Abstract: As the carrier of intellectual capital, the talents of the organization become more significant than ever. The overall objective of this study is to explore an integrative multi-route model between the talents Key Degree Scale (KDS) and withdraw tendency and will discuss the relationship with the influence of job coupling based on theory of social capital. This study mainly focus on the fitness of the proposed model, mediator variables influences of job coupling to withdraw tendency and the interaction of adjustment of moderators with the samples from three Business Schools of China. Path analysis is used to evaluate direct and indirect associations among observed variables referenced the LISREL software programs. OSL hierarchical regressive analysis in SPSS is employed for the complicated relationship of multi-collinearity among the observed variables because of the higher inter-correlation among the observed variables. Analyzed results showed that the effective retaining is a perplexing mediating mechanism not only from on job influences but also from the community factors. The relationship between talents KDS and withdraw tendency is a multi-route system with duality (retention and turnover) and many incentive sectors in which the mediator position of job coupling is extremely prominent. The relatively high homogeneity of MBA samples with more similar features of occupation and population, although from different industries and representative location, thus the application range of the analysis result should be treated carefully to expand into other special industry. The novelty comes from the three issues as follows, firstly, it is a new viewpoint of the social capital to evaluate the problem of talent turnover/retention and proposed and evaluated an integrative turnover/retention model; the second value is the typical sample design of MBA candidates for talent retention in cross-industries and thirdly, this study is a significant application for LISREL in the area of human capital management of social science.

INTRODUCTION

Intellectual capital and knowledge have become critical components of wealth creation during the information era. The dependence of the organization in the intelligence intensive economy pre-dominated by talent capital for competitive advantage and the traditional functional management modes may be influenced. The organization may pursue a more accurate, agile and network-organizational structure so that the organization crosses over the traditional organization boundary of human capital allocation and the internal and external blending and interactive network cooperation pattern may be formed (Wang, 2005). Considering the effective construct of the organization’s core competitive advantages and to obtain resources ability in the network, the organization needs talents more urgently than ever which may be the relatively stable core competitive advantages carrier because the supply of the special characteristic talents with high performance may be unable to follow the market demand growth. According to the investigation report published, there are 43% of essential position personnel and organizational leaders for voluntary turnover in China; comparatively, it is only 5 and 11%, respectively in Singapore and Australia.

Some scholars (Liang and Liao, 2007) believe that the job coupling background on the social capital can deepen the study of talents retention.

Under the present background of intensifying HR competition, the Talent War may be increasingly fierce regardless of how the labour force demands may fluctuate (Capelli, 2000). With the loss of organizational talents, particularly the accompanying essential staff loss in the network nodes, serious damage will be done to the organization’s social structure and social capital could even be destroyed completely. Therefore, regardless of whether the organization is based on knowledge, to control the replacement cost, to retain talent with high organizational performance characteristics and maintain social capital may be vital organization goals (Dess and Shaw, 2001).

However, as for the dynamic development of organizations, employees will not be in some fixed position forever and employee movement behaviour will occur frequently. Rice (2007) regarded this kind of employee movement as a social process in which transfers occur between positions or organizations. The personnel transfers that occur in organizations are inevitable and necessary in the organizational resource allocation and management.

Therefore, the overall objective of this study is to explore an integrative multi-route model between talents’ key degree scale and withdraw tendency and will be discussed it with the influence of job coupling based on theory of social capital.

Job coupling and social capital: Generally, the coupling idea or view comes from a new American economic sociology’s epistemological foundational concept which describes the relationship between economy and society. It means that economic activities should be rooted in the social structure but the core of the social structure should be the social network of people living and coupling. Attention should be paid to analyze the people’s ability with social resources instead of the holding degree to certain specific social resources before (Polanyi, 1957; Zhang and Li, 2005; Li and Zheng, 2007). As demonstrated by mass experience studies, economic activities such as transaction, employment, financing and investing, in people’s daily social life are not random market transaction patterns of individual styles drawn out of classical economics but they may be coupling in social relations (Zhang and Li, 2005).

