Employee IWB

The concept of employee IWB encompasses ‘the intentional creation, introduction and application of new ideas within a work role, group or organization, in order to benefit role performance, the group, or the organization’ (Janssen, Reference Janssen2000, p. 288). This definition implies that IWB is more than mere creativity arousing, although employee creativity is a necessary part of IWB, especially in the beginning, in order to generate new and useful ideas (Scott & Bruce, Reference Scott and Bruce1994). The scope of employee IWB further includes idea promotion and implementation of generated innovations for the expansion and renewal of products and services as well as the evolution of production processes and management systems (Montani, Battistelli, & Odoardi, Reference Montani, Battistelli and Odoardi2017).

In the innovation literature, there are different perspectives on employee IWB. Dorenbosch, Engen, and Verhagen (Reference Dorenbosch, Engen and Verhagen2005) distinguish between the conception and implementation of ideas, while Scott and Bruce (Reference Scott and Bruce1994) elaborate a triadic model comprising the generation of novel and useful ideas, the search for sponsorship, and the implementation of generated and promoted ideas. Indeed, innovation processes can be characterized as a mix of interrelated discontinuous activities (Kanter, Reference Kanter1988), where employees and leaders are most likely to be involved in any combination of these activities at any one point in time (Janssen, Reference Janssen2000; Scott & Bruce, Reference Scott and Bruce1994). We adopted a unidimensional yet comprehensive perspective on IWB, encapsulating Scott and Bruce’s (Reference Scott and Bruce1994) three-stage model, as elaborated below.

First, employees can generate ideas by exploring opportunities to address performance gaps, overcome workflow challenges, and identify sources for customer dissatisfaction. The inclination toward this exploratory behavior hinges on the perceived need, willingness, and ability to innovate, denoting a readiness for change (Holt, Armenakis, Field, & Harris, Reference Holt, Armenakis, Field and Harris2007; Holt & Daspit, Reference Holt and Daspit2015; Tan et al., Reference Tan, Van Dun and Wilderom2021). In the idea championing phase, employees forge alliances with supportive leaders and colleagues who can provide the necessary resources to propel proposed changes (such as personnel, funding, and time) (Amabile, Schatzel, Moneta, & Kramer, Reference Amabile, Schatzel, Moneta and Kramer2004; Denti & Hemlin, Reference Denti and Hemlin2012). Notably, a conducive environment that embraces change is instrumental in fostering innovation champions (Amankwaa, Gyensare, & Susomrith, Reference Amankwaa, Gyensare and Susomrith2019; Scott & Bruce, Reference Scott and Bruce1994; Tan et al., Reference Tan, Van Dun and Wilderom2021). The culmination of the process lies in the implementation phase, where employees invest effort in refining, testing, and actualizing an idea (Kanter, Reference Kanter1988). It is important to recognize that workplace innovation transcends job requirements, underscoring the need for organizational leaders and managers to understand and facilitate IWB (Montani et al., Reference Montani, Battistelli and Odoardi2017).

Leader behaviors and IWBs of followers

Although no ultimate definition of leadership exists (Yukl, Reference Yukl2002), most definitions of leadership reflect basic elements, including ‘group’, ‘influence’, and ‘goal’ (Northouse, Reference Northouse2018). We define leadership here as the interpersonal process of influencing others toward achieving a desired outcome. Although a plethora of leadership styles have demonstrated positive associations with innovation, TFL is the style that has been argued to be generically effective for stimulating creativity and innovation (Mumford & Hemlin, Reference Mumford and Hemlin2017; Mumford, Scott, Gaddis, & Strange, Reference Mumford, Scott, Gaddis and Strange2002). Transformational leaders motivate their followers by communicating appealing visions, encouraging them to think of different ways of solving problems, recognizing their higher-order needs, and serving as role models (Bass & Riggio, Reference Bass and Riggio2006). Especially in Western work settings, TFL has been found to be positively related to employees’ innovative behaviors and organizational innovations (e.g., García-Morales, Jiménez-Barrionuevo, & Gutiérrez-Gutiérrez, Reference García-Morales, Jiménez-Barrionuevo and Gutiérrez-Gutiérrez2012; Gumusluoglu & Ilsev, Reference Gumusluoglu and Ilsev2009; Rosing, Reference Rosing2017; Sarros, Cooper, & Santora, Reference Sarros, Cooper and Santora2008; Steele et al., Reference Steele, Watts and Den Hartog2018). The diversity of these findings on effective leadership may stem from TFL’s theoretical emphasis on idea generation (Hughes et al., Reference Hughes, Lee, Tian, Newman and Legood2018; Lee et al., Reference Lee, Legood, Hughes, Tian, Newman and Knight2020), to the detriment of idea promotion and implementation (Wang et al., Reference Wang, Eva, Newman and Zhou2021). Consequently, scholars have proposed that innovation-inducing leader behaviors must supplement TFL (e.g., Anderson et al., Reference Anderson, Potočnik and Zhou2014; Denti & Hemlin, Reference Denti and Hemlin2012; Hughes et al., Reference Hughes, Lee, Tian, Newman and Legood2018; Siangchokyoo et al., Reference Siangchokyoo, Klinger and Campion2020; Yukl, Mahsud, Prussia, & Hassan, Reference Yukl, Mahsud, Prussia and Hassan2019). Hence, the complexity of the relationship between leader behaviors that promote IWB among followers is likely to include moderating, mediating, including contextual factors (Amankwaa et al., Reference Amankwaa, Gyensare and Susomrith2019; Fischer & Sitkin, Reference Fischer and Sitkin2023; Hughes et al., Reference Hughes, Lee, Tian, Newman and Legood2018; Rosing, Reference Rosing2017; Van Knippenberg & Sitkin, Reference Van Knippenberg and Sitkin2013).

