Implementing Sustainable Human Resources Practices: Leadership Style Matters

: Findings of a positive relationship between high-performance work systems (HPWSs) and organizational performance indicate that an investment in a set of well-conﬁgured HR practices can promote strategic organizational goals. However, recent strategic human resource management (SHRM) research indicates that the positive link between HPWSs and performance might not hold universally across organizations because of poor implementation of the adopted HR systems. Drawing on leadership literature, we address this implementation problem by focusing on the moderating e ﬀ ect of frontline managers’ three leadership styles on the association between HPWSs and employees’ perceived support for innovation. Data collected in mainland China (429 employees on 66 work teams in 14 ﬁrms) showed that some leadership styles had signiﬁcant moderating e ﬀ ects on that association and, subsequently, on team e ﬀ ectiveness. The ﬁndings contribute to the SHRM literature by demonstrating the importance of frontline managers’ leadership styles in the e ﬀ ective implementation of HPWSs to promote organizational innovation and team e ﬀ ectiveness.


Introduction
Does people management matter? To find support for the link between HR systems and firm performance, scholars in the field of strategic human resource management (SHRM) have conducted more than 90 empirical studies during the last two decades (see recent meta-analytic reviews, e.g., [1][2][3]). These studies have shown that the use of high-performance work systems (HPWSs) is positively associated with various dimensions of operational and financial performance across firms (e.g., [4][5][6][7][8][9][10][11]). One important consequence of the use of HPWSs is the existence of heightened levels of organizational capacity to pursue innovation (e.g., [12]). This finding is important because organizational innovation is becoming increasingly relevant in the current economy given that innovation and creativity have been noted as critical factors for organizational success [13][14][15].
Although prior SHRM research has shown a positive association between HPWSs and organizational innovation, the mechanism of this relationship needs better explanation. In recent empirical studies (e.g., [16][17][18]), employee-perceived HR practices have been shown to be substantially different from manager-rated HR practices. This raises concerns about whether HR practices are being effectively implemented within organizations. Whether consciously or unconsciously, it has been assumed that well-designed HR systems will result in enhanced performance because effective implementation of the HR systems will automatically follow, which might not be true. Thus, it is necessary to examine the positive association between HPWSs and organizational innovation in terms of HR system implementation.
To address the issue of HR practice implementation, we focus on the role of frontline managers' leadership in implementing HPWSs. Although the critical roles of frontline managers in implementing HR practices have been observed in previous studies (e.g., [19]), those studies' findings are limited. That is because these studies do not provide insights into specific mechanisms of how frontline managers' leadership can operate as a boundary condition in conveying the effect of HPWSs on organizational outcomes such as innovation. Moreover, the impacts of different leadership styles in implementing HR practices remain unknown. By drawing on leadership literature, we examine whether and how HPWSs and team leaders' leadership style jointly affect subordinates' perceptions of innovation support and, ultimately, team effectiveness.
We suggest the moderating effects of the three leadership styles (transformational, transactional, and laissez-faire leadership) [20], and analyze data collected from 429 employees of 66 work teams in 14 firms in mainland China. By doing so, this study attempts to make three contributions to the SHRM and innovation literature. First, it provides theoretical mechanisms of how the three leadership styles of frontline manager operate in HR system implementation and offers empirical findings. Second, this study both identifies and emphasizes employee perceptions of organizational innovation support as one pathway through which HR systems affect performance. Third, the study provides empirical clues for the relative impact of the three leadership styles, for further examination and future research.

The Effects of HPWSs on Innovation and HR Practices Implementation
SHRM research is focused on examining the impact of HR systems on various dimensions of firm performance. Based on the resource-based view of a firm [21], which finds a source of sustainable competitive advantage in an organization's internal resources, SHRM scholars contend that HR systems that are internally consistent, and connected with firm strategies, lead employees to contribute to achieving a firm's strategic goals [5,8,[22][23][24]. Among SHRM scholars, there is no concrete agreement about which HR practices or polices to include in the HR systems [25] and, thus, how to label them. However, it is generally acknowledged that HR systems that are designed to obtain and develop high-quality employees, motivate them, and provide them with opportunities to contribute are associated with higher organizational performance. These HR systems are often called HPWSs, high involvement work systems, or high commitment work systems [26,27]. The positive association between HPWSs and organizational performance has appeared in many previous SHRM studies (e.g., [1][2][3]28]).
As one plausible mechanism for the positive association between HPWSs and organizational performance, the relational model of HRM [29] offers explanations of how HPWSs promote employee perceptions of innovation support. When a firm utilizes HPWSs that emphasize new knowledge, learning and development opportunities, formal performance appraisal and feedback, employee empowerment and participation, and team-based work, employees are more likely to perceive that their firm is encouraging and supporting creative behaviors and innovation. This is because the use of HPWSs promotes employee communication and fosters shared goals, shared knowledge, and mutual respect and trust among employees, resulting in enhanced organizational effectiveness [12]. For example, Chang et al. [30] provide empirical evidence supporting the link between HPWSs and innovation in workplaces. They analyze data from high-technology firms in China and report that HPWSs positively affected employees' individual creativity.
Although previous studies have found this positive link between HPWSs and innovation, this relationship might not be tenable in different organizational contexts with poor implementation of adopted HPWSs. According to the study by Bowen and Ostroff [22] and Ostroff and Bowen [31], this relationship will appear only when employees perceive that adopted HPWSs are effectively administered.