Social capital as a terminology appeared at first as the community centre deliberation given by Hanifan (1916) which can be described as the most valuable intangible substance in people’s daily life. Modern social capital theory began in the 1970s and rapidly developed in the 1990s. In the 21st century, social capital becomes the resources and information pools which may be brought about by the actor's social relation network. However, as for social connections, trust, standards and the value system, they are the generalized social structures and simultaneously also constitute social capital attributes characteristics in which social capital may be coupled in these structures. Therefore, social capital may enable the organization to create values, solve problems, achieve goals and realize their missions (Nahapiet and Ghoshal, 1998; Guo and Zhu, 2006). From the viewpoint of integrating social capital with organizational knowledge management, Nahapiet and Ghoshal (1998) defined social capital as the social individual’s total sum of practical or potential resource which be embedded into the individual social life and obtained through their relation network. Leana and Van Buren III (1999) defined as the whole resources reflecting the organization social relation characteristics which contains the official and unofficial structures in a traditional organization. The social capital definition among organizations may also be extended as a summation of reflecting the social relations characteristic resources among the interesting group crossing the organization boundary which reflects the organization’s ability and cooperation style through the social network, for example, the organizational strategic alliance, gaining the target resources and creating the value (Guo and Zhu, 2006) formed through the network. Therefore, the organization social capital may be defined as the sum of a new organizational net system with the value-created ability and whole resource reflecting the social relation characteristics coupled inside. Social capital may be integrated by the three-level in the active organization individuals within the organizations and among the organizations.

Evolution of turnover/retention model and new development: The organizational behavior school continually focuses research on employee’s retention/ turnover for the last 60 years. In general, the turnover model may be divided into 2 central parts: the classic school with traditional intermediary variables of job attitude and the New Multi-routes. The classic school, we can trace to the source of function of the executive by Barnard (1997) which is the founding work of the social system school in organizational management. March and Simon (1958) published the classic work organization which laid the theoretical foundation for the organizational equilibrium theory and they are the earliest figures who tried to integrate labor market and individual behaviors for investigating and studying employees turnover behaviors. The Organizational Equilibrium Theory indicates clearly that the movement desirability and apperceived mobility by employees are the most important theoretical precursor variables for their turnover behavior.

Lots of scholars follow this direction, for instance, the psychological process model of Price (1977) with job satisfaction as the direct mediator variable for employees voluntary turnover; the model of Extended Media Chain which put forward by Mobley et al. (1979) and Steers and Mowday (1981)’s turnover model with the introduction of organizational commitment as a mediator variable; taking job satisfaction as the key measuring indicator and reject treating turnover as a process of continuous psychological changes.

Fig. 1:	The traditional turnover model

Fig. 2:	The integrated mediated multi-routes model

Hausknecht et al. (2008) listed the major 12 retention factors from the literature review over the last 60 years which help explain why employees stay or quit.

Griffeth et al. (2000) conducted a review research for the classic school describes that related variables around attitude models reached 11 kinds of demographic predictors; 16 kinds of sub-structure variables related to job satisfaction and organization factors and work environment factors; 6 kinds of variables related to job content and external environment factors; 3 kinds of other behavioural predictors; 9 kinds of adjusting variables for withdraw process. A general research model and analyzing route for traditional attitude research model are shown in Fig. 1.

With regards to the new multi-routes school in the 1980s, Porter and Steers (1973) noticed that the influence of different performance levels, stressed the necessity to take job performance as a factor of turnover. Allen and Griffeth (1999, 2001) included performance level in their model summarizing it into three categories as shown in Fig. 2. This clearly identified the simultaneity effect of the determinant factors contained in movement desirability and apperceived mobility between employees performance and turnover facilitate the practice of organizational behaviors with a multi-route platform to improve the effect mechanism on performance to withdraw tendency and turnover.

Lee and Mitchell (1994), Lee and Maurer (1997) and Lee et al. (2004) put forth and improved the unwrapped job coupling model of voluntary turnover by empirically comparing it with the mediator variables of the job attitude model.

