothesis 3. Managers' paradoxical leadership is positively related to employees' psychological empowerment.
In hospitality organizations, frontline employees sometimes experience role conflict in dealing with opposing components of paradoxes at work (Iverson & Deery, 1997). For example, company policies make employees follow standardized corporate behavior while at the same time continuously requiring the staff to customize service to each client individually. This can cause confusion, frustration, and lack of motivation among employees. For this reason, and given the peculiarities and uncertainty of service encounters, paradoxical leadership may be a good way to manage workers in hospitality. Under the paradoxical leadership approach, managers stress conformity in task performance but allow for exceptions, as Zhang et al. (2015) outline.
Moreover, paradoxical leaders' behavior may provide a good example for employees because these leaders must also consider opposing viewpoints together and integrate them over time (Lewis et al., 2014). For example, on the one hand, subordinates expect their line managers to grant them discretion in delivering guest service; on the other, the organization expects the line manager to supervise and control processes. Paradoxical leadership can thus deal effectively with workplace tensions or contradictions by embracing two possible alternatives simultaneously and integrating them as a pair (Lavine, 2014). Supervisors' paradoxical leadership behaviors could thus serve as a good role model to show employees how to accept and embrace contradictions in complex environments (Lewis, 2000). As a result, employees could learn to fulfill complicated and even ambivalent requirements.
Further, a paradoxical leader's example may be able to teach employees how to meet differing customer expectations, to be more interpersonally flexible. With paradoxical leader behavior as a role model, subordinates could learn greater flexibility and acquire a varied repertoire of responses to unexpected work situations, providing experience that in turn contributes to expanding their holistic understanding. Followers might therefore be more likely to be adaptive and proactive, and to work proficiently, exploring new ways to fulfill their daily job requirements (Zhang et al., 2015) and thus coming to feel more capable and enthusiastic at work. Furthermore, paradoxical leaders adapt their commands to the aptitude, strengths, and capabilities of each subordinate and regulate their expectations to the heterogeneity of each situation, while still assigning equal workload; they also adjust their behavior to the needs of different team members, while keeping homogeneous standards for all, as Zhang et al. (2015) argue. Subsequently, employees might see themselves as more task capable and interactively skilled, becoming in turn more dedicated and eager at work and, as hypothesized in H5, more engaged: Hypothesis 4. Managers' paradoxical leadership is positively related to employee engagement.

Servant leadership
Recent research has identified servant leadership as "the next step in leadership evolution" in hospitality sectors (Brownell, 2010: p. 363). Greenleaf (1977) developed the theory of servant leadership in a ground-breaking essay outlining the importance of prioritizing support for employees, as well as their comfort and development.
A servant leader is an individual who improves his/her followers' necessities, selfconcept, beliefs, and values, through firstly setting an example of high moral standards, integrity, kindheartedness, and dedication. Servant leaders believe in ethical utilization of power and encourage participant behaviors that enhance job efficacy (Liden et al., 2015).
Servant leadership influences employee behavior positively because it fulfills some of workers' psychological needs (Hsiao et al., 2015). In this line, Murari and Gupta (2012) highlight that some characteristics of servant leadership (foresight, persuading, awareness, and stewardship) are very important for employee empowerment. Their study finds that servant leadership and empowerment correlate highly and significantly. Servant leaders emphasize collaboration and the ethical use of power, enabling more autonomy and decision making in employees. Servant leaders include followers' input into important managerial decisions. As a result, followers feel that a significant part of the business' success is their responsibility. Servant leaders' behavior increases employees' self-determination and self-efficacy levels by providing them with autonomy, empowerment, and the conditions in which to grow and develop. Followers consequently gain more self-confidence and conviction in their ability to perform well. Moreover, studies demonstrate the marked effectiveness of servant leadership, given its employee-centered orientation to support and empower subordinates (Liden, Wayne, Liao, & Meuser, 2014). As servant leadership brings out employees' full potential, it is expected to influence employees' psychological empowerment. Based on the foregoing discussion, servant leaders' positive effect on employees' psychological states should produce greater individual psychological empowerment in employees.

Hypothesis 5. Managers' servant leadership is positively related to employees' psychological empowerment.
According to Liden et al. (2015: p. 254), the benefits of servant leadership make it a desirable approach both to leadership "in response to the increasing need for employee engagement" and to "societal demands for higher levels of ethical behavior in organizations". Servant leadership is altruistic and normally creates value for others, both inside and outside the organization; this behavior instills positive energy among their followers. Overcoming their self-interest, servant leaders prioritize serving their followers before attending to their own needs. This type of leader notices their subordinates' needs and desires by the way of personalized attention and takes on a key role in ensuring fairness while focusing on helping others.
Servant leaders express humility, authenticity, and interpersonal acceptance (Murari & Gupta, 2012). According to Owens et al. (2013Owens et al. ( : p.1533, "humble leaders foster learningoriented teams and engaged employees, as well as job satisfaction and employee retention." For this reason, we expect servant leaders to enable their workers to experience personal efficacy by fostering conditions for strengthening motivation to complete tasks. Prior research (e.g., Correia & van Dierendonck, 2014) also shows that servant leadership enhances employee engagement in service companies. Servant leaders promote conditions for psychological safety, one of the necessary conditions signaled by Kahn (1990) for employees to engage at work. The perception that their managers are servant leaders increases employees' likelihood of feeling less constrained and expecting success, and may thus feel more enthusiastic and emotionally attached to their work. We thus hypothesize that servant leadership positively affects employee engagement.
Hypothesis 6. Managers' servant leadership is positively related to employees' engagement.