Frontline Managers' Leadership Style as a Boundary Condition
We believe that this issue of HR practice implementation is highly likely to be influenced by the behaviors of frontline managers who are the key implementers of HR practices. Several studies have specifically noted that the implementation of HR practices is largely dependent on the roles of frontline managers (e.g., [18,32,33]). A study by McGovern et al. [34] found that in implementing HR practices, the motivation of the supervisor was the most influential factor over others such as targets, company values, and career advancement opportunities. Supervisors can create the work context that is needed to nurture creative ideas and innovative behaviors. This can be done by using a set of high-performance work practices to promote open communication, effective distribution of needed resources, and trust building among team members. In this sense, a supervisor's leadership behaviors are expected to be key to the successful implementation of HPWSs.
Although we believe that a frontline manager's leadership behaviors may have a significant effect on the implementation of HR practices, the kind of leadership behaviors that matter and also how they matter are largely unknown. For example, some leadership behaviors can focus on suggesting a strong vision for the organization, whereas others can focus on developing transactional relationships with subordinates. For this reason, understanding the leadership behaviors required for effective implementation of HR practices and distinguishing them from other leadership behaviors are important [19,33].
To gain useful insights into this matter, we focus on three types of leadership behaviors: transformational, transactional, and laissez-faire [35,36]. Transformational leadership (TFL) suggests that transformational leaders proceed beyond exchanging inducements for the desired performance of subordinates, encouraging them to achieve above and beyond normal performance expectations [35,37,38]. Research findings show that by displaying four behavioral components (inspirational motivation, idealized influence, intellectual stimulation, and individualized consideration), TFL becomes a source of a positive influence on multiple organizational outcomes. For example, a study of Howell and Avolio [39] found that TFL of managers was positively associated with business-unit performance in Canadian financial companies. Further, Shin and Zhou [40] showed that supervisors' TFL was positively associated with creativity of subordinate R&D workers in Korean companies, and Gumusluoglu and Ilsev [41] reported that TFL of owner-managers was positively associated with subordinates' creativity and organizational innovation in Turkish software development companies.
Transactional leadership (TXL) is based on the contractual exchanges between leaders and followers [37,42]. This leadership style includes three distinctive behavioral components: contingent reward, active management by exception, and passive management by exception (i.e., the passive use of corrective actions). The use of TXL indicates that followers agree about the use of inducements or avoidance, such as praise, rewards, resources, and disciplinary actions, in exchange for their work [35,43]. In a study of U.S. Army platoons, Bass et al. [42] showed that TXL of platoon leaders positively predicted unit performance. When it comes to knowledge sharing and innovation, however, prior research results are mixed. In a study of a government agency in the Netherlands, Pieterse et al. [44] reported that TXL had a negative association with innovative behavior when subordinates' psychological empowerment level is high while TFL had a positive impact under the same conditions. On the other hand, Analoui et al. [45] showed that TXL, as well as TFL, enhanced knowledge management activity in UK organizations and Masa'deh et al. [46] reported that TXL positively affected knowledge sharing and job performance in a Jordanian government agency.
Laissez-faire leadership (LF) is a leadership style that includes virtually no guidance to subordinates [47,48]. Leaders with this leadership style do not have confidence in their ability to supervise. For that reason, they avoid interacting with subordinates and bury themselves in paperwork. Consequently, they leave heavy responsibility to their subordinates, fail to set clear goals, and avoid making decisions [48][49][50][51]. While the number of prior studies on the effect of LF is largely limited when compared with studies on TXL and TFL, the dominant view of prior research on LF has considered LF detrimental to organizational functionality [52]. For instance, Judge and Piccolo [51] found that LF was negatively associated with leader effectiveness and satisfaction with the leader, and Skogstad et al. [53] reported that LF was positively associated with subordinates' role conflict, role ambiguity, conflicts with coworkers, exposure to bullying at work, and psychological distress in Norway. On the other hand, however, there are a few recent studies that began to focus on the positive side of LF. Yang [52] proposed that LF could be a strategic choice of a leader and that LF may reduce subordinates' dependency and promote subordinates' self-determination and autonomous motivation. She argued that LF is similar to empowering leadership in that non-involvement of LF can be seen as deference to subordinates' abilities and elimination of bureaucratic constraints.
As shown above, many leadership scholars have conducted a remarkable number of leadership studies, generating useful implications in terms of both theory and practice (e.g., [43,51,54,55]). Research findings show that the effects of these leadership styles on organizational process and outcomes differ in magnitude, positively increasing from LF, TXL, and TFL, in that order. However, the efficacy of leadership is not invariable; instead, it is subject to change when coupled with different organizational practices and context [56]. With respect to this study, we believe that each of the three leadership styles can provide an effective organizational context to understand the impact of HPWSs in terms of HR practices implementation.

