The effect of organizational identification on job embeddedness : Evidence from new generation of rural migrant workers in China

Article history: Received: June 16, 2018 Received in revised format: June 24, 2018 Accepted: August 6, 2018 Available online: August 6, 2018 The purpose of this paper was to examine the relationships among work overload (WO), compensation (COM), organizational identification (OI) and organization embeddedness (OE) in the context of manufacturing industry in China through social identity theory. A 37-item questionnaire was filled by 384 new generation of rural migrant workers. Data were examined through a two-stage of first-order of reflective model and second-order reflective-formative hierarchical model using PLSSEM. The empirical results indicate that COM and OI positively and significantly predicted OE, while WO was found to have no direct effect on OE. In addition, COM positively and significantly was associated with OI, whereas WO affected OI, negatively. Further examination of the mediation effects of OI revealed that OI could fully mediate the relationship between WO and OE. Moreover, OI also had a partial mediator role in the relationship between COM and OE. The study concluded with several implications and recommendations for future research. © 2018 by the authors; licensee Growing Science, Canada


Introduction
Job embeddedness is originally theorized to explain why people stay in the job (Mitchell et al., 2001).It was conceptual as "the combined forces that keep an employee from leaving his or her job" and aggregate into three dimensions of fit, links and sacrifice reflecting in organization embeddedness and community embeddedness (Yao et al., 2004, p159).As one of the components of job embeddedness, organization embeddedness primarily focus on work-related elements that keep an employee in his/her current job (Mitchell et al., 2001).Employees embedded in the organization are also embedded in their current jobs.During the past decade, many scholars have suggested some links between organization embeddedness and various work-related outcomes (Lee et al., 2014;Lee et al., 2004;Ng & Feldman, 2009).The sheering number of the empirical studies about organization embeddedness are inspiring, but the construct of organization embeddedness still needs to be addressed (Kiazad et al., 2015;Lee et al., 2014).
First, the composite scale of organization embeddedness has been ignored.Considering the general assessments and statistical advantages (Crossley et al., 2007), most scholars prefer to use the global scale of job embeddedness (Zhang et al., 2012).According to Zhang et al (2012), this global measure was considered as mixing emotional and attitudinal components.In the original theory and research, organization embeddedness was operationalized as composite measurement demonstrated its advantage of theoretical richness and contributions because it explicitly includes non-attitudinal components (Zhang et al., 2012).Indeed, the composite scale completely captures the organization embeddedness construct space and it has provided an insight into the scores for organization embeddedness (Lee et al., 2014).Secondly, the construct of organization embeddedness has been accepted by researchers worldwide (Ferreira et al., 2017;Hom et al., 2017;Lee et al., 2014;Nguyen et al., 2017), but examination in a wider range of national, cultural and organizational context and respondents are still needed (Ampofo et al., 2017;Kiazad et al., 2015;Lee et al., 2014;Zhang et al., 2012).To date, in China, studies about job embeddedness are still limited and most of them are conducted among skilled employees (Li & Jun, 2009;Liu, 2011;Mai et al., 2009;Xiaowei et al., 2005;Yang & Lin, 2013;Zhang, 2011).The measurement of organization embeddedness in the previous studies primarily focus on Crossley's (2007) global job embeddedness scale, widely regarded as ignoring the concern of the three causal indicators (fit, links, and sacrifice) of organization embeddedness.Hence, it is of great importance to further explore the influence of the composite measurement of organization embeddedness and extend to a non-western context of China, and different types of respondents (i.e., low-ranking, low-skilled workers) rather than only skilled workers.
Moreover, over the past years, the majority of previous organization embeddedness researchers have focused on examining the higher predictive abilities as related to work-related attitudes and behaviors (Kiazad et al., 2015;Lee et al., 2014).Limited studies are known about what factors will promote employees greatly embedded into the organizations (Lee et al., 2014;Holtom et al., 2012;Ng & Feldman, 2011).Researchers have suggested that human resource management practices may play important roles in fostering the development of organization embeddedness.However, only few studies have aimed to evaluate this relationship deeply so far (Bergiel et al., 2009;Karatepe, 2013aKaratepe, , 2013b;;Tian et al., 2016).Thus, this study goes beyond prior studies that only look at the consequences of organization embeddedness to further explore the mechanism between these antecedents and organization embeddedness.An unresolved issue is the association of compensation, work overload and organizational identification for the explanation of organization embeddedness.This intention is important because it has theoretical interesting and serious implications for practitioners.
Theoretically, the combination of more specific evaluation of one's task and working conditions as derived from the social identity theory helps us understand the role of organization embeddedness in general.Previous studies have revealed the existing association of compensation and organization embeddedness (Bergiel et al., 2009;Karatepe, 2013a;Tian et al., 2016;Van Dyk et al., 2013).However, these studies did not reveal how compensation triggered organization embeddedness.At the same time, although prior studies have shown that emotional exhaustion could be a full mediator for the impact of work overload in organization embeddedness (Karatepe, 2013b), there are few studies providing evidences from the perspective of social identity theory and organizational identification to explore why work overload cause employees' organization embeddedness.The limitation of the theoretical explanations of job embeddedness was mentioned by Kiazad et al. (2015).Therefore, social identity theory approach will serve as theoretical background for the current study and proposed that compensation and work overload would significantly related to organization embeddedness, and the relationships were mediated through organizational identification.The proposed interplay of compensation, work overload, and organizational identification for their prediction of organizational embeddedness will be empirically tested in the context of manufacturing sector among new generation of rural migrant workers in China.
Practically, this study would be helpful to know how compensation, work overload, and organizational identification are combined to form a new generation of rural migrant workers' embeddedness in their working organizations, and have implications for the management of such attitudes.New generation of rural migrant workers who are highlighted as industrial workers (Ma & Han, 2015) were well known as important labor supply in China (Han, 2015).However, media reports in China have revealed that they have high voluntary turnover behaviors for a long time and currently evolved into short-term employment trend.In terms of new generation of rural migrant workers, who are born after 1980's and 1990's, their job duration are only 1.5 and 0.9 years, respectively (Zhang, 2016).Therefore, the present study intents to provide a new perspective from organization embeddedness to facilitate managers in the manufacturer sectors in China to reduce the new generation of rural migrant workers' high turnover behaviors.This is because organization embeddedness has been demonstrated its significant superiority in explaining turnover behaviors than traditional job attitudes variables and job alternatives (Mitchell et al., 2001).

