Imam's Professorship Journey

Please Cite: Salehudin, Imam (2009) Social/Network Power: Applying Social Capital Concept to Explain the Behavioral Tendency of Individuals in Granting Favors within the Organizational Context. Proceeding of 4^th International Conference on Business and Management Research (ICBMR), Presented in 22^nd November 2009, Bali-Indonesia.

Social/Network Power: Applying Social Capital Concept to Explain the Behavioral Tendency of Individuals in Granting Favors within the Organizational Context

Imam Salehudin, SE.

Department of Management

Faculty of Economics University of Indonesia

gsimam@gmail.com

imams@ui.edu

Abstract:

The concept of Social Capital started from the domain of sociology and was transferred to broader application in other social sciences, such as economics and politics. It has also migrated from the inter-individual to the inter-societal level of society. This study returns to the original context of Social Capital by applying it to explain the behavioral tendency of individuals within the organizational context. The concept of Social/Network Power borrows the concept of Social Capital to explain how someone can access the power of other person, both formal and personal-based, by accessing its power base through social network. The independent variables used in this study are length of relationship, valence of relationship, existence of past favors, existence of potential favors, source of power and gender. This study uses the quasi-experimental method of policy capturing to determine whether social networks enables individuals to access the power base of other person, both formal and personal. This study uses 33 volunteers that were given 48 different scenarios, which yields 1583 unique cases for analysis. The result shows that all independent variable, except gender, has significant influence toward the behavioral tendency of individuals in granting favors by lending their power base, both formal and personal. However, using log linear model, the analysis shows that the effect of past favors toward the tendency to grant favors are moderated by the source of power. Owed favors have greater effect toward influencing the tendency to lend personal power base than formal power base to pay for those favors.

Keywords: Power, Social Capital, Individual Behavior

Summary:

Background

Power and politics plays major and interesting roles in organizational dynamics. Politics, as defined as the acquisition and use of power, determines the shifts of resources and influence decision that affects the entire organization, either for better or for worse. Meanwhile the scope of literature about sources of power in organizational context is severely limited. Majority of research in sources of power has depended on the classification of power by French and Raven (1959). It is necessary to explore possible sources of power in organizational context that have not been explored in the classical classification of power by French and Raven.

Research Statement:

Social Capital is a source of organizational power, in which people with social capital can elicit holders of power as defined by French and Raven to lend their source of power.

Objectives:

The objective of this research is to explore the concept of social capital in the organizational context as a source of power in order to broaden the classification of power by French and Raven.

Methodology:

The methodology used in this research is quasi experimentation, using Policy Capturing method to collect data. The primary data analysis is done with logistic regression, to analyze the effect of each variable, with log linear modeling as secondary data analysis, to analyze the moderation effect of the source of power to the favors owed.

Result:

In situation where the all of the situational variable is absent (constant), there are only 20.13% probability that a request to access a powerbase be granted. The maximum probability of the request to be granted -when all positive variables are included- is 97.97%, while the minimum probability –when only the negative variable is included- is 2.48%.People with long standing relationship have 2.53 times more probability to be granted a favor than people with relatively new relationship. People with positive relationship have 3.20 times more probability to be granted a favor than people with no positive relationship. People with negative relationship have 0.10 times lesser probability to be granted a favor than people with no negative relationship. People with past favors due have 2.22 times more probability to be granted a favor than people with no past favors. People with the potential to repay favors have 1.65 times more probability to be granted a favor than people without potentials. When the favor asked is concerning personal power, it has 4.90 times more probability to be granted than favors regarding formal power. Even though gender is found to be not significant, the result showed that males have a slight tendency to grant favors more, with 1.331 times or 6.73% more probability to grant favors than females do.

The probability to grant requests from individuals with past favors due, when the request asked is concerning the use of personal power, is 1.87 times greater than individuals without past favors. The probability to grant requests from individuals with past favors due, when the request asked is concerning the use of formal power, is only 1.58 times greater than individuals without past favors due. The probability to grant request concerning personal power, when the person who asked have past favors due, is 3.10 times greater than request concerning formal power.  The probability to grant request concerning personal power, when the person who asked did not have past favors due, is only 2.62 times greater than request concerning formal power.

A. Background and Literature Review

Power in Organization

Power and politics plays some major and interesting roles in organizational dynamics. Politics, as defined as the acquisition and use of power, determines the shifts of resources and influence decision that affects the entire organization, either for better or for worse.

Understanding the dynamics of power would benefit organizations by giving insight on how to harness it as well as how to control the players involved in the pursuit of it. Thus, quite a number of researchers have tried to develop theories that explain various dynamics of power and politics. The concept of power has always been inseparable from the behavior of individuals, especially within the organizational context. Effective leaders must understand the sources of power and the proper tactics required in using it to his benefit. It can be said that power is inseparable from leadership.

Max Weber, in his book that was translated to English in 1962, “Basic Concepts in Sociology” defined power as the opportunity within a social relationship that enables one to obtain anything he desires even if there is resistance. Meanwhile, the most commonly used definition of power in the field of political science is as the capacity to influence the behavior of other people, both with and without any resistance.

In accordance with definitions above, Stephen Robbins (2007) in his book “Organizational Behavior” gave the definition of power as the capacity possessed by someone to influence the behavior of others to act according to his desire. Robbins (2007) also described the classification of power by French and Raven (1959) that classifies power according to its source, which is formal and personal power. Formal power is power that is derived from the formal position within an organization.  The sources of formal power are the capacity to coerce by threats of punishments (coercive power), the capacity to promise rewards (reward power), and the legitimate formal authority of the structural position (legitimate power) held by individuals.