Hypotheses
The focus of this study is not only theorizing and testing the mechanisms of how the three leadership styles (TFL, TXL, and LF) can help organizations implement HPWSs effectively but also exploring empirical clues related to the relative impacts of each leadership style on HPWS implementation. For these goals, we tested several hypotheses and one research question (without theoretical predictions of the empirical results). The conceptual model of the study is shown in Figure 1.  It is known that TXL is likely to provide more effective organizational context in fostering performance when coupled with well-defined organizational and job contexts than with turbulent work environments [36,[65][66][67]. Organizations using HPWSs usually have relatively well-defined organizational contexts (i.e., human resource management systems such as rigorous selection procedures, regular performance evaluations, and intensive training). Indeed, the findings of many SHRM studies are typically based on the analyses of large organizations, such as mini-mills (e.g., [63,68]), auto manufacturers (e.g., [23]), banks (e.g., [69]), and large technology firms (e.g., [12]). Based on these findings, we expect that TXL can be an appropriate leadership style to use when organizations implement adopted HR practices.

Hypothesis 1: TFL will moderate the association between HPWSs and employees' perceived support for innovation such that the association between HPWSs and employees' perceived support for innovation will be stronger under high levels of TFL.
The core idea of TXL is to clarify and administer organizational rules and policies regarding contractual exchanges between leaders and followers [35,43]. Under this leadership, followers develop a clear understanding of organizationally expected work behaviors and subsequent returns for providing (or not providing) those behaviors. In other words, the central function of transactional leaders is to communicate organizationally intended messages-which are constructed with the use of their HR practices (i.e., HPWSs in the current study)-by displaying transactional leadership behaviors. We believe that this will reinforce the positive impact of HPWSs on perceived support for