They discovered that the interpretation of job coupling as a structure variable to employees retention where turnover behaviors are higher than job satisfaction and organizational commitment of the attitude model. Job coupling can be not only as the predetermined variable causing the employee to stay or leave but also as the strong effect variable affecting the employee’s performance.

Therefore, in the proposed model as shown in Fig. 3, this study induct the job coupling as a mediator between talents performance character and withdraw tendency, canceled the others less significant.

Fig. 3:	Proposed performance withdraw tendency model

MATERIALS AND METHODS

For ensuring the research work's objectivity and the controllability of obtaining the investigation data quality and the serviceability of the selected investigation samples as far as possible, this research investigation object sample designation will be limited to the specific talent potential community-150 MBAs from 3 Business Schools in China, got satisfied respond are 108. From view of performance and based on the proposed model shown in Fig. 3, the proposed hypothesis of this research includes the following eight items of mediator variables and media routes:

The expansion hypotheses for the interactive effect of model-adjusting variables are:

RESULTS AND DISCUSSION

Primary analysis: The total sample quantity is 150 MBA students of which 108 questionnaires are satisfied. These respondents were between 27 and 42 years, averaging at 34.9 years old and women accounted for 46.0%. The occupational service extends from 7-24 years and the service in the current organization varies from 5-12 years. In addition, SPSS is used for conducting factor analysis, as shown in Table 1 on job coupling structure variables and internal reliability analysis on the internal consistency of questionnaire variables. Based on the results of analyses in Table 1 and 2, Cronbach’s alpha (α) is the average of all possible split-half coefficients resulting from different ways of splitting these scale items.

The coefficient varies from 0-1 and a value of >0.6 generally indicates satisfactory internal consistency reliability (Malhotra, 2004). Result, α values of job coupling variables are all >0.7. Though, the general job coupling has high factor loading in all three factors, organization-linkage and community- linkage variables with high factor loading among its structure variables are distributed in F₃ and organization-fitness and organization-sacrifice which belong to on-job coupling are distributed in F₁; community-fitness and community-sacrifice which belong to off-job coupling are distributed in F₂.

Table 1:	Job coupling variables rotated component matrix (N = 108)

This finding reveals that job coupling measuring indicator has relatively high internal consistency reliability or validity in this type of organization samples. The result is consistent with similar analysis of Lee et al. (2004).

Correlation and OLS multi-regression analysis: For the result of coefficient matrix analysis as shown in Table 3 on the structure variable of sample data, it is obtained by using SPSS statistical software and targeted at investigating the correlation of the sample’s main structure variables.

The result shows that the traditional job attitude variable (job satisfaction and organizational commitment) and withdraw tendency exhibit significant negative correlation (p<0.05 and p<0.01). On-job coupling and off-job coupling exhibit more significant negative correlation to withdraw tendency from organizations (p<0.01 and p<0.01). Job coupling exhibits significant positive correlation with job satisfaction and organizational commitment. KDS is both significantly correlated with on-job coupling (p<0.01) and off-job coupling (p<0.05) but is not so with job satisfaction, organizational commitment and withdraw tendency from organizations (p>0.10).

Table 2:	Statistic information of sample questionnaire (N = 108)
Population and occupation variables, except age, current org-years and career years are all Dummy variables, mean ε (0, 1), Female, unmarried are taken as default values

Table 3:	Sample’s structure variables’ correlation matrix (N = 108)
1: non-local withdraw tendency; 2: local withdraw tendency; 3: non-local apperceived mobility; 4: local apperceived mobility; 5: key degree scale; 6: org-commitment; 7: job satisfaction; 8: off-job coupling; 9: on-job coupling. *p<0.05; **p<0.01; Two-tailed tests

As is indicated, job coupling may be of stronger influence on talents withdraw tendency compared with the traditional attitude variable and is more closely correlated with talents KDS.

