The Influence of Organizational Commitment in Climate for Innovation and Employee Retention Relationship: An Empirical Study in Higher Education Institutions

Organizational commitment is one of the most widely used variables in the research of management and Organizational Behaviour (OB). However, lapses examine its mediating role, specifically between climate for innovation and employee retention relationship. This research sets out as a cross sectional study that incorporates several theories, namely Social Exchange Theory (SET), Three Component Model (TCM) of organizational commitment, and Life-span theory. Judgemental nonprobability sampling was adopted as the method for data collection. A total of 444 academics from both public and private Higher Education Institutions (HEIs) all over Malaysia participated in this research. Data was analyzed using SmartPLS 3.25. Bootstrapping procedure was used to test the mediating effect of climate for innovation. One of the major findings of this research expressed a positive significant association of climate for innovation, organizational commitment, and employee retention. On top of that, climate for innovation was discovered to have a mediating effect on this relationship. This research is significant in terms of theoretical and practical contribution. This study adds to the growing body of research by making up for the absence of reliable evidence in the literature particularly the mediating role of organizational commitment. Together, this paper enlightens practitioners in employee retention planning.


Employee Retention
Employee retention has been carefully defined as a voluntary move made by an organization by initiating a proper environment for a long term engagement (Chiboiwa et al., 2010). Meanwhile, Neog and Barua (2014) referred it as policies and practices crafted by an organization in an effort to motivate employees to keep stay in their job. Furthermore, it is also described as a proactive and consistent effort to pull knowledgeable and competent workforce (Shakeel and But, 2015) to remain serving an organization for a maximum period (Lyria et al., 2017). In relation to this, an established study defined employee retention as a systematic effort which concern on diverse needs of employees for the purpose of earning long term employment relationship (Kumar and Mathimaran, 2017). In this research, employee retention is referred as the decision of academics to keep serving the university as a result of the university's ability to nurture the right environment (Das and Baruah, 2013); (Kyndt et al., 2009).

Review of Employee Retention Studies in Higher Education Industry
According to the recent evidence, various factors concerning the improvement of employee retention in higher education sector have been studies which include organizational climate (leadership, organizational citizenship, compensation, interpersonal relationships and clients, capacity and values) (Erasmus et al., 2015), total rewards (Akhtar et al., 2015), academic growth, intrinsic and extrinsic factors, job satisfaction factors (Selesho and Naile, 2014), organizational factors (culture, and communication) and individual factor such as motivation, opportunity for growth and development, work life balance, engagement (Ngobeni and Bezuidenhout, 2011), and organizational commitment (Darougheha et al., 2013). However, previous studies of employee retention have not dealt with climate for innovation and organizational commitment. To date, the issue related to the evaluation of organizational commitment as a mediator between climate for innovation and employee retention has received scant attention in the research literature.

Climate for Innovation
Commonly, the definition of climate for innovation is revolved around the concept of shared perception about employees' working to happen and reward related innovative behaviours practice (Schneider, 1990). On another note, (Scott and Bruce, 1994) indicated that accepted values, norms and expectations of innovative behavior are normally put into climate for innovation criteria set by certain organization. According to Somech and Drach-Zahavy (2013), climate for innovation was referred to a factor that complement team creativity and innovation practice. In the context of this research, climate for innovation is described as two dimensional aspects focusing on support for innovative behaviour aspect and as availability of adequate resources to be innovative enable them to be innovative (Holliman, 2012); (Scott and Bruce, 1994).

Organizational Commitment
Organizational commitment is expressed as the extent of employees' willingness to exert their effort for the success of the organization as well as the degree of fit between employees' values and organization's value (Mowday et al., 1979). It is also described as multidimensional in nature and strongly related to employee's loyalty Iqbal et al. (2015). Existing research defines organizational commitment as a dynamic process among staff, organization, and environment (Yigit, 2016), including the feeling of deep attachment to an organization or some of its members as a result of improvement in of job satisfaction (Mitić et al., 2016). In the context of this research, the definition of organizational commitment is described as the willingness to put certain efforts as employees have a clear and accepted belief, values and goals as a result of positive mindset towards university (Becker T. E. et al., 1995); (Meyer and Allen, 1997).