Moderating Effects of Transformational, Transactional, and Laissez-Faire Leadership
Previous studies have found that TFL has a positive impact on creativity and organizational innovation (e.g., [57,58]). TFL primarily influences followers' creativity and pursuit of innovation in two ways: the promotion of creative ideas and the escalation of intrinsic motivation (e.g., [58,59]). Transformational leaders promote followers' creative ideas by encouraging them to think "out of the box" through the association of distal concepts or ideas and switching thinking patterns [60]. This intellectual stimulation of transformational leaders can facilitate the adoption of these non-traditional thinking processes. Maintaining strong motivation to pursue innovation is also important to implement ideas generated from creative thinking. Transformational leaders escalate followers' intrinsic motivation by articulating an appealing vision and mission to link followers' identities with their organizations' collective identity [35,38,61,62]. Moreover, individualized considerations help leaders understand problems and difficulties from the viewpoint of followers, which can motivate subordinates by making the articulated goals look more achievable. Finally, supervisors become a respectable role model by displaying innovation-oriented behaviors.
However, the effect of TFL on organizational innovation and employee creativity is actually very similar to that of HPWS. As described above, the use of HPWSs can also become a source of employees' perceived support for innovation because it promotes the generation and implementation of new ideas through intensive employee interactions, participation and development. We expect to observe a positive synergy between HPWSs and TFL in promoting perceptions of innovation support. The synergy argument found in SHRM research is particularly relevant to the current interaction argument. The SHRM literature suggests that a bundle of internally consistent organizational policies related to the management of employees can reinforce each other's positive impact [8,23,63,64]. Thus, it is expected that TFL behaviors displayed by supervisors can reinforce the positive impact of HPWS on employees' perceived support for innovation. Specifically, we believe that the impact of HPWSs on employees' perceived support for innovation would be more positive under higher levels of TFL. Hypothesis 1: TFL will moderate the association between HPWSs and employees' perceived support for innovation such that the association between HPWSs and employees' perceived support for innovation will be stronger under high levels of TFL.
It is known that TXL is likely to provide more effective organizational context in fostering performance when coupled with well-defined organizational and job contexts than with turbulent work environments [36,[65][66][67]. Organizations using HPWSs usually have relatively well-defined organizational contexts (i.e., human resource management systems such as rigorous selection procedures, regular performance evaluations, and intensive training). Indeed, the findings of many SHRM studies are typically based on the analyses of large organizations, such as mini-mills (e.g., [63,68]), auto manufacturers (e.g., [23]), banks (e.g., [69]), and large technology firms (e.g., [12]). Based on these findings, we expect that TXL can be an appropriate leadership style to use when organizations implement adopted HR practices.
The core idea of TXL is to clarify and administer organizational rules and policies regarding contractual exchanges between leaders and followers [35,43]. Under this leadership, followers develop a clear understanding of organizationally expected work behaviors and subsequent returns for providing (or not providing) those behaviors. In other words, the central function of transactional leaders is to communicate organizationally intended messages-which are constructed with the use of their HR practices (i.e., HPWSs in the current study)-by displaying transactional leadership behaviors. We believe that this will reinforce the positive impact of HPWSs on perceived support for innovation. As Bowen and Ostroff [22] and Ostroff and Bowen [31] argue, maintaining consistency between organization-declared HR systems and their actual use by the supervisor strengthens the effect of HR systems. Thus, the impact of HPWSs on perceived support for innovation would be larger under higher levels of TXL. Hypothesis 2: TXL will moderate the association between HPWSs and employees' perceived support for innovation such that the association between HPWSs and employees' perceived support for innovation will be stronger under high levels of TXL.
Findings from studies on LF indicate the somewhat negative effects of this leadership style. However, the impact of LF might not always be negative. We believe that positive aspects of this leadership style can be used to strengthen the positive association between HPWSs and employees' perceived support for innovation because LF can offer appropriate organizational contexts to promote desirable work behaviors under the use of HPWSs. Because HPWSs include team practices, formal programs of participative management, and autonomy-focused job designs that are exemplary of high involvement-oriented employment practices (e.g., [70][71][72]), leadership behaviors that promote employees' job autonomy are particularly relevant to conveying the positive effect of HPWSs on perceived support for innovation. Thus, we presume that subordinates are more likely to perceive high levels of support for innovation under LF.
Because laissez-faire leaders tend to avoid interactions with subordinates, the latter can find their own discretionary ways to solve problems and are more likely to lean on their other team members, not their supervisors, to collect information, accumulate knowledge, and ask for opinions to complete their work. For this reason, the social aspects embedded in subordinates' relationships with other team members become highly salient to them. As prior HPWSs studies have shown (e.g., [12,73]), the formation of a social climate among employees becomes an effective source of organizational innovation. However, under close supervision, policies of higher employee involvement and autonomy could simply mean a heavier workload and stress for subordinates (e.g., [74,75]). In this case, subordinates are therefore less likely to have the time and energy to engage in innovative behaviors. Thus, we believe that the positive association between HPWSs and employees' perceived support for innovation would be stronger under higher levels of LF. Hypothesis 3: LF will moderate the association between HPWSs and employees' perceived support for innovation such that the association between HPWSs and employees' perceived support for innovation will be stronger under high levels of LF.
As explained above, we believe that these three leadership styles will moderate the relationship between HPWSs and perceived support for innovation. However, the extent to which each leadership style moderates the relationship might differ across the three leadership styles. Given that we do not have an a priori theory of the relative strength of these three leadership styles, we are interested in empirically finding the relative strength of moderation by comparing the variance explained by these three moderators.
Research question: What is the relative strength of moderation (R-square increase) by the three moderators (TFL, TXL, and LF) in the association between HPWSs and employees' perceived support for innovation?