Relationships among psychological empowerment, engagement, and leadership
For Quiñones, Van den Broeck, & De Witte (2013), conceiving psychological empowerment as an inner phenomenon is productive for understanding connections between engagement and workplace issues. Since empowered workers are confident in their abilities and the work that they do, they are more likely to be engaged (Stander & Rothmann, 2010). As reported by previous research, employees who feel empowered are not only motivated and engaged but experience more connection, commitment, and feelings of belongingness to their organization (Albrecht & Andreetta, 2011;Macsinga, Sulea, Sârbescu, & Fischmann, 2015). Hence, psychologically empowered employees tend to greater engagement, given their higher self-efficacy as well as competence.
Self-Determination Theory (SDT) as formulated by  asserts that psychological empowerment is also "particularly suitable to understand motivation at the task level, potentially affecting engagement" (Correia & van Dierendonck, 2014: p.880). Some research finds that psychological empowerment plays a relevant role as antecedent of work engagement (e.g., Jose & Mampilly, 2014;Quiñones et al., 2013;Stander & Rothmann, 2010). This study thus hypothesizes that employees who are empowered by their managers will feel better able to perform their tasks and thus more trusted. They will also be more inspired because they view their work as meaningful and satisfying. Such workers' dedication, passion, and engrossment at work increase: they become more engaged.
Hypothesis 7. Employees' psychological empowerment is positively related to employee engagement.
The three leadership styles-empowering, paradoxical, servant-normally show concern for their subordinates' wellbeing and encourage their involvement by developing supportive environments that encourage employee participation. As a result, employees perceive that their managers trust them and believe they will do excellent work.  SDT argues that managers' enabling behaviors may thus satisfy the employees' needs for self-determination and competence, generating psychological empowerment. In studies by Maslach et al. (2001) and subsequently Saks (2006), the perception of support from the organization and one's supervisor emerge as antecedents engagement at work. The climate of "psychological safety" generated by this type of leader may thus be ideal for enabling employees to experience engagement at work (Kahn, 1990).
Empowering leadership takes place in work environments with a fair approach to workload distribution. Paradoxical leaders redistribute workload uniformly, while also considering individuals' capabilities. Servant leaders normally serve as an example of the value of equity and high standards within the organization. Managers displaying these novel leadership styles are thus more likely to be just and equitable, conceding all empowered employees the same possibility to act with latitude and choice in how to perform their work. Maslach et al. (2001) as well as Saks (2006) identify procedural and distributive justice as key elements of employee engagement.
Moreover, participative leadership styles foster a more active role of the employee in daily tasks by providing him/her more latitude. Freedom of choice in how to accomplish duties implies new job demands, as it involves higher levels of job responsibility. One might expect employees who perceive those demands as challenges (opportunities to promote mastery) to be more engaged, as Crawford, LePine, and Rich (2010) argue. Managers who strive to empower employees are thus more likely to inspire engaged employees. Understanding that one has the higher responsibility involved in psychological empowerment can lead employees to perceive a potential for personal growth and thus to experience more engagement.
There are thus several reasons to expect that these three styles of participatory leadership produce beneficial outcomes in followers, including greater engagement through psychological empowerment. Following this logic, we can expect psychological empowerment to mediate in the relationship of managerial leadership style to follower engagement level, as stated in the following hypotheses:

Hypothesis 8. Psychological empowerment mediates the relationship between empowering leadership and employee engagement.
Hypothesis 9. Psychological empowerment mediates the relationship between paradoxical leadership and employee engagement.

Sample and data collection
A survey questionnaire was prepared to measure the perceptions of Spanish hotel employees. The scales were drawn from prior studies, adapting some measures to the context of the hospitality sector. Since the data were collected in a Spanish-speaking country, back-translation processes were followed (Brislin, 1980). Two bilingual experts ensured accuracy of the translation and adapted some expressions to the hotel industry. Prior to final administration, and to ensure comprehensibility of the questionnaire, a pretest with the final translated versions of the measures was piloted on a sample of 12 hoteliers.
Data were collected during the summer of 2015. Full-time hotel workers intensely involved in serving customers (reservation department and front-office staff) were chosen by convenience sampling. Since all questionnaires were personally administered by the research team, who approached the employees directly at their workplaces, participants were selected according their accessibility and availability. For this reason, the geographical scope of the study was limited to a manageable number of cities -ranked among the ten cities most visited throughout the year in Spain. The hospitality sector was chosen for this study due to its importance to the Spanish economy. Spain has over 10,000 hotels, employing a monthly average of over 300,000 workers; in 2016, over 330 million room nights were registered in Spain 1 . It is thus important to better understand how to foster positive outcomes in its workforce.
Seven days prior to administration of the survey, the hotels selected received an invitation by email that explained the research and procedures. The self-completion survey was subsequently given in person to the hotel employees who volunteered to participate in this study: 340 individuals. A cover letter that clarified the research goals and steps taken to ensure confidentiality accompanied the questionnaire. Questionnaires were returned in a closed envelope to guarantee anonymity. Respondents placed the completed surveys in envelopes, which researchers collected from the hotels in person.
Failure to complete the survey or missing data led to elimination of 11 questionnaires. Of the 329 hoteliers who completed the surveys, the sample may be characterized by gender as 193 men (58.7%), by education level as 78.1% holders of university degrees; and by age as 42.8% ages 21-30 and 39.8% ages 31-40. Most (76.6%) responding employees were not managers and had worked an average of 6 years with the company (SD=7.36 years). On average, they had held the position at the time of the survey for 5 years (SD= 6.58) and had been working with the current manager for the last 4 years (SD=4.83).