Organization Embeddedness
As one of the components of job embeddedness, organization embeddedness is conceptualized as workrelated elements of employees (Mitchell et al., 2001).These work-related elements are described as a combined forces of psychological, social, and financial elements operating in and around the job that act "like a net or a web in which an individual can become stuck" (Lee et al., 2004).Employees who bind with the organization are embedded in their current jobs.More specifically, the construct of organization embeddedness are categorized into three critical dimensions, namely links, fit, and sacrifice (Mitchell et al., 2001).

Fits
Fit is defined as the extent to which an individual's perception of compatibility or comfort with his/her organization (Mitchell et al., 2001).It reflects the degree of an individual's personal interesting and values fit with his/her organization, and job.Good organization-individual fit is associated with an individual's personal values, career goals and plans of future, knowledge, skills, and abilities that are wellmatched with the organizational culture and the demands of his/her job.The higher likelihood employee perceived fit with his/her organization, the more he/she feels compatible, the more comfort to his /her organization and job (Mitchell et al., 2001).

Links
Links were originally characterized as the extent to which an individual formally and informally connect to a people and institutions or other people in a social, psychological, and financial web within the workplace (Mitchell et al., 2001).The interaction between people and work team members, other colleagues and their supervisors constitute the links between people and their organizations.Thus, having a good relationship with work team members or colleges, and supervisors are associated with an individual's high organization links.The more strains and important an individual connect to the web, the more possibility an individual might embedded in his/her job and organization (Mitchell et al., 2001).