Perceived Support for Innovation and Team Effectiveness
West and Farr [76] describe perceived support for innovation (PSI) as "the expectation, approval and practical support of attempts to introduce new and improved ways of doing things in the work environment" (p. 38). PSI can be regarded as the extent to which an organization's members believe that their organization provides employees with the support needed to achieve organizational innovation [77]. We believe that PSI is positively associated with team effectiveness through both the social exchange mechanism and the increased confidence of employees. As social exchange theory [78] suggests, employees will work hard to return the favor of their organization when that organization develops long-term reciprocal social exchange relationships with its employees. Thus, high levels of organizational support can shape positive work behaviors of employees that can become a source of team effectiveness [79,80].
Moreover, employees' increased confidence under high PSI will also render work groups more effective. Employees are highly likely to develop the sense that they are equipped with the tangible and intangible resources needed for innovation when they perceive high levels of PSI. For these employees, organizational innovation appears highly achievable. Thus, employees will display more innovative work behaviors, rendering the work group more effective.

Hypothesis 4:
Employees' PSI is positively related to team effectiveness.

Sample
We collected the data in a city located in the southern part of mainland China with the cooperation of 14 organizations. These organizations were in the manufacturing and service industries. The nature of their work involved various functional areas, such as sales, service, manufacturing, customer service, administration, and research and development (R&D). In the collected survey, we limited our analysis to work teams with at least four responses, with an average group size of 6.5. The final sample included 429 employees nested in 66 work teams. For the employee participants, 53.40% were male and 45.10% were female. The average tenure and age were 5.93 and 31.82 years, respectively. With regard to education, 37.80% had at least associate degrees (2-year college) and approximately 22.40% had earned bachelor's degrees or above. Of the supervisor participants, 62.1% were male and 33.3% were female. The average tenure and age were 10.64 and 38.83 years, respectively. In terms of the level of education, 19.7% of participants had associate degrees and 63.6% had bachelor's degrees or above.

Procedure
The local government of the city in which we collected our data provided a list of organizations that we could use to recruit sample organizations. Based on the list, we made initial contacts with companies to explain the study goals and proposed questions to their HR managers and senior executives. With the consent of the participating organizations, our research assistants were present during surveys as a way to increase the confidentiality of the responses. To increase accuracy, the questions were translated into Chinese and then translated back into English [81]. Participation was voluntary and participants were asked to return the completed survey to our research assistants in an envelope that was provided. To minimize common source variance, we collected data from both supervisors and their subordinates. Managers answered questions regarding HPWSs and leadership, whereas subordinates answered questions regarding PSI and team effectiveness.

HPWS Scale
To measure the HPWSs perceived by each work team, we used ten items proposed by Den Hartog et al. [18]. We used frontline managers as the source of this measure. The items covered HR functional areas such as training, development, promotion, performance appraisal, compensation, autonomy, and teamwork. Sample items included "Training is provided regularly to employees", "Employees are informed about the opportunities for development that the organization offers", and "Employees are given room to make decisions in issues relating to their work". The scales for these items ranged from 1 (not at all) to 5 (to a great extent) for all items, and Cronbach's alpha was 0.92 for this measure.

TFL, TXL, and LF
To measure the extent of the three leadership styles of managers, the Multifactor Leadership Questionnaire form 6S (comprising 12 items for TFL, 6 items for TXL, and 3 items for LF) was used. We used frontline managers as the source of this measure. The scales for these items ranged from 1 (not at all) to 5 (to a great extent) for all items, and Cronbach's alphas were 0.93, 0.80, and 0.74 for TFL, TXL, and LF, respectively.

PSI
Eight items proposed by Anderson and West [82] were used to measure this variable. Their measurement validation study includes multiple dimensions capturing organizational innovation at the work group level. We used the "support for innovation" dimension to capture the extent to which members of a work group perceive active support required for innovation, as opposed to merely articulated support. The scales for these items ranged from 1 (not at all) to 5 (to a great extent) for all items, and Cronbach's alpha was 0.93 for this measure. Because this variable was measured at the individual level, we calculated several aggregation statistics (before aggregation was done) to check the validity of this measure as a group level variable. Specifically, r wg and ICC (1) were calculated. The statistic r wg can be considered as a measure of agreement across members who share the same organizational (or group) membership. The median r wg was sufficiently high at 0.82, which is above the recommended cut-off of 0.70 [83]. ICC (1) indicates the extent to which the variance of a measure can be explained by group membership. ICC (1) of this measure was also high at 0.12, which is above the recommended cut-off of 0.05 [84].