Climate for Innovation and Organizational Commitment Relationship
Organizational support perceived by an individual employee in relation to physical and psychological environment is found to influence the feeling of belongingness and commitment to organization (Luchak and Gellatly, 2007); (Reid et al., 2008). Meanwhile, result from the other researches revealed that, supportive environment for innovation was a good indicator to organizational commitment (Holliman, 2012); (Riad et al., 2016). In relation to this, climate for innovation is expected to be a well-established contributing factor for the sense of belonging and organizational commitment. Thus, the following hypothesis is developed: H1: There is a positive relationship between climate for innovation and organizational commitment.

Organizational Commitment and Employee Retention Relationship
A number of published studies (Jehanzeb et al., 2013); (Slattery and Rajan, 2005); (Wu, 2012) reported that highly committed manpower was discovered to have low turnover intention (positively related to employee retention). According to Young (2012), most of these individuals wish to remain active in their current workplace to expand the organizational goals Young (2012) hence, they seems to have no intention to exit. On the other note, employee retention has been observed to be influenced by affective, normative, and continuance forms of commitment (Van and Coetzee, 2012). This finding is further supported by a well-established stream of research rooted in Social Exchange Theory (SET) which revealed that employees' commitment to the organization is mainly derived and motivated by their perceptions regarding the employers' commitment in supporting them, or in other words, how employees perceive their organization is able to reflect the same way they do (Harrison et al., 2006). Hence, it is hypothesized that: H2: There is a positive relationship between organizational commitment and employee retention.

Climate for Innovation and Employee Retention Relationship 2.8. Relationship between Climate for Innovation, Organizational Commitment, and Employee Retention
According to Thakare and Prakash (2015), organizational climates tend to have a strong effect on organizational outcomes based on the influence of organizational processes such as the level of motivation and commitment. Govaerts et al. (2011) in their research indicated that support for innovation such as learning and open climate which encourage creativity tend to be regarded as the strong prediction in the intention to stay. However, employees' perceptions may be different subjected to their evaluation on the support that the organization gives. This is in agreement with the study by Holliman (2012) which states that the degree of organizational support in the aspects of support for innovation and resource supply strongly influence the participation and involvement of different level of employee, including their longevity with the organization. All in all, this further implies that employees' organizational commitment level is highly dependent on the perceived climate for innovation supported by their organization, in which both will reflect their decision to stay in their current organization. Thus, the following hypothesis is developed: H3: Organizational commitment mediates the relationship between climate for innovation and employee retention.

Proposed Conceptual Framework
The framework for the present study utilizes SET, TCM, and Life-span theory as the basis to describe the influence of organizational commitment on the relationship of climate for innovation and employee retention. It is important to note that social exchange develops emotional level/mindset towards positive behavior. Hence, climate for innovation is expected to lead to organizational commitment and employee retention. Figure 1 shows conceptual framework of this research.
In relation to this, an understanding of the employees' expectations on perceived organizational support in terms of climate for innovation tend to assist employers in improving organizational commitment, which is in line with SET (Blau, 1964). Meanwhile, TCM channels the researcher to access information on the three levels of organizational commitment. Most of the established studies affirmed that affective commitment has a strong effect on employee retention. Therefore, it is strongly recommended to fully understand the status of organizational commitment of employees for the purpose of assessing and enhancing organizational practices that is believed to strengthen their emotional attachment which influence their decision to remain in the same organization.

Participants and Procedures
The population of this research was permanent academics who are working on full-time basis and have served for at least six months at their current workplace. These requirements enable them to describe their perception of talent management practice at their institutions. The purpose of choosing judgmental nonprobability sampling based on the fact that the researcher knows a reliable professional or authority that he thinks is capable of assembling a representative sample (Colman and Briggs, 2002) A total of 870 questionnaires were distributed to Malaysian Higher Education Institutions (HEIs). A "drop-off" and "pick-up" method was employed as a result of Higher Education Institutions (HEIs) proximity to the researcher. Out of the 870 questionnaires distributed, a total of 468 were returned but only 444 questionnaires were usable for further analysis, which yielded a 53.8% response rate.