Measures
Up-to-date multi-item scales, recently developed in the management literature to measure new leader behaviors, were employed in this exploratory study. A 7-item Likert scale was used to record responses, from (1) totally disagree to (7) totally agree. All scales used this 7-point Likert response format.
Engagement was assessed with the UWES-9 (the short version of the Utrecht Work Engagement Scale) (Schaufeli et al., 2002) (e.g., "I feel happy when I am working intensely", "my job inspires me"). The scale yielded Cronbach Alpha of .93.
Psychological empowerment. This construct was operationalized using Spreitzer's (1995) scale (e.g., "I can decide on my own how to go about doing my work"). The four dimensions were reflected in a 12-item scale: significance, skill level, autonomy, and efficacy. The scale yielded a Cronbach Alpha of .90.
Empowering leadership. Empowering leadership was measured using the 12-item instrument from Ahearne et al. (2005). This instrument is divided into multi-item subscales measuring four dimensions: (1) greater significance of one's job, (2) encouraging participatory decision-making processes, (3) exhibiting trust that employees will perform with excellence, and (4) permitting independence from bureaucratic rigidity. The scale yielded a Cronbach Alpha of .95.
Paradoxical leadership. Paradoxical leadership was measured using the instrument by Zhang et al. (2015). This 22-item scale has five subscales that correspond to its five dimensions: (1) treating subordinates consistently but also permitting individual variation, (2) combining self-centeredness with other-centeredness, (3) controlling decision-making and permitting some autonomy, (4) enforcing regulations for performing tasks but also granting flexibility, and (5) maintaining both hierarchical authority and mutual respect. Therefore, high levels of paradoxical leadership behavior imply the degree to which leaders connect and embrace two paradoxical poles consistently. The scale yielded a Cronbach Alpha of .94.
Servant leadership. The short version of the servant leadership scale (SL-7), recently refined by Liden et al. (2015), measured employee perceptions of the degree to which management adopted a servant leadership style. It consisted of 7 items (e.g., "My supervisor puts my best interests ahead of his/her own") representing a single construct of servant leadership style. The scale yielded a Cronbach Alpha of .87.
Control variables. Two employee characteristics were controlled: gender (coding female 1 and male 2) and level of education, because both factors have been shown to relate to engagement (Macsinga et al., 2015;Quiñones et al., 2013). A 4-point scale was employed to measure education, on which higher values represented a higher educational level: up to completion of middle school (1), high school or equivalent (2), completion of an undergraduate degree (3), graduate study (4).

Data-analysis strategy
The data gathered from the questionnaire were analyzed using IBM's Statistical Package for Social Science (SPSS), version 22. Both the measurement and the structural model were assessed through EQS 6.2 (Bentler & Wu, 1995;Byrne, 2006) because this software provides robust statistics for the standard errors, thus correcting for the effects of violation of the principle of multivariate normality (Byrne, 2012).
Scholars disagree on whether to analyze the measurement models independently of structural ones (Wallace & Sheetz, 2014). In this study, the recommendations of Schumacker and Lomax (2010) have been followed. Those authors advise following two steps: (1) assess the measurement model; and (2) examine the structural model because "it is only after latent variables are adequately defined (measured) that it makes sense to examine latent variable relationships in a structural model" (Schumacker & Lomax, 2010: p.192). This perspective is consistent with prior research (Elicker, Levy, & Hall, 2006;Wallace, Keil, & Rai, 2004). It is also suitable in this case because it permits us to refine the measures before the structural model is tested (Wallace & Sheetz, 2014).

Assessing common method bias
To reduce the possibility of common method bias due to self-report measures, a set of procedures was implemented prior to data collection (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). For example, variables were scrambled to prevent respondents from intuiting the research model and questions, as it reduced the likelihood that they would adjust their answers to what they believed were the expected results (Terglav et al., 2016).
In addition, only previously tested scales were used, and "filler" items and changed instructions were added to separate the variables in the minds of the respondents (Alfes, Shantz, Truss, & Soane, 2013).
Common method bias (CMB) was also tested using Harman's single-factor test (Podsakoff & Organ, 1986). The results identified seventeen factors. Since the first factor explained only 20.185% and balanced distribution across the other factors explained the rest of the variance, CMB does not seem to be a problem (Podsakoff & Organ, 1986). A common latent variable comprised of all survey items was also included in the model . Following the recommendation of Matzler, Strobl, Stokburger-Sauer, Bobovnicky, and Bauer (2016), substantial and method variance were contrasted. The indicator's average substantive explained variance was 0.678 and the average method-based variance 0.007a ratio of 96.857:1. Furthermore, given the nonsignificance of the majority of method factor loadings, indicating only minor method variance, one can conclude that method bias does not affect the study results substantially.

Assessment of psychometric properties of the measures
The measurement model was evaluated using confirmatory factor analysis (CFA). Full CFA assesses whether item variables load excessively on previously determined factors but not on non-related factors (Schumacker & Lomax, 2010).