Team Effectiveness
To measure this variable, we used the scale developed by Campion et al. [85]. We decided to use only four of the items because some items seemed inappropriate for our sample. Employees' responses were used to assess the level of perceived effectiveness of their work group. Sample items included "Please indicate the quality of work done by your team" and "Please indicate the overall performance level of your team". The scales for these items ranged from 1 (not at all) to 5 (to a great extent) for all items, and Cronbach's alpha was 0.90 for this measure. We also calculated aggregation statistics because this variable was measured at the individual level. The values of r wg and ICC (1) for this measure were high at 0.87 and 0.18, respectively.

Control Variable
To rule out alternative explanations, we included team size as a control variable in testing hypotheses. Previous studies found that team size is a significant predictor of team process and effectiveness (e.g., [86,87]). Table 1 represents the means, standard deviations, and correlations of the variables included in the study. In Hypothesis 1, we suggested a moderating effect of TFL on the association between HPWSs and PSI. To test the hypothesis, we ran a series of ordinary least squares regression analyses in a hierarchical manner. In Model 1, we first entered a control variable ( Table 2). In the following model (Model 2), HPWSs and TFL were added to the previous model. In Model 3, we added an interaction term. Before running the main analyses, we centered HPWSs and TFL to create the interaction term by following the procedure suggested by Aiken and West [88]. The result showed that TFL was not a significant moderator of the relationship between HPWSs and PSI (β = 0.14, p > 0.05). Thus, Hypothesis 1 was not supported. However, the HPWS-TFL interaction term explained an additional 3% of the total variance and the regression coefficient was positive. In Hypothesis 2, we proposed the moderating effect of TXL on the relationship between HPWSs and PSI. The procedure for testing Hypothesis 2 was consistent with the procedure used to test Hypothesis 1. In Model 4, we entered a control variable. In Model 5, HPWSs and FXL were added. In Model 6, we included an interaction term created by multiplying HPWSs and TXL. The interaction was significant (β = 0.25, p < 0.05) and the R-square increase was 0.09 (p < 0.05). To probe the nature of this interaction, we plotted the slope between HPWSs and PSI at three levels of TXL (low, mean, and high TXL). As expected, the slope was more positive at the high level of TXL (1 SD above the mean) than at the low level of TXL (1 SD below the mean; Figure 2).  In Hypothesis 2, we proposed the moderating effect of TXL on the relationship between HPWSs and PSI. The procedure for testing Hypothesis 2 was consistent with the procedure used to test Hypothesis 1. In Model 4, we entered a control variable. In Model 5, HPWSs and FXL were added. In Model 6, we included an interaction term created by multiplying HPWSs and TXL. The interaction was significant (β = 0.25, p < 0.05) and the R-square increase was 0.09 (p < 0.05). To probe the nature of this interaction, we plotted the slope between HPWSs and PSI at three levels of TXL (low, mean, and high TXL). As expected, the slope was more positive at the high level of TXL (1 SD above the mean) than at the low level of TXL (1 SD below the mean; Figure 2). In Hypothesis 3, we suggested the moderating effect of LF on the relationship between HPWSs and PSI. The control variable was entered first (Model 7). Subsequently, HPWSs and LF were entered in Model 8. Finally, an interaction term of HPWSs and LF was added in Model 9. According to the result, the interaction was significant (β = 0.18, p < 0.05), and the increase in R-square was 0.07 (p < 0.05). We plotted this result to better understand the nature of this interaction, following the same procedure used for the testing of Hypothesis 2. The pattern of interaction was consistent with the finding in the testing of Hypothesis 2. Specifically, the slope between HPWSs and PSI was positive at the high level of LF (1 SD above the mean), whereas the slope was negative at the low level of LF (1 SD below the mean; Figure 3). In Hypothesis 3, we suggested the moderating effect of LF on the relationship between HPWSs and PSI. The control variable was entered first (Model 7). Subsequently, HPWSs and LF were entered in Model 8. Finally, an interaction term of HPWSs and LF was added in Model 9. According to the result, the interaction was significant (β = 0.18, p < 0.05), and the increase in R-square was 0.07 (p < 0.05). We plotted this result to better understand the nature of this interaction, following the same procedure used for the testing of Hypothesis 2. The pattern of interaction was consistent with the finding in the testing of Hypothesis 2. Specifically, the slope between HPWSs and PSI was positive at the high level of LF (1 SD above the mean), whereas the slope was negative at the low level of LF (1 SD below the mean; Figure 3). To probe the research question regarding the relative strength of moderation across the three leadership styles, we compared the variance explained by the three interaction terms. Specifically, PSI was regressed on HPWSs, all three leadership styles, and the control variable in Model 1 of Table  3. Next, each of the three interaction terms was added one at a time in the subsequent models (Models 2, 3, and 4) for the comparison. The results showed that, in order, the relative strength of moderation was TXL, LF, and TFL. Because the moderation of TFL was not statistically significant at the 0.10 level in Hypothesis 1, TFL explained the least variance of the criterion variable (R-square = 0.03, p > 0.05). In terms of the relative strength of moderation between TXL and LF, the sizes of the explained Rsquare were comparable. The R-squares explained by TXL and LF are 0.07 (p < 0.05) and 0.07 (p < 0.10), although the significance levels slightly differed. Finally, we tested the association between PSI and team effectiveness. In Model 5 of Table 3, team effectiveness was regressed on the control variable. In the subsequent model, PSI was added to the regression. The result showed a positive association between PSI and team effectiveness (β = 0.52, p < 0.05). To probe the research question regarding the relative strength of moderation across the three leadership styles, we compared the variance explained by the three interaction terms. Specifically, PSI was regressed on HPWSs, all three leadership styles, and the control variable in Model 1 of Table 3. Next, each of the three interaction terms was added one at a time in the subsequent models (Models 2, 3, and 4) for the comparison. The results showed that, in order, the relative strength of moderation was TXL, LF, and TFL. Because the moderation of TFL was not statistically significant at the 0.10 level in Hypothesis 1, TFL explained the least variance of the criterion variable (R-square = 0.03, p > 0.05). In terms of the relative strength of moderation between TXL and LF, the sizes of the explained R-square were comparable. The R-squares explained by TXL and LF are 0.07 (p < 0.05) and 0.07 (p < 0.10), although the significance levels slightly differed. Finally, we tested the association between PSI and team effectiveness. In Model 5 of Table 3, team effectiveness was regressed on the control variable. In the subsequent model, PSI was added to the regression. The result showed a positive association between PSI and team effectiveness (β = 0.52, p < 0.05).