Therefore, it needs to further conduct Multi- regression analysis on sample data to investigate the net influence of introduction of job coupling variable to the withdraw tendency. In this study, multi-regression by the basis of SPSS statistic software on various variables and withdraw tendency, draw on the thought of hierarchical multi-regression analysis, design 5 linear OLS regression model with the dependent variable withdraw tendency from organizations:

According to the definitions of Lee et al. (2004) on-job coupling and off-job coupling variables are adopted as the averages of their own 3 structure variables, respectively since the collinearity of general job coupling with its structure variable OLS regression is conducted next step, through investigations of significance, from standard regression coefficient of corresponding variables and R² compared to Model 2. This study compares the significance of the variables newly-introduced in each model which influences to talents withdraw tendency as shown in Table 4. The results show that the F-test value and D-W test value both meet the requirement of effectiveness.

Discussion and revelation of OLS multi-regress result: In terms of the relation between the population, occupation features and talents’ withdraw tendency from organizations of Model 1, most respondents are 28-45 years old in the prime period of their career. It has formed stable dependence on the career development with stable social linkage. With their linkage, they can get various resources. The population factors have relatively strong influence on the withdraw tendency such as gender, marriage, education and length of career service and rewards etc., respectively exhibit positive or negative effects. Along with introduction of other variables in the investigated samples, the variables of population and occupation show less influence on withdraw tendency of Model 2 with the introduction of traditional movement desirability and apperceived mobility. Meanwhile, the KDS as the independent variable which reflect the degree of talents performance character has been introduced exploratively but not significant.

In Model 3 and 4 with introduce of job coupling element such variables are all showed significant negative effect to the withdraw tendency, obviously elevate the coefficient of determination R² from 0.17-0.52. This might explain the responders who are in their golden career period are more profoundly dependent on their on-job and off-job social resources and affections they have coupled. Wherein, linkage factors actually characterize the combination of categories and scale of social relations what they have to face and choose in their actual organizational social life which is the base for forming their individuals social capital and meeting their various requirements (Mitchell et al., 2001a, b).

In terms of the basic functions of organizational social capital such factors of non-affective social relation are considered as indispensable context factors in the socialization process of talents’ job satisfaction and organizational commitment (Watson and Papamarcos, 2002).

Table 4:	OSL regress analysis results for withdraw tendency (N = 108)
Significance “+” means p<0.1 ; “*” means p<0.05 ; “**”means p<0.01; R2 ANOVA, p<0.000; two-tailed tests

Fig. 4:	Routes analysis result of talent’s KDS withdraw tendency model

The introduction of job coupling variable has lowered the significance of job satisfaction and organizational commitment exhibit in the Model 3 and 4 which exhibit an effect that job coupling has more influence on the interpretation. This seems to mirror the above views. Model 5 with the introduction of on-job coupling and off-job coupling show an effect similar to Model 3 and 4.

Moreover, its significance of net effect in influencing withdraw tendency from organizations shows the relative independence and this is similar to the research result of Lee et al. (2004).

Evaluation of model fitness: In the process of the routes analysis of the hypothesized model, goodness-of-fit of LISREL for the hypothesized route relation of the model shown in Fig. 4 is first investigated according to the addition strategy mentioned before in allusion to the mediator route relation of the model variables which are to be tested.

As is indicated by the fitness analysis result of LISREL, the route relations of selectable model variables are tried step by step which is judged on the basis whether χ² is improved or significantly improved (Allen and Griffeth, 2001).

In this way, the following route relations are added successively: talents KDS to withdraw tendency directly, off-job coupling, respectively to job satisfaction and organizational commitment and on-job coupling and off-job coupling, respectively to apperceived mobility directly but no significant improvement are found to the increment of model fitting index of χ². In Fig. 4 and Table 5 for the final route relations of model variables fitted by LISREL and Goodness-of-fit is shown.

Wherein, χ² is the most common chi-square test on model’s goodness-of-fit, the ratio between χ² and responsive degree-of-freedom (df) is usually used as a rough rule of thumb for appraising the model’s goodness-of-fit. If the ratio is <2, the fitness is deemed as desirable (Joreskog and Sorbom, 1993; Hou et al., 2004). The test value of the model fitting in the study is about 1.93. RMSEA, the root mean square error of approximation is an index for model goodness-of-fit put forward by Seiger (1990) belonging to an index of absolute fitting degree. Therefore, its appraisal should combine the investigation on the confidence interval of the value.