Figure 2. Measurement model. (Full Confirmatory Factor Analysis)
Estimation of the measurement model yielded global fit indexes suggesting poor datamodel fit. Causes of misfit were then investigated. To do so, the guidelines of Byrne (2010: p.67) were followed: "in reviewing the model parameter estimates, three criteria are of interest: a) the feasibility of the parameter estimates, b) the appropriateness of the standard error, and c) the statistical significance of the parameter estimates." First, correctness of the magnitude and sign of the parameter estimates were confirmed. Although all had the expected positive sign, some already revealed problems of size. Still, since size is not sufficient reason to eliminate an indicator, the research team continued the analysis. Second, the standard errors of each indicator in the measurement model were analyzed. These were small for most indicators (suggesting accurate estimation). For items Svl1, Svl6, Svl7, Plso2, Plrf1, Plrf2, Pldc1 and Pldc2, however, large standard errors suggested inadequate fit. In the third step, it was confirmed that the parameter estimates were indeed not statistically significant for the items mentioned, which indicated that they should be deleted from the model (Byrne, 2010). To eliminate these items from the measurement model, the single-step approach recommended by Boomsma (2000) was followed: the items were eliminated one at a time, proceeding from greater to smaller standard error, and then the full measurement model was estimated after eliminating each item. This enabled the researchers to confirm two things: 1) that none of the parameter estimates became significant during the scale refinement process with successive elimination of the items; and 2) that, for each estimation of the full measurement model, the same parameters were obtained, indicating the model's robustness. For reflective constructs, Mackenzie, Podsakoff, and Podsakoff, (2011) argue that it is possible to omit several items without distorting critical elements the construct domain, since the remaining items can register the effect of these constructs. Similarly, drawing on Nunnally and Bernstein (1994), Hair, Hult, Ringle, and Sarstedt (2017) argue that reflective indicators (sometimes known as effect indicators) constitute a representative subset of all items present in the conceptual domain of the construct. Through association with a specific construct, all indicator items should be closely correlated and all individual items exchangeable. Eliminating individual items from trustworthy constructs should thus not generally alter the meaning of the construct.
The following fit indexes were obtained for the full measurement model: Chi square=2576.017; d.f. 1241, p=0.000; CFI=0.951; AGFI=0.841; GFI=0.884; SRMR=0.048; RMSEA=0.042; 90% CI=0.038-0.045. The proposed measurement model had significant goodness-of-fit statistics, but other indexes indicated good model fit. The model fit was evaluated using the two-index strategy proposed by Hu and Bentler (1999). According to their strategy, a model is acceptable if it satisfies one of two pairs of conditions: either that RMSEA ≤ 0.06 and CFI, TLI or RNI ≥ 0.95; or that RMSEA ≤ 0.06 and SRMS ≤ 0.08. In that case, the full measurement model satisfies both conditions, although the GFI and AGFI values are just below the recommended minimum of 0.90.
Other studies (e.g., Foote, Seipel, Johnson, & Duffy, (2005: p. 213) attribute this minimal difference, however, to the "artifact of the large degrees of freedom relative to sample size and the relatively small number of estimated parameters". The last step for evaluation of the measurement model was a specification search (Byrne, 2010). Although the fit indexes reported show good model fit, it might have been possible to improve the fit. The modification indexes provided by EQS (Lagrange test) were thus calculated. The largest modification indexes obtained corresponded to the correlation of the error variances. According to Boomsma (2000), the measurement model should only be modified if the changes make substantive sense. Since the changes detected through the Lagrange test had no substantive meaning, no changes were thus made. Table A1 presents the data obtained in the final full CFA (see Appendix).
Convergent validity of the scales was also evaluated. Scale reliability was assessed through composite reliability (CR), average variance extracted (AVE), and Cronbach's Alpha (Hair, Black, Babin, Anderson, & Tatham, 2010). The constructs' CR values were above the recommended minimum of 0.70 (Bagozzi & Yi, 1988), and the AVEs above 0.5 (see Table A1). These statistics conform to the criteria proposed by Fornell and Larcker (1981). The reliability indicated by the Cronbach's Alpha coefficients, which evaluate the constructs' internal consistency, also exceeded the recommended minimum of 0.80 (Nunnally & Bernstein, 1978). The results presented in Table A1 thus confirm internal consistency and reliability of all scales.
The approach of Voorhees, Brady, Calantone, and Ramirez (2016) was followed to measure construct-level discriminant validity. Firstly, Fornell and Larcker's (1981) procedure comparing the square root of the AVE and the construct correlations was used (the main diagonal in Table 2 shows the square root of the AVE). As all values are greater than the correlations of the constructs, the results confirm discriminant validity among the model constructs. Second, heterotrait-monotrait (HTMT) ratios were determined for each pair of constructs (Henseler, Ringle, & Sarstedt, 2015). The ratios, displayed in Table 3, take values below 0.85 for each pairing, also confirming discriminant validity.

Table 3. Heterotrait-monotrait Ratio
The scales that measure engagement (Schaufeli et al., 2002), empowering leadership (Ahearne et al., 2005), psychological empowerment (Spreitzer, 1995), and paradoxical leadership (Zhang, et al., 2015) have been conceptualized as second-order constructs in the prior literature. Higher-order models should, however, be evaluated critically as compared to alternative lower-order models (Hair et al., 2010). Evaluation here followed the strategy of competitive models in Marsh and Hocevar (1988) and Rindskopf and Rose (1988), and used in Cao and Zhang (2010), Huang, Lai, and Lo, (2012), and Chen, Lu, Chau, and Gupta (2014). Firstly, for each construct, it was specified a first-order onefactor model, constraining all items for each model to load on a single latent, unobserved factor. A second, target model required each scale's items to load on the number of factors defined as dimensions of the construct and all factors to correlate amongst themselves. A third, nearly identical model differed from the second only in replacing the correlation paths between the factors by a shared higher-order factor. Finally, to compare these three alternative models for each construct, the target coefficient was calculated. Marsh and Hocevar (1988) recommend evaluating the explanatory power of second-order models through the target (T) coefficient [T= Chi Square (baseline model)/Chi Square (alternative model)]. The target coefficient indicates how well the higher-order factor model explains covariance in the first-order factors, measured as percentage variation of the first-order factors attributed to the second-order construct (Huang et al., 2012).
Although T-coefficients ranging from 0.80-1.00 show that a second-order construct is present, the second-order model provides a more parsimonious explanation of covariance in the first-order factors (Rindskopf & Rose, 1988). Since the second-order factor alone does not explain the shared variance in the first-order factors, the fit indices of the higherorder model can never be better than the corresponding first-order model (Segars & Grover, 1998). Table A2 shows the target coefficients calculated for engagement (T=1), empowering leadership (T= 0.914), psychological empowerment (T= 0.806), and paradoxical leadership (T=0.947). These values support the conclusion that a secondorder model provides a better structure in all second-order model constructs.