Measurement
In this research, the dependent variable which refers to employee retention was adopted from Kyndt et al. (2009). There were 11 items in the original scale with a Likert scale 1-5 represented by strongly disagree to strongly agree.
On another note, an independent variable of climate for innovation was measured by adopting 22 items from (Scott and Bruce, 1994) . With a Likert scale 1-5 represented by strongly disagree to strongly agree. Support for innovation measures to what extent individuals perceive their organization is supportive and tolerant of creative ideas, innovational changes and diversities of its members in problem solving. Meanwhile, resource supply measures the adequacy of the resources (personnel, funding, time) provided by the university.
Meanwhile, organizational commitment was measured by adopting organizational commitment theory developed by Meyer and Allen (1997). The theory comprised of three dimensions, namely affective, normative, and continuance commitment. The items were illustrated into 22 questions with the Likert scale 1-7 ranging from strongly disagree to strongly agree.
The present study utilized Partial Least Squares (PLS) to predict and maximize the explained variance in employee retention (Urbach and Ahlemann, 2010). Anderson and Gerbing (1988) two step analytical procedures were adopted to analyse data in PLS. The first step involves the evaluation of the measurement model followed by tests for all hypotheses using structural model. The measurement model was assessed using two types of validity, namely convergent validity and discriminant validity (Hair J. F. et al., 2011). The convergent validity of the measurement is usually ascertained by examining the loadings, average variance extracted (AVE), and the CR (Hair J. F. et al., 2014). On the other hand, the discriminant validity of the measures (the degree to which items differentiate among constructs or measure distinct concepts) was assessed based on three criteria including crossloadings, Fornell-Larcker criterion, and HTMT as suggested by Hair J. J. F. et al. (2016). In assessing the cross loading, the outer loadings of each indicator on its respective latent construct must be greater than its loadings on any other constructs (Chin, 1998). The second approach to examine discriminant validity is Fornell-Larker Criterion. The square root of the AVE for all constructs should be greater than the correlation between the constructs (Fornell and Larcker, 1981). Finally, Hair J. F. et al. (2014) suggested looking at the coefficient of determination R 2 , estimation of path coefficient (β), and the corresponding t-values via a bootstrapping procedure with a resample of 5,000. In addition to these basic measures, it was further recommended for the researchers to report the predictive relevance (Q 2 ) and the effect sizes (f 2 ). Table 1 presents the demographic characteristics of the respondents. The total of male respondents is 229 which are represented by 51.6%. Majority of the respondents represented by 45.9% are in the age range of 26 to 35 years old. Most of the respondents (46.2%) possess Master's degrees. Meanwhile, about 65.5% hold the lecturer position. In terms of work experience, majority (29.9%) of the respondents possess five to ten years work experience. In relation to the length of service, majority (48.4%) of the respondents have been working for their current institutions for duration of less than five years. In terms of marital status, 60.6% of the respondents are married. Finally, 56.5% of the respondents are working in the public university.

Common Method Variance (CMV)
In this research, there were no issues on common method bias based on the results of the analysis obtained from Harman One-factor test as recommended by Podsakoff et al. (2003). On top of that, the Principal Components Analyses returned eleven factors emerging using the eigenvalue greater-than-one rule (Kleinbaum et al., 1988) with a total variance of 74.46%, with the first factor accounting for 28.33%.

Evaluation of Measurement Model at the First Order 4.3.1.1. Convergent Validity
The convergent validity results illustrated in Table 2 show that the loadings are above 0.5 specifically in the range of 0.636 to 0.982, which further demonstrate that all indicators are reliable (Hulland John, 1999a). Meanwhile, the composite reliabilities (CR) are recorded to be higher than 0.7 (ranging from 0.908-0.977), thus suggesting acceptable internal consistency reliability (Henseler Jörg et al., 2009). Finally, the AVE for the two constructs are found to be higher than 0.5 that is in the range of 0.622-0.879, which managed to satisfy the conditions of convergent validity (Bagozzi and Yi, 1988); (Fornell and Larcker, 1981).