Sacrifice
Sacrifice refers to the perception losses of material, psychological or social benefits when an individual leaves his/her organization (Mitchell et al., 2001).An individual gives up his/her job-related benefits if he/she leaves his/her job, such as stock options, pension compensation, colleagues, interesting projects and job advancement opportunities.The more benefits an employee might give up when he/she leaves his/her organization, the more feeling of sacrifices when he/she leaves the job (Mitchell et al., 2001).

Compensation
In current study, compensation refers to the satisfactory level of the monetary compensation, including salary and benefits.Both of the direct and indirect compensation can increase employees' motivation and attachment to the organization (Appelbaum & Kalleberg, 2000;Arthur, 1994;Bergiel et al., 2009;Tian et al., 2016).Managers view compensation as a competitive determinant to retain employee through which influence employee attitudes and behaviors and improve organizational performance.Whereas, employees view compensation as the return for the work they perform and see it as their financial security.A satisfied compensation system in the company might arouse the sense of individual about holding valued resources in the organization.Being provided with satisfied compensation is likely to increase the extent to which individuals feel the potential sacrifice associated by leaving the job.Therefore, we believe there is a direct relationship between compensation and organization embeddedness.
Moreover, the empirical relationship between compensation and organization embeddedness has been demonstrated in previous studies (Bergiel et al., 2009;Tian et al., 2016;Van Dyk et al., 2013).Bergiel et al. (2009) indicated that organization embeddedness fully mediated the relationship between compensation and intention to quit in a state department of corrections in the Southeastern USA.Tian et al. (2016) also revealed that compensation contribute to organization embeddedness in a sample of Chinese state-own firm employees.Hence, this study anticipates that, the more satisfied compensation employee gain in the organization, the more embeddedness in the organization.Thus, we propose the following hypothesis:

Hypothesis (1):
There is a positive and significant relationship between compensation and organization embeddedness.

Work Overload
The present study defines work overload as the situations where an individual feels that there are too much workload based on available time, abilities, and other constraints.Generally, work overload is considered as quantitative job demands, which emphasis on the incompatibility between job requirements, time constraints and resources associated with work that can be used to meet these requirements (Maslach et al., 2001).The mismatch of workload is usually considered to be excessive overload and too much demand exhausts an individual's energy to the extent that it is impossible to recover (Maslach et al., 2001).Therefore, when an employee recognizes heavy workload in his/her job, the perception of incompatibility or uncomfortable with his/her organization will arouse.The high demands of the job continue hinder the individual's personal interesting and values fit with his/her organization.Employees who do not fit with the organization in turn are less embedded in their jobs and display negative or poor employee outcomes.
Additionally, previous empirical studies have indicated that there is an existing relationship between work overload and organization embeddedness.For example, by conducting a study in 110 full-time frontline hotel employees and their managers in Romania, Karatepe (2013) found emotional exhaustion fully mediated the relationship between work overload and organization embeddedness.Hence, this study anticipates that, for new generation of rural migrant workers, the higher perception of work overload, the less embedded in the organization.This lead to the following hypothesis:

Hypothesis (2):
There is a negative relationship between work overload and organization embeddedness.