Structural model
The analysis was based on confirmatory modeling, a method that specifies a relationship model constructed from theory and statistically evaluates that model's significance using structural equation modelling (SEM). Again, non-normality of the data (Mardia's coefficient=568.879; t-value=71.546) established by the multivariate normality test led us to choose the robust maximum likelihood ML estimation method (Bentler & Bonett, 1980). In estimating the structural model, empowering leadership, paradoxical leadership and servant leadership were allowed to correlate, since the study goal is to analyze the joint impact of these three types of leadership on psychological empowerment and engagement.
Before assessing fit of the structural model, the second-order constructs must be evaluated by analyzing the factor loadings, the R 2 , CR and AVE of each higher-order construct's dimension. CR and AVE were calculated using the procedure for second-order constructs recommended in Gefen, Rigdon, and Straub (2011) and used in Pavlou and Gefen (2005). In the formula proposed by Gefen et al., (2011: p. 11), "reliability in second-order constructs is calculated as multiplication of the standardized loadings of the first-order factors by the standardized loadings of the second-order factor". The CR and AVE were also calculated using the traditional procedure for reflective second-order constructs specified in Mackenzie et al. (2011).
The results show the validity of all second-order constructs except psychological empowerment, whose dimensions self-determination and impact do not perform well (second-order construct factor loadings below 0.7; their R 2 below 0.5) These factor loadings were thus eliminated, following the guidelines of Jarvis, MacKenzie, and  and Mackenzie et al. (2011) for reflective constructs when a subdimension lacks validity. In fact, some previous researchers using Spreitzer's (1995) scale omitted some of the construct's dimensions from the analysis due to similar problems (e.g., competence, in Albrecht and Andreetta (2011); and self-determination, in Dimitriades (2005) and Jose and Mampilly (2014). Further, multiple prior studies have operationalized psychological empowerment as a single averaged score, disregarding distinctions among its dimensions (Fong & Snape, 2015;Koberg, Boss, Senjem, & Goodman, 1999;Quiñones et al., 2013). According to Edwards (2001), this operationalization procedure involves two assumptions: a) since the different dimensions express the reflective second-order construct to the same extent, changing the construct's units changes each dimension to the same extent; b) each dimension has the same quality in indicating the reflective higher-order construct. Based on Edwards's (2001) premise, distinctions among dimensions can thus be disregarded, such that the larger numbers in the remaining dimensions signify higher levels of psychological empowerment as a whole.
Confirmation of the hypotheses demands analysis of the proposed model's global fit. There are two main classes of fit statistics: 1) model test statistics, and 2) approximate fit indexes. Each class assesses model fit differently (Kline, 2011). The model was evaluated using both types of fit statistics: First, the model test statistics (or Chi-square test) were calculated. Hayduk, Cummings, Boadu, Pazderka-Robinson, & Boulianne (2007), (Boomsma, 2000), and Kline (2005Kline ( , 2011 have reviewed the importance of the result of this test in evaluating the structural model. The following result in the model was obtained: Chi square= 2412.825; d.f.= 1149 p =0.000 (see Table A3 in the Appendix). The exact-fit hypothesis at the 0.05 level was thus rejected. Since the discrepancy between the observed and the model-implied covariances is statistically significant, the model fails the Chi-square test. Interpreting this result required taking into account that: 1) the Chi-square test assumed normal distribution of the data (an assumption that this data did not fulfill, based on the result of Mardia's test) (Kline, 2005;Hayduk et al., 2007); 2) the test registers correlation magnitude (a larger correlation could lead to rejection of the null hypothesis [Kline, 2005]); and 3) that "the p values for test statistics are also estimated in sampling distributions that assume random sampling from known population. (In fact) random sampling is a crucial part of the population inference model" (Kline, 2011: p. 195). The data come from a convenience sample that does not fulfill the assumption of multivariate normal distribution, and some variables show relatively high correlations, which could explain the unexpected result of this test.
A significant Chi-square p-value does not necessarily invalidate the model (Kline, 2011). It does, however, require diagnosing why the test failed. One can diagnose this reason by recording and describing the correlation residuals, attending especially to absolute values >0.10 (Kline, 2011). EQS provides the bulk of these residuals, showing that 92% have values ranging from -0.1 to 0.1, which seems to indicate the absence of misfit. The parameters whose residuals take the absolute values >0.10 correspond to parameters of the relationship of the items measuring servant leadership to empowering leadership, empowering leadership and paradoxical leadership; and between engagement and psychological empowerment. Since the vast majority of residuals is shown to be positive, the model under-predicts the associations observed (Kline, 2011).
The second way of evaluating model fit is through the approximate fit indexes (Kline, 2005(Kline, , 2011. The SEM literature contains many indexes of model fit (Kline, 2005), and one challenge of SEM research is to determine which indexes must be reported in studies that use this methodology. In this case, the recommendations in Kline (2005Kline ( , 2011 were followed, which advise reporting a set of four approximate fit indexes. Each index takes a different perspective from which to describe model fit: a) RMSEA, with its 90% confidence interval, and whose scale measures poorness of fit, with 0 representing perfect fit (it is a parsimony-adjusted index); b) Jöreskog and Sorbom's (1982) absolute fit index, GFI (Goodness of Fit Index); c) Bentler's (1990) Comparative Fit Index (CFI), measuring incremental fit; d) Standardized Root Mean Square (SRMS). In contrast to the foregoing indexes, is related to the correlation residuals.
In addition to the foregoing, Gefen et al. (2011) recommend reporting the AGFI (see Table A2 in the Appendix). Hu and Bentler (1999) assess fit of the model using a strategy involving two indices. According to their strategy, a model is acceptable if it satisfies one of two pairs of conditions: either that RMSEA ≤ 0.06 and CFI, TLI or RNI ≥ 0.95; or that RMSEA ≤ 0.06 and SRMS ≤ 0.08. On the basis of these guidelines, this model does not satisfy the first pair of conditions but does satisfy the second, since SRMS=0.057 and RMSEA= 0.044. So, the results suggest that the model fit was acceptable, although it does not obtain a CFI value of 0.95. Recall, however, that the standardized residuals indicate some degree of misfit with respect to servant leadership, empowering leadership, and paradoxical leadership.
Further, Boomsma (2000) indicates that evaluation of the structural model must also take into account the squared multiple correlation coefficients' (R 2 ) magnitudes, which indicate each equations' fit individually. Figure 3 reports each equation's R² plus the size and significance of the estimated coefficients. The data match the theoretical predictions.
Based on Riedl, Kaufmann, and Gaeckler (2014), it was adopted the method in MacCallum, Browne, and Sugawara (1996) for determining the statistical power of the structural equations model. A result of 1 (for ɛ0 =0.05; ɛa=0.08), within the acceptable range of 0.8-1, confirmed reliability and validity. This high statistical power could explain the failure of Chi Square test because higher power increases the probability of registering discrepancies between the observed and implied covariance matrices. That is, high power increases the risk of rejecting correct models (Riedl et al., 2014). This result provides more evidence in favor of the research model. The criteria used to assess the structural model lead to the conclusion that the equations from which the model is constructed represent the data well.

Hypothesis Testing
The estimation values and respective significance levels for the standardized coefficients of parameters representing each of the hypotheses are displayed in Figure 3.