Discriminant Validity, Cross Loading and Heterotriate-Monotrait ratio of correlations (HTMT))
Generally, comparison is made between loadings of the construct and other construct. The value must be greater than the value of loading of other constructs. In this case, AVE was utilized to evaluate discriminant validity. On top of that, the square root of the AVE of each diagonal construct should exceed the correlation that is shared between the construct and other constructs in the model (Fornell and Larcker, 1981); (Fornell and Cha, 1994). The result of the analysis confirmed that all constructs exceed the threshold value of 0.5 (Bagozzi and Yi, 1988). The AVE values are in the range of 0.622 and 0.879, whereas the square roots of AVE that appears in the diagonal line are found to be larger than any correlation between the associated construct and other constructs (Chin, 1998). Table 3 depicts the discriminant validity results. The assessment of the loadings for each construct at the first order signifies that the measures are adequate in terms of individual validity. However, Hulland J. (1999b) suggested the use of cross loading to test discriminant validity as the second assessment in determining whether the items were loaded on another construct equally to their theorized construct. Table 4 illustrates the result of cross-loadings. It can be observed that the loading clearly separate each latent variable as theorized in the conceptual model and the correlations between indicators and latent variable are higher compared to other constructs. Hence indicates discriminant validity have been satisfied. In the third assessment, the HTMT was tested to assess discriminant validity: (i) as a criterion, the figure was compared with a predefined threshold HTMT value of HTMT.90 (Clark and Watson, 1995); (Kline, 2015). According to the results, the correlations between factors in the measurement model of all items are less than HTMT. 90, which further indicates the difference in the true correlation between the two constructs. Apart from that, the result also suggests that the discriminant validity has been ascertained. (ii) As a statistical test, the null hypothesis (H0:HTMT≥1) was tested against the alternative hypothesis (H1: HTMT< 1). The results of the HTMT inference revealed that the confidence interval values did not contain the value of one. The HTMT and confidence interval results of the research are depicted in Table 5 (in Appendix). Thus, The result of the AVE analysis, cross loadings, and HTMT managed to illustrate that the measurement model successfully displayed discriminant validity, which fulfils the criteria of Fornell-Larcker and HTMT Kline (2015).

Evaluation of Formative Model at the Second Order
This research adopted a two-stage analysis to measure the indicator weight and VIF based on the recommendation of Becker J.-M. et al. (2012). Next, the inter-construct validation was tested to determine the construct validity.

Variance Inflation Factor (VIF)
The purpose of testing VIF is to identify collinearity to indicate to which extent the variance of an indicator can be explained by other factors. As can be observed in Table 5, the VIF for five indicators of two formative second order constructs are in the range of 1.233 to 1.899, which is far below the conservative threshold value of 5. Hence, no issue of multicollinearity was detected across the indicators. Hence, there was no difficulty with the estimation of PLS model.

Indicator Validity
In this research, indicator validity is determined by testing the significance of each dimension. Table 5 shows that the weights of all items (the relationship within formative measurement model from indicator to latent variable) are above the value of 0. Several researchers seem to suggest that the path coefficient (estimation) should be greater than 0.10 or 0.20 (Chin, 1998); (Lohmöller, 1989). There is a presence of empirical support to retain the indicator based on the significant results at the respective t-value. Moreover, the indicator will only be dropped when both the outer loading and weight are insignificant (no empirical support) and the presence of weak conceptual support for an indicator inclusion as suggested by Ringle et al. (2013). Smart PLS version 3.2.5 recommended by Ringle et al. (2013) was utilised to examine the significance and relevance of indicators' weights. The bootstrapping procedure conducted using 5000 resamples was used to assess the significance of weights of the formative indicators (Hair J. F. et al., 2011); (Hair J. F. et al., 2014). Table 5 depicts the validity results for second-order formative construct. According to the analysis, significant t values was detected in resource supply, support for innovation, and affective, normative, and continuance commitment, which provide an empirical support to retain the indicators (Hair J. F. et al., 2011). The results further suggested that the two dimensions of climate for innovation as well as the three dimensions of organizational commitment have a significant contribution on the overarching construct. On another note, the strong-tie support for innovation acts as a more important contributor to climate for innovation, whereas affective commitment is regarded as the most important contributor to organizational commitment. Hence, the utilization of operationalization in this research has confirmed the uniqueness of various dimensions of climate for both innovation and organizational commitment.
As suggested by Henseler J. et al. (2015), another discriminant validity test was performed using HTMT as prescribed in the measurement model, in terms of the criterion and statistical test. As shown in Table 6, all the values suggest that the discriminant validity is ascertained. Hence, the figures indicated that reliability and validity results of each construct and its dimensions are confirmed to be categorized as formative construct the measurement model adopted in this research is valid and fit. Therefore, the hierarchical conceptualization of climate for innovation and organizational commitment have been justified for structural model estimation.