Organizational Identification
Organization identification is defined as 'individuals' perception of oneness with or belongingness to an organization, where the individual defines him or herself in terms of organizations in which he/she is a member' (Mael & Asiforth, 1992).It has been widely described as an important psychological state that reflects the potential intrinsic link between employees and the organization and, therefore may be able to explain and predict many key attitudes and behaviors in the workplace (Edwards, 2005).In fact, organizational identification is originally constructed from the theory of social identity which emphasize that individual prefer to enhance their social identity through categorizing themselves in a positive identity (Fuller et al., 2006).By categorizing themselves into different groups, such as gender, age cohort, race, nationality, people achieve their salience within these out-groups.Consistent with the social categories between different groups, people continue to compare with different members in the same group.Based on their uniqueness or status in the particular group, individuals tent to maintain their identity through the intergroup comparisons.The negative and positive intergroup comparisons are able to affect their self-esteem (Ashforth & Mael, 1989).Thus, individuals' particular categories in the group provide a source of positive identity and finally enhance their positive self-image (Edwards, 2005).
In fact, organization is a typical source that can assist people to assess self-worth (Tyler, 1999).Organizational identification is a specific form of individual's social identity in the context of organization (Ashforth & Mael, 1989).Organizational identification occurs when people effectively categorized themselves into the organization for which they work (Ashforth & Mael, 1989).It is viewed as a cognitive process reflecting in the perception of similarities between individual's self-concept and the organizational identity (Dutton et al., 1994).The more an employee copes with his/her organization, the more organization's values, norms, and interests are incorporated in his/her self-concept (Ashforth & Mael, 1989;van Knippenberg, 2000).Indeed, organizational identification reflects the process of the gradually integration and congruence of values and goals between individual and organization (Hall et al., 1970;Schneider et al. 1971).Employee with high identification in the organization displays high psychologically intertwined with the organization, as well as values and emotional attachment to his /her organization, which therefore lead to employees' attitudes and behaviors in the workplace (Tajfel, 1978).
Therefore, people tent to identify with their organizations which are perceived to be prestigious because membership in a salient organization enhances self-esteem (Fuller et al., 2006).However, individuals not only evaluate their membership status in the eyes of others outside the group, but also assess their own identity in the eyes of others in the group (Tyler & Blader, 2003).Thus, the perceptions of individual's status within the group, which has been termed as "respect", are key sources that will determine the extent to which the individuals cognitively merge their self-identity with the organizational identity (Tyler & Blader, 2003).The perception of "respect" can enhance employees' self-evaluation, and shape their self-esteem in the organizations (Pierce & Gardner, 2004).There are a wide variety of information (i.e.organizational structure, police, culture, and norms) in the organizations are viewed as cues of organizational "respect" for the individuals (Fuller et al., 2006).Specifically, satisfied levels of compensation are one way to recognize employee accomplishments and likely to be noted as an important signal that employees assess they are valued within the organization.It is also served as an important role in determining employees' social economic status and the ability to provide security.The extent to which employees receive compensation from the organization will influence their perception of "respect" within the organization and convert into their self-esteem.As noted by Doby and Caplan (1995), rewards serve to convey the employees' worth to the organization.Therefore, this study proposes that compensation will facilitate employees' identification in the organization, which leads to the following hypothesis:

Hypothesis (3):
There is a positive and significant relationship between compensation and organizational identification On the other hand, it is likely that employees will experience more workload when they work in the mechanistically designed organizational system.The more mechanistically designed organizational system, the more likely that employees experienced low levels of organization-based self-esteem (Pierce & Gardner, 2004).Organizations which are characterized as high level of system-imposed control through a division of labor, rigid hierarchy, centralization, standardization, and formalization are more likely to promote the perception of employees' incompatible within the organization context (Pierce & Gardner, 2004).By way of contrast, employees who are working in less structured, non-routine technologies, high involvement and supported designed organization tend to see themselves as valuable resource in the organization.Therefore, as employees experience high levers of workload, there is an increased likelihood that they will attribute low level of self-esteem and lead to negative identification in the organization.The above discussions lead to the hypothesis:

Hypothesis (4):
There is a negative and significant relationship between work overload and organizational identification Moreover, employee with high identification in the organization shows psychologically intertwined with the organization, as well as value and emotional significance attached to his /her organization.High OI would help individuals develop more cooperation and interaction among co-workers (Dukerich et al., 2002), which prompt individual to develop "links" within the organization.In addition, identification in the organization reinforces the relationship between "socialization" and "internalization", facilitates individual's feeling of loyalty and commitment with the organization, and enables the internalization of organizational values and beliefs (Ashforth & Mael, 1989).A meta-analysis from Riketta (2005) also pointed out that an individual's fit to the organizational values, norms, beliefs and practices as well as emotional attachment to the organization is associated with organizational identification (Riketta, 2005).Hence, individual with higher level of organizational identification might better fit with his/her organization.The stronger organizational identification is likely to increase the links, fit and sacrifice components which comprise organization embeddedness.Furthermore, previous empirical study conducted in an India among executives has indicated the strong association among organizational identification and job embeddedness (Ghosh, 2015).However, the existing empirical study on this linkage is still limited.Hence, this study anticipates that the high perception of organizational identification, the more embeddedness in the organization, which lead to the following hypotheses:

Hypothesis (5):
There is a positive relationship between organizational identification and organization embeddedness Hypothesis (6): Organizational identification will play as a mediating role on the relationship between compensation and organization embeddedness

Hypothesis (7):
There is a mediation effect of organizational identification on the relationship between work overload and organization embeddedness

Research Framework
Guided by the prior literature reviews, this study proposed the research framework which was depicted as Fig. 1.

Respondents and procedure
Respondents in the current study targeted new generation of rural migrant workers who are working in manufacturer sector of China.New generation of rural migrant workers were well known as an important part of labor supply in China (Han, 2015).They were highlighted as industrial workers (Ma & Han, 2015).To some extent, the high-speed economic development of China can be attributed to labor force comparative advantage (Perkins, 2012) where a large numbers shifted from farming labor to non-agriculture sectors.The large number of shifted from farming labor is called rural migrant workers and provided low cost labor force for industries in mainland China, and promote the society transform from an agricultural to an industrial society (Li & Li, 2007).There were about 281.71 million rural migrant workers accounting for 36.3% of the total labor force (776.03 million) at the year of 2016 in mainland China (China National Statistics Bureau, 2016).The number of new generation of rural migrant workers was accounted for 140 million, about 49.7% of the total number of rural migrant workers (China National Statistics Bureau, 2016).Unfortunately, rural migrant workers have shown high voluntary turnover behaviors for a long time and currently evolved into short-term employment trend.In terms of new generation of rural migrant workers, who were born after 1980's and 1990's, their job duration are only 1.5 and 0.9 years respectively (Zhang, 2016).
Data was collected from November 2017 to March 2018 in 7 factories in Guizhou province which is located at southwest of mainland China.By assistance of the managers from each human resource department in the factory described the purpose of this study to the new generation of rural migrant workers, the respondents complete the questionnaire voluntarily and anonymously within two days.A total of 600 questionnaires were distributed and 384 questionnaires were finally selected for this study.The demographic descriptive of the respondents were showed as Table 1.(2008) "salary and benefits" scale.Four items of work overload were adopted from Karatepe's (2013) work overload scale, one item was adopted from Boxall and Macky's (2014) role overload.Organizational identification scale consisted of six items from Mael & Asiforth (1992).Answers were elicited on a 7-point Likert scale ranging from 7= 'strongest agree' to 1 = 'strongly disagree'.The scales of these variables are originally written in English, and accomplished into Chinese version through a multistage translation and back-translation procedure suggested by Brislin (1980).

Results
For data analysis the current study employs the smartPLS 3. The smartPLS-3 is gaining popularity day by day due to its model parsimony and many studies used this software for accurate results (Imran et al., 2017;Imran et al., 2018aImran et al., , 2018b)).Data received from the respondents were loaded into Partial Least Squares Structural Equation Model (PLS-SEM) to test the hypotheses proposed before.Measurement model and structural model were assessed in this study.Reliability and validity assessment were provided in the measurement model evaluation, while path coefficient and mediation test were assessed in the structural equation model.

Assessment of Measurement Model
The present study applies a two-stage approach of first-order reflective model and second-order reflective-formative hierarchical model to assess the measurements.