The introducer of the index believes that in the confidence interval of 90%, the fitting degree is generally acceptable if RMSEA is below 0.10 and it is fine if the RMSEA is <0.08.

The test value of the model fitting in the study is 0.067 within the required confidence interval. GFI (Goodness of-fit Index), AGFI (Adjusted Goodness of-fit Index) are two general goodness-of-fit indexes, put forward by Joreskog and Sorbom (1993), CFI (Comparative Fit Index) and NFI (Norm Fit Index) are put forward by Bentler and Chou (1987) and Bentler and Bonnett (1980). It is generally believed that the model fits can be acceptable when they are >0.90. PNFI (parsimony normed- fit-index), PGFI (parsimony goodness-of-fit index), put forward by James et al. (1982) if their values are 0.50 or above, the model is acceptable. Therefore, based on the result of goodness-of-fit estimated by LISREL for hypothesized model of the study, the indexes of fitness are all in the reasonably acceptable.

HRA on the adjusting variables: In order to inspect the interactive effects which may exist theoretically between the adjusting variables hypothesized in the model and the major route effect variables, based on the structure hierarchies of routes and variables in the model, this study divides the interactive effects into 3 analytic hierarchies: exogenous independent variables (KDS), possible endogenous mediating variables (on-job coupling) and social linkage variables (talents social linkage are defined as the synthetic value of organization-linkage and community-linkage).

Moreover, with the front route effect variable of the corresponding hierarchy as the dependent variable, the interactive effect of adjusting variables are inspected in two steps, through covering into corresponding variable and then increasing interactive factors (Allen and Griffth, 2001). Wherein, the maximum VIF value is <4.7 (Table 6).

In accordance with result of LISREL on the basic model hypothesis for the relation of talents KDS-withdraw tendency, the 5 hypotheses (H¹-H⁵) in the expansion part for relations between the model’s mediator variables are all preliminarily verified. H⁶ and H⁷ are verified in significant mediator effect, especially, the mediator effect of on-job coupling to talents organizational performance to withdraw tendency seems to be more significant comparing with traditional job attitude variables (its maximum complete effect coefficient: -0.62, p<0.05). The hypotheses on the interactive effects of adjusting variables in the model gain acceptable significance level verification as shown in Table 7.

Limitation and future domains: This empirical study is developed on the basis of the MBA group. Thus, the application range of the analysis result should be treated carefully to expand into some special industry. The relatively high homogeneity of samples, although from different industries and representative location but with more similar feature of occupation and population.

Table 5:	Index for the goodness-of-fit of model estimated in LISREL

Table 6:	The interactive effect HRA result (N = 108)

Table 7:	The verified result of the mediators and the interactive effects of adjusting variables hypothesis

For previous variables of job coupling, there are several research on this function already, however the conceptual model to this function are relatively limited (Allen et al., 2003) and job coupling and other outcome variables may be is another topic can deepens in the future.

CONCLUSION

The introduction of coupling perspective into the measurement of talents’ job performance characteristic is not only in favor of catching characteristic performance features possessed by talents in creating organizational values but also consistent with the job coupling analytic mode, thus it is in favor of revealing the internal relations talents organizational behavior performance features and their retention factors. Following the results mentioned before, it can be basically concluded: relation between talents organizational performance and withdraw tendency of voluntary turnover is a multi-route system with duality (retention and turnover) and many incentive sectors. Secondly, in such a relational effect system, the mediator position of job coupling is extremely prominent with dominative restriction position to the withdraw tendency of talents. Thirdly, the off-job reward in connection with individual social network and family responsibility may be important retention factors that restrict withdraw tendency. Therefore, the introduction of job coupling based on the social capital is in favor of identifying and expanding key management domains for talent retention and cultivating core competitiveness based on intellectual capital of organizations.