Indeed, several researchers have emphasized that effective leadership for IWB needs to consider both idea generation and the processes necessary for the implementation of ideas (De Jong & Den Hartog, Reference De Jong and Den Hartog2007). Indeed, the qualitative study of 12 leaders of knowledge-intensive services revealed a broader range of leader behaviors tapping either idea generation (e.g., providing vision) or idea application (e.g., providing resources) or both (e.g., monitoring progress) (De Jong & Den Hartog, Reference De Jong and Den Hartog2007). The concept of ambidextrous leadership goes one step further; leaders may stimulate idea generation and idea implementation by using paradoxical leader behaviors (Rosing, Frese, & Bausch, Reference Rosing, Frese and Bausch2011). Whereas opening behaviors (e.g., encouraging alternative ways of accomplishing tasks) enhance employees’ exploration behaviors, closing behaviors (e.g., taking corrective action) foster employees’ exploitation activities (Rosing et al., Reference Rosing, Frese and Bausch2011). Hence, leaders need to switch between opposing behaviors (e.g., allowing for errors vs. sanctioning errors) to foster both idea generation and idea implementation. Already in their repertoire of behaviors, leaders wishing to effectively rouse and establish effective innovative behavior among their followers must reconcile the potentially conflicting perceptions that others, and especially their followers, may have about them.

Given the complexity of effective organizational leadership for IWB, and for resolving the followers’ paradox of idea generation, development/promotion, and implementation, we argue here that leaders’ broad behavioral repertoire must carefully match followers’ activities in the innovation process; only then leaders can effectively evoke employees’ IWB. There is already some research that focused on the various stages of the innovation process (e.g., Amabile et al., Reference Amabile, Schatzel, Moneta and Kramer2004; George & Zhou, Reference George and Zhou2007), but very few studies have focused on the idea that different stages of innovation may require different leader behaviors in facilitating the various activities of the followers involved (Basu & Green, Reference Basu and Green1997; Gerpott, Bledow, & Kühnel, Reference Gerpott, Bledow and Kühnel2022). Therefore, we still know little about which and to what extent leader behaviors are effective in evoking employees’ innovative behaviors, let alone in Asian contexts. The different stages of innovation provide a useful theoretical foundation for identifying specific leader behaviors that promote employees’ innovative efforts. Specifically, we propose that different leader behaviors are required for stimulating followers’ idea generation, promotion, and implementation. In the below, we review scholarly leadership insights underpinning the three-pronged empirical approach.

Leader behaviors for follower idea generation, promotion, and implementation in Asian contexts

We do know that Asian leadership styles often reflect cultural values such as collectivism, harmony, and respect for authority (Zheng, Graham, Farh, & Huang, Reference Zheng, Graham, Farh and Huang2021). Leaders in Asian workplaces tend to adopt a more participative and inclusive approach to encourage idea generation among followers (Hofstede, Reference Hofstede2001). This participative leadership style fosters a sense of belonging and psychological safety among employees, empowering them to contribute innovative ideas without fear of criticism (Javed, Naqvi, Khan, Arjoon, & Tayyeb, Reference Javed, Naqvi, Khan, Arjoon and Tayyeb2019). Such leadership behaviors include clearly communicating shared objectives for change, appreciating, and encouraging employees’ creative endeavors, actively listening to their inputs, and demonstrating a willingness to provide support. These actions foster a positive attitude among the followers, thereby encouraging collective engagement in the exploration of new and innovative ideas and solutions (Cheng & Hong, Reference Cheng and Hong2017; Janssen, Van de Vliert, & West, Reference Janssen, Van de Vliert and West2004). Leaders in Asian cultures often emphasize the importance of humility, which creates an environment where followers are more likely to share their ideas openly (Hu, Erdogan, Jiang, Bauer, & Liu, Reference Hu, Erdogan, Jiang, Bauer and Liu2018).