Discussion and Implications
Recent SHRM studies have emphasized the effective implementation of HR practices to leverage the positive impact of HR systems on organizational performance [22]. Although it is important to adopt the appropriate set of HR practices to take advantage of the positive effect of HR systems, it is also important to successfully implement those adopted HR practices within organizations. In other words, the effective use of HR practices simultaneously includes the issue of both "what" and "how". In this study, we examined that claim by focusing on the role of frontline managers' leadership styles in implementing HPWSs, which has a positive impact on PSI and, subsequently, team effectiveness.
Consistent with our prediction, the results showed that the slope between HPWSs and PSI was more positive at the high level of TXL. This suggests that maintaining clear communications with respect to organizational expectations through leadership behaviors, such as contingent reward and management-by-exception, will be a necessary condition to strengthen the positive association between HPWSs and PSI. Similarly, the slope between HPWSs and PSI was more positive under the high level of LF. This result suggests that employees working for organizations that intensively utilize HPWSs are more likely to perceive strong support for innovation when their supervisors minimize managerial interventions and supervising behaviors. This finding is not surprising because subordinates can better focus on deep-level thinking and develop their ideas when they are less interrupted by their supervisors [89]. Distinctive from the case of TXL that showed a flat (slightly positive) slope for the HPWS-PSI relationship at the low level of TXL, the slope for the HPWS-PSI relationship at the low level of LF was negative. This result demonstrates that HPWSs could have negative impacts on PSI when supervisors do not allow a certain level of freedom and autonomy within their work groups. Indeed, several previous studies (e.g., [75]) have reported a negative impact from the use of HPWSs.
Although these findings are consistent with our expectations, we expect that the findings could change in a different cultural context. It is possible that the moderating effects by TXL and LF might diminish or even disappear with samples collected in Western cultures. Sometimes, observing rules are not strictly enforced and non-compliance is accepted in China when good justifications are given. Guanxi can be one example that reflects this cultural trait because strong guanxi can be a key to solve some restrictions in China [90]. Thus, we expect that TXL that sets clear rules and expectations can be more effective in implementing HPWSs in China than in other countries such as US that requires high levels of compliance to rules in general. In fact, our literature review showed that the effect of TXL on knowledge sharing or innovative behavior is positive in Jordan [46] but negative in the Netherlands [44].
Significant moderation by LF also needs some attention. LF turned out to be a positive influence in the current study as a recent study suggested [52]. However, LF can be negative to organizational functionality as many studies on LF demonstrated (e.g., [51][52][53]). These mixed findings can be attributed to cultural contexts in which research studies were conducted. Different from Western leaders, who place high emphasis on providing accurate and constructive feedback to subordinates, Asian leaders (e.g., China, Korea) place high emphasis on displaying deference to subordinates' abilities and maintaining less intervention. Some researchers argued that LF could be perceived as empowerment leadership in some settings since both the leadership styles allow a large degree of subordinates' discretion and autonomy [52,91]. For this reason, LF can be effective in China whereas being less effective or negative in Western countries in promoting the effect of innovation-oriented policies such as HPWSs.
Contrary to our expectations, TFL was not a significant moderator of the relationship between HPWSs and PSI. We believe that this is the result of the conceptual overlap between the two variables because both TFL and HPWSs can be a source of high employee involvement and provide a strong social climate within w ork groups. Indeed, the correlation between the two variables was 0.76 in the preliminary analysis. A recent study [92] testing the interactive effects of two highly related constructs (managers' service leadership and service-oriented HPWS) has also found that high levels of service leadership did not strengthen the positive relationship between service-oriented HPWS and collective customer knowledge. We believe that the substitution effect [93] emerged in the test.