Reflective Measurement Assessment
Internal consistency, indicator reliability, convergent validity and discriminant validity are tested to examine the reflective measurement in this study.The internal consistency and indicator reliability of the reflective model are assessed with Cronbach Alpha (CA), significance of item loadings, and Composite Reliability (CR) (see Table 2 and Table 3).Based on these findings, we can see that the reflective measurement achieved the value higher than the recommended criterion of 0.7 for CA (George & Mallery, 2003) and CR (Nunnally & Berstein, 1994) (see Table 2).Table 3 depicts the factor loadings in this study.The loadings of item OEL1 (0.699) and OEL5 (0.581) are below 0.708, but higher than 0.4.According to Hair et al. (2016), indicators with outer loadings between 0.40 and 0.70 should be considered for removal from the scale only when deleting the indicators leads to an increasing of composite reliability above the suggested threshold value.However, in current study, composite reliability and AVE have achieved the threshold values.Hence, we still consider retain this item in resent study.
On the other hand, we examined the average variance extracted (AVE) to assess the convergent validity, whereas test the cross-loadings, Fornell-Larcker and Heterotrait-Monotrait ratio (HTMT) to evaluate the discriminant validity.The AVE value of the constructs in this study (see Table 2) are much higher than the criterion of 0.50 indicating the construct have convergent validity (Fornell & Larcker 1981).Table 3 shows that the loadings of the indicators are much higher in their own constructs than on other construct in the model.Table 4 indicates that the square root of AVE (diagonal) is higher than the correlations (offdiagonal) for all reflective constructs.Table 5 reveals that all the values fulfill the criterion of HTMT (0.9) (Gold et al., 2001).Therefore, the results of Table 3, Table 4, and Table 5 all confirm the discriminant validity of the measurement in this study.

Formative Measurement Assessment
In this stage, the study assessed the potential collinearity problem through the estimation of variance inflation factor (VIF).A variance inflation factor (VIF) with 5 or higher respectively indicates a potential collinearity problem (Hair et al., 2011).Additionally, the study investigated the significance for outer weights of each indicator to the formative construct.The results have shown that all the indicator of OE (OEF, OEL, and OES) were significant to its formative construct.The second-order, two-stage approach formative measurement model and findings are demonstrated in Table 6.Therefore, the current study has achieved all threshold values of the formative measurement), and a further analysis of the structural model can continue.

Assessment of Structural Model
Following Hair's et al. (2016) procedure of structural model assessment, this part has investigated structural model collinearity issues, significance and relevance of the structural model relationships, R 2 level, f 2 effect sizes.A VIF value of 5 or higher is considered as critical levels of collinearity (Hair et al., 2011).Table 7 indicated the assessment of the Structural Model for collinearity issues.From Table 7 we understand that the VIF values in current study were clearly below the threshold of 5. Hence, collinearity among the predictor constructs is not a critical issue in the structural model, and we can continue investigating the results reports.

The Coefficient of Determination (R 2 ), Effect Size (f 2 )
With respect to evaluate the model's predictive power, the study has assessed the coefficient of determination (R 2 value).The value of R 2 represents the exogenous latent variables' combined effects on the endogenous latent variable (Hair et al., 2016).R 2 values of 0.75, 0.50, or 0.25 for endogenous latent variables have been described as substantial, moderate, or weak coefficient of determination respectively (Hair et al., 2011;Henseler et al., 2009).In addition, this study has evaluated the effect size of the predictor constructs using Cohen's ƒ 2 (Cohen, 1988).The value of ƒ 2 [ƒ 2 = (R 2 include-R 2 exclude)/(1-R 2 include)] assesses how strong the exogenous variable's contribution into R 2 values of the endogenous variable.The ƒ 2 value of 0.02, 0.15, and 0.35 represent small, medium, and large effects of exogenous latent variable respectively, while effect size value of less than 0.02 indicate that there is no effect (Cohen, 1988).Table 8 illustrates the coefficient of determination (R 2 ) and its effect size (ƒ 2 ).

Assessment of Structural Model Path Coefficients
Smart-PLS provides an inner-model evaluation of the path coefficients, which represents the hypothesized relationships among the latent variables.The bootstrapping procedure enables researchers to compute empirical Beta-values, T-values and P-values for all structural path coefficients.Recent study used two-tailed tests of significance level 5% for testing the T-values.Thus, a critical value of 1.96 is considered as the threshold to examine the hypotheses in our current study (Hair et al., 2016).In addition, researchers have claimed that using T-values and p-values is not sufficient when reporting the significance of the structural model relationships.When performing bootstrapping test, the confidence intervals bias corrected the results of the upper and lower bound should be out of 0 point (Hair et al., 2016).Table 9 indicates the results of the structural Model Path Coefficients.From Table 9, we can observe that the effects of COM on OE were significant and positive (β=0.516,t=11.042,p=0), supporting hypothesis 1.The effect of COM on OI was significant and positive (β =0.275, t=5.179, p=0), supporting hypothesis 2. The effect of OI on OE were significant and positive (β =0.217, t=4.218, p=0), supporting hypothesis 3. The effect of WO on OI was significant and positive (β =-0.294, t=5.103, p=0), supporting hypothesis 5.However, the effect of WO on OE was not significant (β=-0.041,t=0.701, p=0.483).Thus, hypothesis 4 was rejected.