The promotion of ideas in Asian work environments is influenced by the cultural emphasis on interpersonal relations and communications (Lee & Kim, Reference Lee and Kim2021). To propagate employees’ ideas, managers should connect with the organizational decision-makers (Byrne, Mumford, Barrett, & Vessey, Reference Byrne, Mumford, Barrett and Vessey2009), and in many cases, champion innovation themselves (Rogers, Reference Rogers2010). Achieving this may necessitate active relationship-building with members of other business units, stimulating cross-functional collaboration and enhancing employee ideal success. High-quality exchanges between leaders and followers not only empower greater autonomy but also provide access to better information, mitigating the fear of failure and thereby bolstering employee courage in promoting their ideas (Prieto & Pérez-Santana, Reference Prieto and Pérez-Santana2014). These relations-oriented behaviors (Yukl, Reference Yukl2012; Yukl et al., Reference Yukl, Mahsud, Prussia and Hassan2019) help leaders maintain their followers’ task commitment and cooperation.

In Asian cultures, where followers are more anchored in their leader’s role modeling (Engelen et al., Reference Engelen, Schmidt, Strenger and Brettel2014), we assume that leaders demonstrating IWB themselves can effectively catalyze employee IWB (De Jong & Den Hartog, Reference De Jong and Den Hartog2007). Employees can learn from their leaders’ innovation capacity through observation and renewed interaction with others (Tierney & Farmer, Reference Tierney and Farmer2004). This reasoning is also in line with the social learning theory, which states that learning takes place vicariously by modeling and self-control (Bandura, Reference Bandura1982).

Implementing ideas in Asian workplaces requires leaders who can navigate the diverse cultural landscape and ensure alignment of innovative initiatives with organizational goals (Ahmad, Butt, Chen, & Liu, Reference Ahmad, Butt, Chen and Liu2023). Authentic leadership, characterized by transparency and ethical decision-making, is particularly valuable in gaining employees’ trust and facilitating idea implementation (Lv, Jiang, Chen, & Zhang, Reference Lv, Jiang, Chen and Zhang2022). Leaders can leverage on their extensive network, both internal and external, to garner support and resources for implementing innovative ideas, exemplifying the importance of team leadership and social capital in idea realization (Hitt, Lee, & Yucel, Reference Hitt, Lee and Yucel2002). To ensure employees can implement (their) ideas, leaders may adopt task-oriented behaviors (Yukl, Reference Yukl2012; Yukl et al., Reference Yukl, Mahsud, Prussia and Hassan2019) encompassing planning improvement activities, clarifying roles and objectives, role-modeling, and monitoring (innovation) performance, and fostering mutual accountability for exceeding team objectives (Mumford et al., Reference Mumford, Scott, Gaddis and Strange2002; Wang et al., Reference Wang, Eva, Newman and Zhou2021). Leaders can thus grant employees some autonomy to determine independently how to execute a job or task and to participate in decision making (Lee, Choi, & Kang, Reference Lee, Choi and Kang2021; Yukl, Reference Yukl2012). The opportunity to exercise self-direction and control makes employees feel important, competent, and respected (Xiong Chen & Aryee, Reference Xiong Chen and Aryee2007). This will therefore motivate employees to experiment in their jobs, leading to IWB. Leaders can also use constructive feedback to communicate to employees how (idea implementation) tasks should be executed and assessing whether their performance aligns with desired outcomes (Lee et al., Reference Lee, Choi and Kang2021).

In sum, follower IWB is substantially influenced by leader behaviors that stimulate its three process dimensions. The subsequent sections delve into the development of a measurement instrument to assess leader innovation-stimulating behaviors that impact employee IWB.

Paper overview: Three empirical studies conducted in Singapore

Three consecutive empirical studies were conducted in public and private service organizations in Singapore (see Table 1). This city state as the research site is driven by a distinctively Asian blend of characteristics and therefore a relevant setting for investigating the Asian interplay between specific leader behaviors and employee IWBs. Singapore’s economic growth hinges already on innovation, and specifically on organizational innovation (Ng, Reference Ng2012). Singapore is often seen as a microcosm of Asia, showcasing a multicultural society that encompasses Chinese, Malay, Indian, and other ethnic groups (Pang & De Meyer, Reference Pang and De Meyer2015). Its strong emphasis on collectivism resonates with the broader Asian cultural trait of valuing harmony, group cohesion, and shared goals. Consequently, it provides a good site for developing a new survey instrument for the study of effective leader behaviors that stimulate effective follower IWB in Asian work environments.

Table 1. Scale development process

Note: KMO = Kaiser–Meyer–Olkin.

While Singapore’s leadership landscape has been influenced by Western leadership theories and practices, the nuances of the broader cultural context call for a better understanding of leader behaviors that promote followers’ IWBs. Addressing this gap has implications for leadership development, talent management, and organizational innovation strategies. In Study 1, we followed Hinkin (Reference Hinkin1998) scale-development steps by generating survey items through semi-structured interviews with leaders and employees in private and public service systems. In Study 2, we examined the reliability and factor structure of the ILS in a public service organization, following the methodology outlined by Lambert and Newman (Reference Lambert and Newman2022). Study 3 validates the scale with a multi-actor sample comprising both support staff and academic staff from a higher educational institute located in Singapore.