Evaluation of Structural Model 4.4.1. Coefficient of Determination (R 2 )
In the current study, PLS bootstrapping function was utilized to generate the t-statistics values. It has been wellestablished that the bootstrapping is able to generate 5000 samples from 444 cases. This evaluation was guided by Chin (1998) with the coefficient of determination (R 2 ) value around 0.67, 0.333, and 0.19 which are considered substantial, average, and weak, respectively. On the other hand, the minimum coefficient of determination (R 2 ) should be 0.10 (Falk and Miller, 1992), which ensures the nomological validity of the model. Table 7 displays the coefficient of determination (R 2 ) value for the current research. The results demonstrate that the overall model explained 57.6% of the variance in employee retention. Climate for innovation (CI) and organizational commitment (OC) explained 57.6% of the variance in employee retention (ER). Meanwhile, climate for innovation (CI) explained 35.2% of the variance in organizational commitment (OC). While as shown in Table 8, results of path coefficient and bootstrapping illustrate that employee retention can be seen to have no direct influence with climate for innovation (CI) (β=0.022, t=0.542, not significant), thus indicating that H1 is not supported. On the other hand, organizational commitment was discovered to be influenced by climate for innovation (β=0.342, t=7.516, p<0.001) which proves that H2 is supported. Meanwhile employee retention is directly influenced by organizational commitment (β=0.63, t=16.632, p<0.001) which suggests that H3 is supported.

Effect Sizes (f²)
In this research, employee retention was predicted by climate for innovation at f² value 0.001, while organizational commitment at the f² value of 0.607. In addition, organizational commitment was predicted by climate for innovation at the f² value of 0.122. According to Cohen (1988), the f² value of 0.001 was less than a small threshold value of 0.02. On another note, the f² value of 0.122 was indicated to be higher than the medium threshold value of 0.15, whereas the f² value of 0.607 was revealed to be higher than a large threshold value of 0.35.
The present research validated strong significant effect of climate for innovation represented by (β=0.342, t=7.516, p<0.001) on organizational commitment of academics in Higher Education Institutions (HEIs). This finding broadly supports the work of established studies in this area. Hence, climate for innovation was found to be important to ensure the commitment of the academics in their profession.
Referring to Table 9, strong significant effect of organizational commitment on employee retention was also confirmed (β=0.63; t-statistic=16.632, p<0.001). As supported by a well-established stream of research rooted in SET, employees' commitment to the organization was revealed to be derived from their perceptions of the employers' commitment in supporting them, or concerning the perception of employees regarding the same reflection from the part of organization (Hutchison and Garstka, 1996). Hence, committed employees who are not looking for employment elsewhere will remain longer with a positive perception about the organization (Eshiteti et al., 2013) and will be inclined to exhibit more positivity on-the-job behaviors (Harrison et al., 2006).
It is clear that the present study has validated the insignificant association of climate for innovation on employee retention in Higher Education Institutions (HEIs). On top of that, climate for innovation was measured through the two aspects of support for innovation and resource supply and proven to be important to climate for innovation at the significant weightage of β=0.751; t-statistic=11.971 and β=0.423; t-value=5.449, respectively. However, the t effect of climate for innovation on employee retention was found to be insignificant (β=0.022, t=0.542). Meanwhile, the omission of this predictor had 0.001 effects on employee retention. In other words, the retention of academics was 0.001 attributable to the presence of climate for innovation. As suggested by Cohen (1988), the figure is far from a smaller effect size of 0.02. Considering the demographic background, majority (45.9%) of the respondents were categorized as Generation Y who are generally known to be easily comfortable and confident with technology advancement (Queiri et al., 2015) and put priority more to themselves rather than to their organizations (Solnet et al., 2012); (Twenge et al., 2010). As claimed by Daly and Dee (2006), heavy workload for instance large class capacity may cause dissatisfaction and reduce commitment to organization. Other factors were most probably came from improper administration of courses, schedule and additional administrative tasks such as assignment of new courses, frequent changes of timetable, long hours of work, irregular breaks, handling students' discipline and challenging targets or deadlines, Oredein and Alao (2010). The mentioned factors are believed to harm the academics wellbeing (Metcalf et al., 2005). On top of that, a concern should also be on work-life balance as it reflects the academics retention among Higher Education Institutions (HEIs) (Karatepe, 2013); (Mukhtar, 2012). Hence, further studies should investigate factors that may align the three variables; climate for innovation that can lead to organizational commitment and employee retention.