Testing Mediating Effect
According to Hayes and Scharkow (2013), bias-corrected bootstrap confidence interval is the best approach for detecting mediating effects.According to Hair et al., (2013), bootstrapping the sampling distribution of the indirect effect is the simplest approach that researchers can implement.We have performed the steps of mediator analysis procedure suggested by Hair et al. (2016) and the study combined the evaluation of the direct and indirect effects of the mediation model from PLS-SEM to determine the mediating effects and its types.Table 10 indicates the indirect effects in the mediation model.(Preacher & Hayes, 2004, 2008), confirming hypothesis 6 and 7.
On the other hand, Zhao et al. (2010) provided deeper insights into the mediation analysis and identified mediators into five types.From Table 9, we noticed that, in current study, COM was positively and significantly (β=0.516,t=11.042,p=0) associated with OE, while WO did not have any significant relationship (β=-0.041,t=0.701, p=0.483) with OE.Regarding to the definition of complementary mediation (mediation effect and direct effect both exist and point at the same direction) and indirect-only mediation (the indirect effect is significant but not the direct effect), we can conclude that OI play the complementary mediation (partial mediation) and indirect-only mediation role in the relationship between COM and OE, and WO and OE respectively.

Summary of Findings
This paper investigated the impact of compensation and work overload on organization embeddedness through the theoretical framework of social identity and organizational identification.The authors argued that compensation and work overload facilitate organizational identification, which in turn enhance organization embeddedness.The empirical study has shown that COM maintained a positive and significant relationship with OE (β=0.516,t=11.042,p=0.000), which was consistent with previous studies (Bergiel et al., 2009;Tian et al., 2016;Van Dyk et al., 2013).However, WO did not significantly predict OE (β=0.041,t=0.701, p=0.483), which was congruent with prior study (Karatepe, 2013b).On the other hand, this study has found that COM (β=0.275,t=5.179, p=0.000) and WO (β=-0.294,t=5.103, p=0.001) predicted OI, significantly.These findings filled the gap in the literature about the linkage between COM, WO and OI.Furthermore, this study demonstrated that organizational identification did positive and significant impact on organization embeddedness.The findings were the same with the previous studies (e.g.Ghosh, 2015).Finally, the current study has found that organizational identification paly as a partial mediator between the relationship of COM and OE, whereas it had a full mediator between the relationship of WO and OE.To all, the study has provided some evidences and it has demonstrated the authors' arguments were true.It particularly fills the gap in social identity theory on how organizational identification triggers the evaluation of one's task and working conditions into organization embeddedness.