Item generation

For item development, we followed Hinkin’s (Reference Hinkin1998) recommendation to use both deductive and inductive approaches. In the first step, we specified the key employee tasks required for innovation, using existing IWBs (e.g., De Jong & Den Hartog, Reference De Jong and Den Hartog2010; Janssen, Reference Janssen2000). Utilizing the earlier-mentioned three-stage model of innovation, Janssen’s (Reference Janssen2000) IWB scale measures individual’s idea generation, promotion, and realization behavior in the workplace. The items developed by De Jong and Den Hartog (Reference De Jong and Den Hartog2010) enabled us to further define employee activities involved in idea generation, championing, and application.

Next, we reviewed the literature (presented in the above) on the relationship between leadership and innovation to identify a broad range of leadership behaviors that may promote employees’ innovative efforts within the innovation process. Moreover, we conducted semi-structured interviews with 20 leaders and 40 employees from four large service organizations in Singapore: an information and technology corporation; a public service organization which facilitates trade documentation and security administration; a company providing event and advertising services; and an educational institute targeted at special needs of individuals (see Table 2). These four organizations were purposefully selected as they had attained either a Singapore Quality Class award, Innovation Award, or People Developer Award; we thus sought to identify the best leadership behavior that support high employee IWB (Eisenhardt & Graebner, Reference Eisenhardt and Graebner2007). The number of leaders ranged between 4 and 15 leaders per organization, to allow for a representative leader sample in each organization.

Table 2. Profile of the interviewees in Study 1

Furthermore, following qualitative research principles, the sample size was co-determined based on carefully tracing signs of data saturation, as advocated by Guest, Bunce, and Johnson (Reference Guest, Bunce and Johnson2006), and reinforced by recent studies by Khan, Hossain, Jahed, Akter, and Pappas (Reference Khan, Hossain, Jahed, Akter and Pappas2024) and Patton (Reference Patton2023). When we started to approach our sample size of 60 participants, new themes and insights ceased to emerge from the interviews, pointing to thematic saturation. Hence, the sample size of 60 allowed for gathering in-depth insights while remaining operationally feasible.

To ensure a broad representation of leader behaviors stimulating innovation across various industries, we have sampled a diverse set of leaders and employees, reflecting the heterogeneous business environments in Asia (Hennink, Kaiser, & Marconi, Reference Hennink, Kaiser and Marconi2023). The diversity of participants also enhances the transferability of the findings across various organizational context (Patton, Reference Patton2023). The final set of interviewees were predominantly male (55%), ranged between 28 and 65 years of age, and nearly every educational and socioeconomic background was represented in the sample. The employees had worked for at least 2 years with their organizations, and had been nominated by their Human Resources departments as high-performing staff: all of them agreed to participate. The inclusion of high-performing employees ensures that the study was not just based on leaders’ own self-reported behaviors, but also employees’ reflection on the relationship between leadership and (successful) innovation. The leaders were C-suite level, departmental heads, or frontline supervisors. All the participants were informed of the purpose of the in-depth interviews: to gather information about leader behaviors that evoke employee IWB. After being assured of anonymity and confidentiality, all the participants agreed to being audio-recorded.

At the start of each interview, we provided the participant with our definition of innovation-stimulating behaviors adapted from West and Anderson (Reference West and Anderson1996). In addition, the participants read brief descriptions of the different innovation stages based on Scott and Bruce (Reference Scott and Bruce1994). The interview format consisted of three open-ended questions. The questions for the employees were (1) What do managers of a highly innovative work team do? That is, what behaviors come to mind when you think of a manager who has led you effectively to be innovative? (2) Describe a time when your manager successfully convinced you that a change in process or approach was necessary for success: What behaviors did they demonstrate? And (3) Think of a situation when your manager affected the team/individual’s performance: What did your manager do to block the team/individual from being innovative? How did your manager’s behavior diminish the team/individual’s success? The same three questions were rephrased to the self-context for the leaders’ interviews. The average interview duration was 45 minutes. All the responses were transcribed immediately afterward, and all interviewees confirmed their transcript.

Leader behaviors were identified through independent transcript content coding by two researchers using Gioia’s inductive approach (Gioia, Corley, & Hamilton, Reference Gioia, Corley and Hamilton2013; Locke, Feldman, & Golden-Biddle, Reference Locke, Feldman and Golden-Biddle2020). Both researchers generated a list of behaviors mentioned in the interviews and copied each one onto separate index cards. Differences were discussed and resolved. On removing the redundant cards, the two researchers sorted the total 110 behaviors into groups according to conceptual similarities. As recommended by Corbin and Strauss (Reference Corbin and Strauss2014) and Mathieu, Luciano, D’Innocenzo, Klock, and LePine (Reference Mathieu, Luciano, D’Innocenzo, Klock and LePine2020), we also used the literature to think of innovation-stimulating leader behavior categorizations. We then used Yukl (Reference Yukl2008, Reference Yukl2012) meta-categories and comprehensive hierarchical taxonomy of daily managerial work to classify the specific leader behaviors. The final classification consisted of 11 leader behaviors which were mapped onto Scott and Bruce’s (Reference Scott and Bruce1994) three employee IWB stages. Table 3 shows their initial classification and definitions based on the wider literature.