Figure 3. Structural Equation Model
Hypothesis 1 anticipated that the manager's empowering leadership style has a positive association with psychological empowerment. Hypothesis 2 predicted that the manager's empowering leadership relates positively to workers' engagement. Hypotheses 1 and 2 both received empirical support. Specifically, Figure 2 presents the significant and positive direct effects of empowering leadership, on psychological empowerment (β = .389, p <0.01) and job engagement (β = .270, p <0.01) respectively.
The results also indicated a positive association of paradoxical leadership with psychological empowerment (β =.173, p<0.01), supporting Hypothesis 3. Nevertheless, the expected association of paradoxical leadership with engagement turned out to be nonsignificant, providing no support for Hypothesis 4.
Hypotheses 5 and 6 predicted that servant leadership would exert a positive influence, respectively, on psychological empowerment and engagement. Hypothesis 5 was rejected, whereas the results support H6. Specifically, servant leadership does not affect employees' feeling of psychological empowerment significantly but exerts a positive effect on engagement of employees (β = .165, p <0.01).
To summarize, this model is suitable to predict the effects of different leadership styles on engagement due to the relatively high explained variance obtained for engagement (R²= .734). This engagement variance is explained by empowering leadership (β =.270), servant leadership (β = .165), and psychological empowerment (β = .633). Results also reveal a moderate explained variance of psychological empowerment (R² = .201), obtained thanks to the effects received from paradoxical leadership (β = .173) and empowering leadership (β = .389). Although there is no influence of servant leadership on psychological empowerment, this leadership style still affects job engagement directly and significantly (β = .165).
The SEM results indicate the positive relationship between psychological empowerment and engagement predicted, with a positive, significant path coefficient from psychological empowerment to engagement (β=.633, p<0.01), supporting Hypothesis 7.
Mediation analysis (test of hypotheses 8 to 10) To improve interpretation of the relationships hypothesized in the model, Zhang and Bartol (2010) advise conducting decomposition of effects, as this procedure provides fuller understanding of the direct and indirect effects. The direct, indirect, and total influence of each leadership style on engagement was thus assessed, following Zhao, Lynch Jr., and Chen (2010). In addition, 95% bias-corrected confidence intervals were calculated and bootstrapping (5000 resamples) performed, as recommended in Preacher and Hayes (2008) and Hayes (2013), using SPSS v.22. Zhao et al. (2010 and Hayes (2013) consider the mediating effect as significant when confidence intervals for the indirect effect do not include zero.

Table 5. Mediation Analysis
"Specific indirect effects represent the portion of the total effect that works through a single intervening variable" (Zhang et al., 2010: p. 118). For coefficient (β), effects decomposition yielded a value of .246 (p <0.05) for empowering leadership's indirect influence on engagement at work, via psychological empowerment. This effect accounted for 47.67% of empowering leadership's full effect, as hypothesized in H8 (Table 5). Table  5 also displays paradoxical leadership's indirect effect on job engagement through psychological empowerment, which is positive and significant (indirect effect β = .109, t = 2.103, p <0.05), confirming hypothesis 9. There is no evidence, however, that servant leadership influences engagement indirectly via psychological empowerment, and thus no evidence to support H10.