The finding of insignificant moderation by TFL also can be attributed to the Chinese context and the result could be different in other Western countries. As Hofstede [94] suggests, Western culture is characterized by low collectivism (i.e., high individualism), contrary to Chinese culture, and developing a high level of a social climate for knowledge and information sharing could be more difficult in Western countries than in China. On the other hand, Chinese culture places a high priority on maintaining harmony and achieving collective goals in organizations and society in general [95]. Thus, TFL might not add over and above the effect of HPWSs to the promotion of a social climate for knowledge sharing and innovation in China, but it could strengthen the effect of HPWSs in Western countries.
The test of our research question regarding the relative strength of the moderating effect by each leadership style showed that the size of effects is TXL, LF, and TFL, in that order. However, the R-square increase explained by TXL was almost the same as the increase explained by LF, although theoretical mechanisms for these moderating effects can be different. This suggests that both leadership styles are equally useful in taking advantage of HPWSs to nurture high PSI. However, the use of HPWSs will only reduce PSI when a certain level of LF behavior is not present. In this case, HPWSs not coupled with a high degree of LF can only hurt organizational innovation. HPWSs that involve close supervision might increase employee work stress, leading to perceptions of reduced support for innovation.
This study has several implications and contributions. Although the critical roles of frontline managers in implementing HR practices are being recognized [19], our understanding of the roles of specific leadership styles of frontline managers in HR practice implementation was limited. This study contributes to the SHRM and leadership literature by examining the impact of several different leadership styles on implementing HPWSs and providing new insight into the mechanisms through which HPWSs affect performance. This finding implies that frontline managers, as well as HR staff, are responsible for HR activities and that managers should provide leadership behaviors that deliver clear organizational messages and guarantee employee discretion and autonomy to enhance HPWSs' impact on PSI. Second, this study also contributes to the SHRM literature by finding PSI to be one of the pathways through which HPWSs influence performance. Finally, the integration of the leadership theories with the SHRM literature is another contribution. Instead of relying on one leadership style or the general roles of supervision, we took advantage of well-established theoretical models (TFL, TXL, and LF) to expand our understanding of the relationship between HR systems and leadership. The study also broadens our understanding of the relative importance of those three leadership styles in promoting PSI through the use of HPWSs. This finding provides practical recommendations for managers to use in promoting innovation and creativity in their work groups.
This study is not without limitations. The use of cross-sectional data can affect its findings. Although multiple sources were used to collect data, the examination of the relationship between PSI and team effectiveness is not free from the influence of common source bias. Second, the limited sample size is another limitation. With a bigger sample, more sophisticated and rigorous analytical procedures can be employed. Finally, the use of a Chinese sample might limit the generalizability of the current study. With a sample collected in other national contexts, the three leadership styles could show different patterns as moderators for the hypothesized model.

Conclusions
This study addresses the issue of HR practice implementation by focusing on leadership styles of frontline managers. We believe that integrating the leadership literature with SHRM research findings provides useful insights to understand how frontline managers implement HR practices. Specifically, our results reinforce the notion that frontline managers' leadership plays a key role in HR system implementation. Moreover, the study expands this notion by testing the idea that different leadership styles might have distinctive impacts in implementing HPWSs. These findings provide insights into the role of leadership as a means of advancing our knowledge of how to use HR systems to improve organizational performance, suggesting new research directions for revealing the black box in the link between HPWSs and organizational effectiveness.