Table 3. Overview of Study 1’s leader behaviors related to the IWB subdimensions based on Scott and Bruce (Reference Scott and Bruce1994)

Note: GI = generate idea; CI = champion idea; II = implement idea.

Content validity

Based on the 11 leader behaviors, we assessed the content validity of the generated leader behavior items that evoke employee IWB with a panel of experts, following Lambert and Newman (Reference Lambert and Newman2022). Where possible, the survey item phrases were taken from the original interviews. In line with Kaplan and Saccuzzo (Reference Kaplan and Saccuzzo2017), we excluded draft items that appeared too complex, ambiguous, had double negatives, or did not describe a certain behavior. The remaining 42 items were then arranged in a logical and systematic way (also utilizing Yukl, Reference Yukl2012; Yukl, Gordon, & Taber, Reference Yukl, Gordon and Taber2002; Yukl et al., Reference Yukl, Mahsud, Prussia and Hassan2019) to assist the participants to easily complete the survey involving the 11 leader behaviors.

At the item-development stage, the item-objective congruence (IOC) index was used to evaluate content validity (Rovinelli & Hambleton, Reference Rovinelli and Hambleton1977; Turner & Carlson, Reference Turner and Carlson2003). Here, the 42 items were sent to a panel of five experts (Rovinelli & Hambleton, Reference Rovinelli and Hambleton1977) with an IOC form which asked them to evaluate whether the items did denote what they were supposed to measure by assigning a score of +1 (clearly measures a depicted leader behavior), 0 (the content measurement is not clear or ambiguous), or −1 (clearly does not measure a depicted leader behavior); a space was provided for comments regarding the clarity and understandability of the items. The five experts had a master or a PhD degree, had worked in various innovative industries, and had been in their current role for more than 10 years. This panel consisted of two content experts with relevant quantitative research experience and three lay experts who were potential research subjects. To determine their general degree of agreement over each statement, an average IOC was calculated (Turner & Carlson, Reference Turner and Carlson2003). If the score was greater than 0.5, the item was deemed to be fitting (Shaikh, Reference Shaikh2018).

Results

As shown in Table 4, most of the items achieved the required IOC scores. Two of the 42 items had a low IOC score and had to be rephrased. Item 10 was amended from ‘Implements idea generation activities’ to ‘Implements idea generation activities (e.g., brainstorming sessions, focus group discussions, strategy sessions, etc.)’. Item 36 was amended from ‘Is curious and frequently challenges the status quo during idea generation activities’ to ‘Is curious and frequently challenges the status quo’.

Table 4. Index of item-objective congruence (IOC) of each statement included in the Innovative Leader Survey (ILS) scale

Note:

^a Items were deleted as a result of correlation analysis in Study 2. The italic-faced items were amended because of a low IOC.

Sampling procedure

The data were obtained from one of the organizations that had previously participated in Study 1. This large public-service organization has received a Digital Service Award for their innovative way of facilitating Singapore’s cross-border trades. They had embarked on a leadership development program for their upper and middle management. One of the researchers was engaged to conduct the program.

Several weeks before the program started, a questionnaire was distributed with an accompanying note outlining this study’s purpose. The final sample size was 1,037 respondents, consisting of focal leaders (N = 86), the focal leaders’ bosses (some had two bosses due to a matrix structure, N = 117), at least three followers (N = 390), at least three peers (N = 319), and at least three external partners or customers (N = 125). The focal leaders’ ages ranged from 30 to 65 years, with a mean of 41 years. Of these, 59.9% were males and 40.1% were females. Moreover, 64% of focal leaders had occupied their current positions for over a year.

The multilevel sampling strategy ensured diverse representation of perspectives within the organization, thereby enhancing the validity and generalizability of the study’s findings (Ruvio, Shoham, Vigoda‐Gadot, & Schwabsky, Reference Ruvio, Shoham, Vigoda‐Gadot and Schwabsky2014; Van Dun, Hicks, & Wilderom, Reference Van Dun, Hicks and Wilderom2017). Moreover, the inclusion of a minimum of three respondents from each rater type, such as followers, peers and external partners or customers, further enhanced the depth and breadth of insights gathered. This approach allowed for a comprehensive examination of leadership behaviors and their impacts across the organizational contexts and relationships (Van Dun et al., Reference Van Dun, Hicks and Wilderom2017; Van Dun & Wilderom, Reference Van Dun and Wilderom2021).