DISCUSSION AND CONCLUSIONS
Hospitality workers must frequently take prompt action at the frontline level to provide good service. For this reason, the inability or unwillingness of customer-contact employees to respond quickly to service contingencies may result in unsatisfactory service encounters. Empowerment represents a promising solution in hospitality organizations, since it empowers employees, enabling them to make decisions to achieve successful service recovery and, in turn, satisfaction, and retention of customers. In hotel settings, providing outstanding service sometimes indeed requires departing from the script. In line with Baum's (2015) recommendations, companies should empower their front-line workforce to make decisions independently, without deferring to supervisors and managers. The present research findings especially advocate empowering and paradoxical leadership behaviors for employees to experience psychological empowerment in hotel settings. This study's findings also advance the literature by showing (in the context of hospitality) that psychological empowerment clearly precedes job engagement. Only a few studies (Jose & Mampilly, 2014;Macsinga et al., 2015;Quiñones et al., 2013) anticipated this positive association: that employees who experience workplace significance, skill, autonomy, and efficacy-in other words, psychological empowerment, are more engaged. Alfes et al. (2013) find that engaged employees are intensely engrossed in their work, exerting more energy and enthusiasm and excelling in service. Many practitioners and organizations in the service sector are thus investing great effort in finding ways to increase their employees' engagement. According to the study results, the employee's gender is not a determining factor in his/her experience of engagement, but education level is. It seems logical that more educated employees in hospitality are more skilled when performing their jobs and tend to feel more pride and enthusiasm about their own performance.
This study suggests that achieving engaged employees in hotel businesses requires managerial tools that foster psychological empowerment. In this line, empowering leadership is found to be a key leadership style for hospitality managers, given its strong and positive influence on customer-contact staff. Employees who identify their managers as employing an empowering leadership style are more likely to feel engaged and empowered at work. These findings reinforce those of prior research (Albrecht & Andreetta, 2011;Tuckey et al., 2012). Staff-enabling actions such as providing autonomy, encouraging employee participation, expressing confidence in workers' performance, and helping employees to understand the importance of their work are essential actions for leaders to perform, not only to enhance psychological empowerment but also to increase engagement levels among hospitality workers. People are so diverse that every service encounter differs. Yet frontline workers can respond to the unexpected when management empowers their judgment, expertise, and discretion in service delivery. This latitude and responsibility will make employees feel more capable, more enthused, and more dedicated at work-in sum, more engaged.
Furthermore, like the findings of De Klerk and Stander (2014), the partial mediation results in this study confirm the indirect effect of psychological empowerment in the relation of empowering leadership to job engagement. Enhancing managers' empowering leadership behavior thus produces greater psychological empowerment and subsequently job engagement in hospitality employees. Leaders who empower followers through increased responsibility, access to information and power, and reward for autonomous action heighten follower self-efficacy, increasing their employees' level of engagement.
Evidence was also found for the positive relationship of paradoxical leadership to psychological empowerment. Managers following paradoxical behavior leadership approach grant autonomy while simultaneously preserving decision control. When managing their staff, such managers maintain both distance and closeness, achieving a perfect combination of self-centeredness and other-centeredness to share the leadership role in some aspects of the daily work. In this way, employees believe they have some flexibility to resolve work issues using their creative personal judgment.
Contrary to expectations, no relationship was found between employee engagement and paradoxical leadership among hotel staff surveyed: employees do not feel "directly" engaged when their managers display this type of paradoxical leadership behavior; paradoxical style only engages them when they experience psychological empowerment first. This full mediation -in terms of the Baron and Kenny (1986) approach-explains that hotel employees feel trusted and empowered when leaders engage in paradoxical behavior, but that, at the same time, some ambiguous behaviors associated with paradoxical leadership may raise suspicion among employees rather than engage them. This result can be explained because, traditionally, hospitality organizations in Spain have had very hierarchical and autocratic structures. In such tightly structured contexts, unconventional approaches to work problems are not popular. In fact, paradoxical leadership style is more suited to organic workplace structures than to workplaces with mechanistic processes, characterized by higher-level authorities (Zhang et al., 2015). Consequently, it is possible that the employees surveyed in this study are not yet ready cognitively to identify the benefits of paradoxical mindsets in their leaders. On the contrary, some employees may even perceive those paradoxical behaviors as ambiguous management moves and thus be confused about their superiors' real intentions.
Despite its exploratory nature, this study also produced some compelling findings from its analysis of servant leadership's connection to job engagement: belief that one's manager is a servant leader increases likelihood of engagement in one's job. This finding can be explained by drawing on Liden and colleagues (2015), who note that servant leaders normally stimulate strong relationships with followers and encourage employee enthusiasm and dedication. Moreover, servant leaders motivate their followers by helping them to realize their full potential, as well as by paying attention to their needs. Employees frequently see this type of leader as a role model and normally display positive and ethical behaviors at work in return (Greenleaf, 1970).
This study did not find a significant relation of servant leadership to psychological empowerment among hospitality employees. This finding diverges from Correia and Van Dierendonck (2014), who conclude that servant leaders enable workers to participate in organizational processes and to become agents of change, inducing greater levels of psychological empowerment. One possible explanation for this difference is that what constitutes an effective style of leadership differs by sector, culture, and momentum of the organization. For example, Humphreys (2005) postulates that servant leadership is better suited to stable contexts and less effective in periods of renewal and change. Contrary to the ideal servant leadership scenario, Spain has in recent years been characterized by turbulence and fluctuations due to economic crisis. In this context of uncertainty and frequent layoffs, servant leaders who prioritize employees' needs over their own are surely a minority; or they may on some level avoid fostering psychological empowerment fully among their employees so that they themselves become indispensable in the company. Further, cultural contexts like Spain, with higher levels of power distance (Hofstede, Hofstede, & Micheal, 2010), could provide a less fertile environment for employees to feel empowered by the servant leadership style, especially in traditionally hierarchical sectors such as hospitality, where pyramidal charts dominate HR configurations. One last explanation for this lack of connection between servant leadership and psychological empowerment could be that, despite servant leaders' empowerment of their employees, bureaucratic constraints or even psychological barriers (culture) persist in the workplace that limit the staff's ability to perceive that their servant managers are in fact empowering them when displaying this type of retainer leadership. Servant leaders in this study were observed to be good at fulfilling some needs of their followers but not at effectively empowering their workforce, as shown by the absence of evidence that psychological empowerment mediates the relationship between servant leadership and engagement.
The findings of this research provide both scholars and practitioners with empirical proof that empowering, paradoxical, and servant leadership are valuable in producing workforce empowerment and engaged employees.

Theoretical implications
This study answers the call for specific research to determine which aspects of management practice are most effective in fostering positive attitudes and behavior in employees in the 21 st -century hotel context.
Whereas traditional leadership styles such as transformational leadership have been widely studied in many sectors, only a handful of studies have gathered evidence to assess how newly identified leadership types affect service delivery. For example, exceptionally little research explores how servant leadership affects hospitality (e.g., Ling, Lin, & Wu, 2016;Wu, Tse, Fu, Kwan, & Liu, 2013), when one considers that the foundation of the hospitality industry is the culture of serving others. Moreover, ours is the first study to apply the recently-developed paradoxical leadership scale (Zhang et al., 2015) in a Western country. In addition, whereas empowering leadership has been posited as one of the best leadership approaches for managing service workers, most prior research has been conducted outside Europe (e.g., Albrecht & Andreetta, 2011;Bester et al., 2015;De Klerk & Stander, 2014;Lorinkova & Perry, 2014;Zhang & Bartol, 2010). This study advances prior scholarship on leadership by stressing the importance of innovative leadership methods to increase more positive states of mind in Spanish hotel employees.
Further, the concept of employee engagement is rather new and is thus an emerging construct in academia (Saks, 2006). This investigation also helps to identify antecedents of job engagement in one service sector-hospitality-where determined, contented employees make a significant difference in performance of the organization. The study can also increase scholars' comprehension of how psychological empowerment relates to engagement.
Finally, SDT helps to interpret the study's findings on improving employees' psychological empowerment and engagement , in accordance with Jose and Mampilly (2014) and Menguc, Auh, Fisher, and Haddad (2013). Some leadership styles favor the satisfaction of two basic human needs: self-determination and competence. Hence, managerial support encouraging employees' participation and autonomy can produce greater levels of engagement in their subordinates by increasing feelings of power and control over their jobs.
Building on the recent positive psychology paradigm (Seligman & Csikszentmihalyi, 2014), the study results recommend further studies on the ways different leadership styles produce varying positive results to enhance employee well-being.