Contributions
Theoretically, current study contributes to the organization embeddedness literature.In particular, it sheds light on the extending examination of organization embeddedness to a wider range of national, cultural and organizational context and respondents.The previous studies are usually focused on skilled workers in western context, whereas this study is one of the various papers that addressed the issues in an Asian country with low-skilled workers (i.e.new generation of rural migrant workers) in the manufacturer sector of China.To the best of our knowledge, this study is the first paper to investigate the relationship between compensation, work overload, organizational identification, and organization embeddedness in the manufacturer sector among new generation of rural migrant workers in China.Our contribution includes the incorporation of the social identity theory in job embeddedness studies.The study illuminated how organizational identification in the manufacturer sector acts as a mediating role between compensation, worker overload and organization embeddedness.The concept of organizational identification is increasingly playing an important role in organization embeddedness.The findings of the factors (i.e.compensation and work overload) which promote and enhance the relationship between organizational identification and organization embeddedness are also important contribution of this paper.Our study is an important addition to job embeddedness models in use in this field of research.
Practically, this study has important implications for practitioners and managers in the manufacturer sector of China.As more and more manufacturers in China are facing high voluntary turnover problems among new generation of rural migrant workers, this study should of interest among managers in this context.This is because organization embeddedness has been demonstrated its prominent ability to prevent employees' turnover behaviors (Mitchell et al., 2001).Employees who embedded in the organizations may show low turnover intentions.This study shows that the role of organizational identification as a strategic pillar has significant implications for the manufacturer sector among new generation of rural migrant workers, because it is helpful for enhancing organization embeddedness among new generation of rural migrant workers.This information also facilitate the particular implications for providing satisfied compensation and suitable workload to new generation of rural migrant workers in the manufacturer sector in China, because it is useful for them to identify their memberships in the working organizations and finally embedded them into the jobs.Furthermore, the indirect-only mediation of organizational embeddedness between work overload and organization embeddedness indicates that work overload is not one significant factor influencing new generation of rural migrant workers' organization embeddedness, but if the workers have good experience, suitable work overload in the working place would also trigger their high identification with the organizations and enhance their organization embeddedness.Therefore, manufacturers in China should encourage managers to provide supportive implications for new generation of rural migrant workers to help them identify their memberships in their organizations.

Limitations of this study
The present study is one of the first papers empirically examining the relationships between compensation, work overload, organizational identification, and organizational embeddedness.This study was exploratory in nature and subject to several limitations.The data collection was limited to the context of manufacturer sector among new generation of rural migrant workers of China.Thus, future study could generalize it to different context and different industries.Additionally, this study is a cross-sectional study.A longitudinal methodology might be necessary to validate this study in future.Thirdly, future studies can combine other important variables into the research model.Other high commitment management practices (status within the organization) and prestige antecedents (status of the organization) should be incorporated in this research model and show their functions in facilitating employees identify themselves in the organization and finally triggering their organization embeddedness.Particularly, studies including "visibility within the organization, growth opportunities, participation in decision-making, visibility of the organization, success in firm goal achievement, average status level of employees, and prestige level of the employee" as independent variables in the research model would be useful to enhance our understandings of what factors can promote organization embeddedness through organizational identification.Finally, as China is a collectivist society, a cross-cultural research in future studies would be helpful.

Fig
Fig. 1.Research Framework Mitchell et al., (2001)ss were operationalized fromMitchell et al., (2001), and Yang and Lin's, (2013) organization embeddedness scale.A final scale of 17-item of organization embeddedness were implied for the present study.Four items of compensation were operationalized from Ellenbecker et al.

Table 2
Findings of the measurement model(First-Order, Reflective)

Table 6
Findings of Measurement Model for Formative construct (Organization Embeddedness) From Table8we can observe that COM, WO and OI can describe 42.6% variance of OE in manufacturing sector among new generation of rural migrant workers in China, while COM and WO only explain 22.5% variance of OI.Additionally, COM has a relatively big effect (ƒ 2 =0.363) on all exogenous variables' (COM, WO, and OI) contribution to the values of OE R 2 .However, WO did not show any effect (ƒ 2 =0.002) on the endogenous variable's (OE) R 2 values.Moreover, both COM and WO have small to medium effect (ƒ 2 =0.083; ƒ 2 =0.095) on their contribution of OI R 2 values.

Table 9
Results of the Structural Model Path Coefficient Table 10, we realize that indirect effect of COM on OE through OI was significant and positive (COM → OI → OE, β=0.060, t=3.347), supporting hypothesis 6.The indirect effect of WO on OE through OI (WO → OI → OE, β=-0.064,t=3.140) was significant but negative, supporting hypothesis 7. The indirect effects of 95% Boot CI Bias Corrected are:[LL=0.031,UL=0.098], and [LL=-0.10,UL=-0.031] in the relationship with COM → OI → OE and WO → OI → OE do not straddle 0 in between indicating there are some mediations