Data collection procedure

The survey was administered on an online platform, with focal leaders providing a self-assessment of their own behaviors. The other raters were tasked to assess the frequency of each behavior exhibited by their focal leader. We adopted the frequency scale rather than the Likert-type agreement scale to avoid the negatively sounding ‘disagreement’ rating points: as Shipper, Hoffman, and Rotondo (Reference Shipper, Hoffman and Rotondo2007) recognized, criticism should be avoided in most Asian cultures. This is also why we chose a 7-point scale with three points above the midpoint to allow for more variation, where 1 = ‘Never’ and 7 = ‘Always’. The raters were informed that the focal leaders would receive composite feedback on their behaviors during the workshop. All the participants were assured that the individual ratings would be used exclusively for research purposes and would remain confidential.

Data analysis

Given the known differences in self- and other-ratings, particularly in Asian cultures (Dorfman et al., Reference Dorfman, Javidan, Hanges, Dastmalchian and House2012), we commenced our analysis by comparing self-ratings to other-ratings. As the data was not normally distributed, Mann–Whitney U tests were performed. Subsequently, we conducted intercorrelation analyses across the 11 behavior scales and 42 items. After that, we performed exploratory factor analysis using R (version 4.2.1) to ascertain whether the behaviors and items converged into a single construct. Principal axis analysis and a promax rotation (with 200 iterations) were employed in Lavaan, version 0.6-12. The use of promax rotation, which accounts for correlations between the components, was preferred over varimax rotation (Kline, Reference Kline2014). Items had to have a minimum factor loading of .4 on one factor only and a minimum difference in factor loading of .2 (Kline, Reference Kline2014). Also, confirmatory factor analysis (CFA) was used to evaluate the assumed factor structure (Kline, Reference Kline2014). We considered both single-factor and multiple-factor models. The appropriateness of each model was examined using several indices of fit such as the ratio of chi-square to its degree of freedom, its P-value, the root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), and the comparative fit index (CFI) (Hu & Bentler, Reference Hu and Bentler1999).

Results

The internal consistency reliability of the 11 behaviors in the scale was computed separately for the self and others ratings. The resulting Cronbach’s alpha coefficients are presented in Table 5, demonstrating satisfactory levels overall. However, the Cronbach’s alpha of one leader-behavioral construct fell below the recommended threshold of 0.70. As suggested by Briggs and Cheek (Reference Briggs and Cheek1986) and DeVellis (Reference DeVellis2016), a lower Cronbach’s alpha might still be acceptable especially to the earlier phase of construct development (similar to Luthans, Avolio, Avey, & Norman, Reference Luthans, Avolio, Avey and Norman2007) if the indicators show effective measurement of the intended construct, particularly in cases where the scale is brief or when measuring complex constructs. Table 6 displays the intercorrelations among these 11 behaviors, indicating moderate levels of correlation (p < .01). Hence, the ILS instrument appears to effectively capture the set of innovative leader behaviors.

Table 5. Descriptive statistics for the leader behavior scales and Mann–Whitney U Tests for the comparison between the focal leaders and all-others ratings in Study 2

Note: N _leaders = 86 (Focal Leaders); N _AllOthers = 951 (N _Boss = 117; N _Follower = 390; N _Peer = 319; N _Customer = 125.

* p < .001 (two-tailed, based on a Mann–Whitney U two-independent-samples test).

Table 6. Intercorrelation among the 11 Scales in Study 2

Note: The lower left triangle shows correlations among the scales from self-ratings; the upper right triangle shows correlations among scales for ratings from all others (i.e., boss, followers, peers, customers/partners). All correlations are significant at p < .01 level.

Factor analysis and model fit

Next, we obtained a correlation matrix for the items associated with the 11 behaviors and deleted eight items with low inter-item and item-total correlations (see Table 7 for the deleted items). We then performed exploratory factor analysis on 34 items which converged into a three-factor model. The three factors accounted for 50.90% of the variance. The results showed a high degree of sampling adequacy (Kaiser–Meyer–Olkin), of .9873, χ² = 1283, df = 462, and a p-value of <.001.

Table 7. Results of the exploratory factor analysis in Study 2

Notes. KMO = .9873; χ² = 1032; df = 375, p < .001. Items 6 and 19 were deleted due to cross-loading. Item 29 was deleted due to low factor loadings.

Three items had to be deleted due to either cross-loadings (items 6 and 19, Table 6) or low factor loadings (item 29, Table 7). After running another exploratory factor analysis, the results showed a Kaiser–Meyer–Olkin of .9873, χ² = 1032, df = 375, and a p-value of <.001. The same three factors emerged and accounted for 52.1% of the variance.

We then conducted CFA with the 31 remaining items. The three-factor model fitted the data well: CMIN/df = 4.075, p < .001, CFI = 0.953, RMSEA = 0.054, SRMR = 0.039. The reliability coefficient for the three retrieved subscales revealed good internal consistency (α = .959 for factor 1, .904 for factor 2, and .963 for factor 3). To label the three factors, we modified Yukl’s three dimensions (Reference Yukl2012, Reference Yukl, Mahsud, Prussia and Hassan2019): change-oriented behaviors, relations-oriented behaviors, and task-oriented behaviors, to: Envision, Energize, and Enable, respectively.