Blindfolding and Predictive Relevance (Q²)
This research employed the blindfolding procedure to obtain predictive relevance (Q²) value. The Q² value determined the level of predictive relevance that the exogenous/predictor has for the endogenous/ dependent constructs. There is path model's predictive relevance for a particular construct when Q² values are larger than zero for a certain endogenous latent variable. The results of Q² value implied that all path models possessed predictive relevance for all endogenous construct. Following Hair et al., (2014) a threshold value of 0.02, 0.15, and 0.35 are respectively considered as small, medium, and large. Thus, Q² value for the two paths (CI  ER, and OC  ER) showed the presence of 35.3% predictive relevance (considered as large) on employee retention (ER). Finally, the Q² value for path model of CI  OC demonstrated that climate for innovation (CI) possessed 18.5% predictive relevance (considered as medium) on organizational commitment (OC).

Mediation Analysis
The adoption of PLS had led to the employment of procedure which is known as indirect effect to assess mediation. Using bootstrapping, by resampling the original data set of N using a computer, followed by the formation of new sample (called a 'resample' or bootstrap sample) known as size N. In this research, 5,000 times for bootstrapping purposes were utilized, the results of mediation are presented in Table 9. As displayed in Table 9, results from the analysis performed using bootstrapping procedure to test the mediating effect of the respective variable show a significant indirect effect β = 0.364 (0.529*0.688) with a t-value of 11.892. This indirect effect of 0.364 at 97.5% Boot Confidence interval (CI): [LL = 0.305, UL = 0.423] which does not straddle a 0 in between seems to indicate the presence of mediation (Preacher and Hayes, 2008). Therefore, it can be concluded that the mediation effect was statistically significant, which suggests that H4 is fully supported and validated in this research. Therefore, empirical and theoretical support is provided on the significant role of organizational commitment as a mediator in between climate for innovation and employee retention variables in the context of this study.

Conclusion
The notion that expressed climate for innovation was empirically tested and supported, which then led to organizational commitment that enhances employee retention. Hence, these findings make several contributions to the current literature. The result indicating the insignificant association between climate for innovation and employee retention was somewhat counterintuitive and considered as the single most striking observation to emerge from the data which should be further investigated.
Some of the issues emerging from this finding relate specifically to the benefit that can be gained by managers, which lies in the fact that organizational commitment of employees can only increase if climate for innovation is taken care well. In addition, the increase in organizational commitment will also improve employee retention.
There is abundant room for further progress by replicating the proposed conceptual framework to be tested in different sectors. On top of that, this research adopted judgmental nonprobability sampling, which indicates the generalization to the theory of the phenomenon may have wider applicability. Several other types of sampling design can also be considered for population generalizability This research was undertaken to investigate the relationship between climate for innovation and employee retention that is mediated by organizational commitment. Furthermore, an intensive literature review was conducted in assisting the development of hypotheses for the relationships. Meanwhile, an exploratory research performed utilizing interview and focus group was extremely encouraged to obtain information that was not discovered which may contribute to a central dimension to both practitioners and academicians.
The insignificant result of the association of climate for innovation on employee retention requires an attention for a thorough study in the aspects of resource supply and support for innovation. The principal limitation of this analysis can be addressed in future studies by examining other variables such as workload. This is consistent with a previous finding that demonstrated workload to be closely related to the intention to stay Ng'ethe et al. (2012). On top of that, performance achievement is recommended as one of the constructs considering that decision to remain is subjected to concern on personal achievement (Saeed et al., 2014). It is strongly connected to employees in generation Y category which is known to be achievement oriented (Alexander and Sysko, 2011).
In this research, climate for innovation was measured using Scott and Bruce (1994). Further research is suggested to revise the items to improve its mediating effect. The items are suggested to be linked with organizational commitment in to the effort of predicting employee retention.