Practical implications
As employees play a decisive role in service encounters, it is imperative to learn ways to motivate and engage them so that they display positive attitudes and behaviors that result in outstanding performance. The present study outlines various practical implications for hospitality managers by identifying leadership styles that foster positive behavior in followers.
Trends in hospitality research and practice show the increasing importance attributed to innovation in styles of leadership. For Liden et al. (2014Liden et al. ( : p. 1447, "as employees' education levels increase, autocratic leadership approaches will no longer be tolerated." Contemporary service employees prefer collaborative, people-oriented leadership styles that mesh with what companies themselves expect from them when attending to customers, and managers must be exemplary role models. Managers must, therefore, understand not merely how their behavior influences intermediate organizational outcomes but also which of their behaviors produce each effect (Namasivayam et al., 2014).
Unlike leadership approaches with a top-down hierarchical tone, servant, empowering, and paradoxical leadership styles emphasize trust and collaboration and enhance employee self-efficacy. This research thus outlines how servant and empowering leaders can influence employee engagement significantly. Servant leaders are especially beneficial to service organizations such as the hospitality industry, because servant leaders focus on power sharing and satisfaction of their followers' needs, even placing others' interests ahead of their own. Servant leaders serve as an example of orientation towards others. The research of Liden and colleagues (2015) confirms a major tenet in the theory of servant leaders: that servant leaders' subordinates follow the example of their supervisors' ethical behavior, sometimes giving priority to the needs of others above their own. Since servant leadership can permeate the whole organization (Hsiao et al., 2015), this kind of leadership is especially advocated for the hospitality industry, where "serving culture" is the prime principle.
The study results urge hospitality managers to adopt stronger participatory styles of leadership, particularly behavior that makes employees feel empowered and engaged psychologically. Organizations in this sector must encourage their managers to enable followers by encouraging them to take responsibility and assert autonomy independently of "always getting a stamp of approval" (Tuckey et al., 2012). But such empowerment is not only delegation: the organization must both give employees the resources necessary to make such decisions and reward them for doing so.
Nevertheless, acceptance of empowerment by first-line managers and frontline employees is not a sure thing, as Karatepe (2013) highlights. In a traditionally hierarchical sector like hospitality, it is not an easy task to unanimously accept employees' freedom of choice. On the one hand, employees may perceive that they are being given more responsibilities for the same salary, or even see the empowering philosophy as demagogic or as mere lip service (Biron & Bamberger, 2011) with no real authority delegated. On the other, managers frequently do not fully understand, embrace, or encourage empowerment because they do not even know how to delegate authority and power. In both cases, extensive training programs should be implemented. The empowerment process should in any case be accompanied by a proper rewards system, not only economic but also emotional. On the practical level, hospitality companies must train the workforce at all levels of the hierarchy: 1) managers in successfully delegating power, and 2) employees in accepting more responsibility and authority to act in challenging service encounters.
Finally, and of equal importance, this study has interesting outlines for educators. As Brownell (2010) stresses, it is important to reflect growing trends in leadership curricula. For example, appropriate measures by professors can inculcate principles of modern leadership in students by using pedagogical tools that help learners adopt paradoxical thinking and servant behaviors, such as, for example, role-playing. Case studies of successful organizations run under the empowerment philosophy, such as Semco (Semler, 1989) or Ritz Carlton (Michelli, 2008), can also make people rethink their mental frameworks.

Study Limitations and Future Lines of Research
The limitations suggest issues for subsequent investigation.
First, the study is cross-sectional, evaluating participants in specific situations at a specific moment in time, while participants are likely to change their perceptions over time. Caution must thus be exercised when using the results to draw conclusions about causal relationships. Future studies could "perform longitudinal surveys to conduct cross-period research and collect long-term data on organizations" (Hsiao et al., 2015: p. 54).
Second, as one cannot be certain that convenience sampling accurately represents workers in Spanish hotels, caution is necessary when making generalizations to the population as a whole. Further studies must improve the generalizability of the results by ensuring representative samples through random sampling. Similarly, a larger sample size is advisable, as mentioned in the methodology section.
Third, although the model was not modified based on specification searches, changes were made in the measurement instruments (some items of different constructs were removed) that could correspond to specific characteristics of the sample. Whether these changes can be generalized to different samples or the general population thus remains to be confirmed in further research, and the resulting model should be evaluated using independent samples (MacCallum, Roznowski, & Necowitz, 1992). This limitation is mitigated, however, by the exploratory nature of this study, which admits the changes described above (MacCallum et al., 1992). Similarly, the remedy prescribed by MacCallum et al. (1992)-constructing competing models, as in the case of the multidimensional constructs-helps to avoid capitalizing on chance.
Fourth, the self-report measures have both benefits and limitations. While they are crucial to gathering information on employee perceptions-information that managers cannot provide-such single-source data is vulnerable to common method bias, as bias due to self-interest may distort some findings. The questionnaire was constructed, however, to minimize this distortion, and the results suggest that method bias is not a problem.
However, data from multiple sources is usually desirable. Future multilevel investigations are advised to reveal the leaders' vision of their own behaviors (Wong, 2016).
Further, the boundary conditions must be determined within which leaders' influence obtains optimal outcomes, and this requires exploration of factors moderating connections between engagement and leadership styles. For example, as Barnes and Collier (2013) hypothesize, the kind of relationship that managers and employees have developed seems likely to influence employees' level of engagement. It could thus be interesting to include a theoretical approach based on Leader-Member Exchange (LMX, Graen & Uhl-Bein, 1995), as do Garg and Dhar (2016) and Luo, Wang, Marnburg, and Øgaard (2016), in the proposed model.
On the other hand, it is necessary for certain contextual and personal factors to converge for employees to experience positive feelings at work. Prior research has identified personality factors as relevant to psychological empowerment (Kim, Shin, & Swanger, 2009) and job engagement (e.g., Macsinga et al., 2015). Employees' perceptions and experiences in the workplace could also be influenced by their personalities. Future research could incorporate the Five-Factor personality model in the proposed model to provide a more comprehensive portrait of critical factors influencing empowerment and engagement of hospitality employees.
These limitations appear not to undermine the study's contributions in advancing understanding of the role of leadership in hospitality management. The research does illuminate part of the relationship of emerging leadership styles to employee psychological empowerment and engagement. Future studies could build on this investigation of empowering, paradoxical, and servant leadership by investigating additional styles, for example, authentic (Walumbwa, Wang, Wang, Schaubroeck, & Avolio, 2010) or charismatic leadership (Babcock-Roberson & Strickland, 2010). This study thus opens a stream of potential research that seeks to understand how other leadership behaviors result in employee wellbeing in hospitality.