Study 2 thus showed that three meta-categories of behaviors seem to be important for an Asian leader’s promotion of IWB among his or her followers. Although this procedure was the first step in establishing the ILS’s factor structure, a more rigorous cross-validation with an independent sample was carried out next.

Sampling procedure

We used a mixture of purposive and convenience sampling techniques, leveraging the first researcher’s access to the sample and the difficulty of gathering a large, diverse group of participants from one organization (May & Perry, Reference May and Perry2022). First of all, we purposively selected a leading educational institute in Singapore for Study 3, which is renowned for its innovation in education and research, providing an ideal setting to explore the leader behavioral repertoire in progressive, innovation-oriented organizations. The institution’s proactive nurturing of innovation facilitated a diverse and large pool of leaders for our study, enhancing the credibility and scope of our data collection and analysis. To allow for the representation of diverse viewpoints and experiences, both academic and non-academic support-staff leaders and employees were invited to participate, thereby enriching the dataset and the applicability of our findings across different organizational contexts (Eagly & Chin, Reference Eagly and Chin2010; Esen, Bellibas, & Gumus, Reference Esen, Bellibas and Gumus2020).

The convenience element of the sampling strategy was due to the first author’s involvement in a leadership development course at the studied organization (Etikan, Musa, & Alkassim, Reference Etikan, Musa and Alkassim2016), which also led the organization to allocate time for its leaders to participate voluntarily. This practical access allowed direct engagement with participants across the organization (both in highly professionalized roles to more administrative support roles) that were interested in leadership development (Patton, Reference Patton2023). This streamlined participant recruitment and increased the likelihood of participation (Belliveau & Yakovenko, Reference Belliveau and Yakovenko2022).

Several weeks before the first leadership development workshop was to be conducted, the 26 focal leaders comprising 12 academic and 14 non-academic staff (middle and upper-level) were asked to nominate their raters. Similar to Study 2, in terms of sample size we aimed to include a minimum of three respondents from each rater type per focal leader. Thus, apart from their direct bosses, we asked them to name at least three of their followers, at least three peers (i.e., same-level colleagues from the same unit), and at least three external partners or customers. This multisource feedback approach facilitated a comprehensive evaluation of leader behaviors from diverse viewpoints. Crucially, the nomination was overseen and approved by the HR manager responsible for staff development, ensuring fairness and transparency in the selection of raters. All the respondents were assured that their individual reports would not be seen by their managers or anyone else in the organization and that only composite feedback on the behaviors would be presented to the focal leaders for leadership development purposes. The response rate was 99.3% (287 out of 289 invited raters, N _Leader = 26, N _Boss = 26, N _Follower = 78, N _Peer = 71, N _Others = 86). In the second wave of data collection 4 months later, 76 (97.4%) of the 78 followers responded.

Measures

The online ILS instrument included the 31 items representing the three subscales (Table 7). All the raters were asked to assess the frequency of the focal leader’s behaviors with the 7-point scale (1 = never, 7 = always). The Cronbach’s alpha for the Envision subscale (α_Leader = .952, α_All-others = .96) was .973, for Energize (α_Leader = .873, α_All-others = .921) was .924, and for Enable (α_Leader = .953, α_All-others = .972) was .981.

To measure the discriminant validity of the ILS conducted 4 months later, we also assessed the raters’ IWB using the nine Janssen (Reference Janssen2000) items, again on a 7-point scale (1 = never, 7 = always). An example item is: ‘I seek new working methods, techniques or instruments’. Its Cronbach’s alpha was .930.

Data analysis

First, we ran Mann–Whitney U tests to compare the self-ratings with all the other raters to confirm Study 2’s findings. Then, we investigated whether the developed factor structure of the ILS could be replicated by conducting a CFA on the item-level data to examine the factor structure of the proposed instrument. The model fit was examined with several indices such as the chi-square, the RMSEA, CFI, as well as SRMR.

Next, we ran a multiple linear regression analysis of the leaders’ Envision, Energize, and Enable behaviors with the dependent variable: their followers’ IWB 4 months later. This approach was chosen to delve deeply into the nuanced interactions between these leader behaviors and the assumed subsequent follower outcome. Furthermore, we employed structural equation modeling to investigate the moderating influence of Envision, Energize, and Enable on followers’ IWB. Structural equation modeling was selected due to its ability to construct comprehensive measurement models for both the independent and dependent latent variables, each supported by multiple observed variables. This method not only allowed us to assess the individual impact of Envision, Energize, and Enable on IWB but also to explore the interrelationships between the leader behaviors. Considering the constraint of small sample size and the non-normal, categorical data, we utilized a robust variant of the diagonally weighted least squares estimator within the Lavaan version 0.6-12 software. This robust approach was essential for accurate analysis, providing robust standard errors and a scaled test statistic as suggested by Yuan and Bentler (Reference Yuan and Bentler2007).

Results

ILS factor confirmation

The entire sample’s data were tested for factorability which yielded a Kaiser–Meyer–Olkin of .99. The presumed three-factor model could be confirmed through CFA in terms of its model fit indices: χ²₍₄₃₁₎ = 985.789 with p-value = .000, CFI = .935, RMSEA = .067, and SRMR = .039. All 31 items loaded positively with standardized factor loadings of at least .719 on the three ILS scales (Table 8).

Table 8. Results of Study 3’s confirmatory factor analysis

Relationship between ILS and IWB

The correlations in Table 9 indicated that there was a strong relationship among the factors (r _min = .870, r _max = .947), i.e., the behaviors observed in the ILS seemed to be compatible or homogeneous. Moreover, they correlated significantly with employees’ IWB.

Table 9. Summary of Study 3’s descriptive statistics and zero-order correlations of the ILS factors and followers’ IWB

Note: Cronbach’s alpha values are presented on the diagonal between brackets. All correlations are significant with p < .01.

We wanted to know, though, if the single factors were also significantly related to employee IWB. Thus, we tested different models, starting with the direct influence of the three factors (Envision, Energize, and Enable) on employee IWB by performing a simple linear regression. The result showed a significant relationship (p-value = .000, β = .899, R ² = .879). Although we saw that the ILS can be used to predict IWB, we wished to break down the ILS in terms of its latent variables: Envision, Energize, and Enable, and show their relationship with IWB. Hence, we conducted a multiple linear regression with these three predictors and IWB as the dependent variable. However, this model was rejected due to strong covariances between Envision, Energize, and Enable, indicated by high variance inflation (VIF_Envision = 9.67, VIF_Energize = 6.41, VIF_Enable = 14.67), leading to insignificant relationships between Energize and IWB as well as Enable and IWB. Therefore, we used structural equation modeling starting with the moderating effect of Energize between Enable and employee IWB (Model 1, Table 10). The model fit indicators were all unsatisfactory, with CFI = .018, TLI = .039, RMSEA = .085 and SRMR = .164 (Schreiber, Nora, Stage, Barlow & King, Reference Schreiber, Nora, Stage, Barlow and King2006). Similar results were obtained on testing the moderating effect of Energize between Envision and Enable (Model 2, Table 10). Swapping the predictors and moderators around did not yield better results.

Table 10. Model fit indices to establish the relationship between ILS and IWB in Study 3

Note: DWLS = diagonally weighted least squares estimator, ML = maximum likelihood.

Considering that Energize had the weakest direct correlation with IWB (Table 10), it could be assumed that Energize functions better as an intermediate variable between other predictors. Hence, we tested Energize as the mediator. Model 3 (Table 10) shows better fit indices than achieved with the moderation models. However, the CFI, TLI, RMSEA, and SRMR were still mediocre. Additionally, a p-value of .000 for the Chi-Squared test highlighted that there were significant differences between the model and the data (Kline, Reference Kline2014; Schreiber et al., Reference Schreiber, Nora, Stage, Barlow and King2006). Since we had ordinal discrete data, analyzing Model 3 with the maximum likelihood estimation method delivered non-satisfactory model fit indices (see Table 10). Hoyle and Isherwood (Reference Hoyle and Isherwood2013) noted maximum likelihood was developed for continuous data but that the diagonally weighted least squares estimaton method is robust against deviations from normality and so better suited for ordinal discrete data. Hence, Model 4 was analyzed utilizing the diagonally weighted least squares estimation method leading to much better model fit indices such as CFI = .930, TLI = .925, RMSEA = .023, and SRMR = .041 (see Table 9). The p-value for the Chi-Squared test was .224 meaning this model did not deviate significantly from the analyzed data (Table 10 and Fig. 1).

Notes. Env = Envision, Enr = Energize, Enb = Enable, IWB = Innovative work behavior of followers. Survey items 6, 19, and 29 are not included. Description and item labels can be found in Table 7. All effects are significant with p < .001.

Figure 1. Structural model on effects of leader behaviors on followers’ IWB, Study 3.

In addition, we used the previously established structural equation model with each of the innovation process stages, i.e., idea generation, championing, and implementation. Although there is a significant relationship between all the subscales and stages (Table 11), the results show that employees’ idea generation behaviors are mostly driven by leaders’ Envision behaviors (β = 1.037), whereas leaders’ Energize behaviors support more employee idea championing (β = .748), and leaders help employees best via Enable behaviors to implement their ideas (β = .838). Thus, the three ILS subscales paralleled the logic of the three employee-innovation stages our research began with.

Table 11. Effects of the ILS subscales on employees’ idea generation, championing, and implementation behaviors, study 3

Note: All effects are significant with p < .001.