Personal power is power that is derived from the personal characteristic of an individual. The sources of personal power are the expertise (expert power) and the desirable traits that induce identification (reference power) owned by individuals. Robbins explains that those types of power (coercive, reward, legitimate, expert, and referent) all come from the dependency of a client to the resources held by a patron. Greater dependency of a client to a patron creates greater power of the patron to the client. The extent of the dependency is based on the extent of importance, scarcity and non-substitutability of the resources owned by the patron.

Social Capital

Social capital is a trendy phrase nowadays in the circle of social scientists and practitioners. This phrase is first used in the field of sociology in the individual scope of view, but then spreads to other field of science with wider scope of view. Portes (2000) explored the usage of this phrase and stated that Bourdieu (1985) is the first one who used this phrase in his paper to explain his opinion that one purpose of individuals in building relationships with other individuals is to obtain future benefits. Meanwhile, Putnam (1993) expanded the concept of individually-owned social capital as defined by Bordieu into community-owned social capital used with bigger scope of applications.

This expansion of application from the individual scope into bigger societal role often caused disambiguation among researcher. In order to avoid disambiguation, this paper limits the definition of social capital to its individual scope of application only.

Cornwell and Cornwell (2008) summarized previous researches concerning Social Capital (Burt 1992; Coleman 1988; Granovetter 1973; Lin 1999; Portes 1998) and conclude that the core of Social Capital Theory proposes that individuals can access resources owned by others through social connections or relationships with the owners.

Social capital is the social structure and relationship that enables individuals to access certain resources owned by other people. It is different than personal resources that is possessed and used by solely individuals, in which the usage of social capital incorporates interpersonal relationship and social dynamics of its user. Cornwell and Cornwell (2008) summarize at least three benefits of social capital at the individual level identified in previous researches, which is: (1) access to information, (2) social control, and (3) social support and solidarity (Coleman 1988; Sandefur and Laumann 1998).

Social/Network Power

Based on the discussions above, because the source of power is the dependency to a certain resources and resources can be accessed with certain social structure and relationships with the owner of the resources, thus it can be concluded by basic logic that someone can access the power base of other people through social relationship. The concept of Social/Network Power can be defined as the power that comes from the capacity to access the powerbase of others, aside of the power base -either formal or personal, through social relationships.

The concept of Social/Network Power emphasizes that as someone with social capital can access the resources of others through social relationship, social capital can be converted into power through the access to the power base of other people, either formal or personal. Thus, in order to observe how social/network power affects the behavior of others, we have to see it from the point of view of the owner of the powerbase. People with Social/Network Power would have better chance of success in asking a person with power to lend his or her powerbase. This is why observing the factor influencing the decision of individuals whether to grant a favor or not is relevant in measuring the effect of social/network power.

Social/network power comes from the social capital owned by individuals. Therefore, factors that build social capital would also build individual social/network power. The first factor used is the length of relationship. This factor is relevant in determining the trust upon a relationship is built. McAllister (1995) quotes the finding of Zucker (1986), Cook & Wall (1980) and Granovetter (1985) that one factor influencing inter-individual trust is the frequency and consistency of past successful interaction between individuals. This is because the personal nature of interaction that makes it possible for people to keep track record of past behaviors of each other.

The second factor is the valence of the relationship or whether the individual sees the relationship with the other individual as a positive, negative or neutral relationship. Positive valence would signal a relationship with strong trust, while negative valence would signal a relationship with strong distrust. Neutral valence would mean the relationship has neither strong trust nor distrust. Valence of the relationship would influence the social capital since people could keep track of their behavior to each other and would behave consistently based on the principle of fairness and reciprocity (Lindskold, 1978; Stack, 1988 in McAllister, 1995).

The third and fourth variable concerns the principle of reciprocity generally prevalent within eastern culture. Abdulkadiro˘glu & Bagwell (2005) discovered that individuals would only exhibit trust and facilitates cooperation if such behavior is seen as a favor that must be reciprocated, both instantly or deferred.

The fifth variable is the source of power (powerbase) that is being accessed. Powerbase can be classified into two groups, either formal or personal source of power, as explained previously. In general, the social access to formal-based power is more limited and regulated by organizational norms and rules compared to the social access to personal-based power. Thus this variable acts as moderating variable between the previous variables and the dependent variable.

B. Methodology

Research Design

This research is designed as a quasi experimental research using the policy capturing method. Kline and Sulsky (1995) elaborated that the main question in researches using policy capturing is “What decision would individuals take with the available information?”

This method is executed by exhibiting a series of scenarios to each research subject, in which each scenario is based on a combination of information cues derived from the independent variables used in the research, and then measuring their response to each scenario.  Waller and Novack (1995) describes their decision to use policy capturing by quoting several previous researches that explains that in individual decision making, there is usually not enough time to consider all the detailed information and implications regarding the decision. Meanwhile, psychological studies shows that 70% of variations in human judgement can be explained by a linear model called judgmental structure. This judgmental structure is built by observing the available information cues. Policy capturing is used to capture the judgmental structure of individuals.

Therefore, it can be concluded that the purpose of this approach is to understand the individual judgmental structure in making decisions, by observing the relationships between various information cues used with the final decision made in each scenario.

Selection of Subjects

This research involved 40 subjects participating in the data collection. However, only data from 33 subjects are used in the data analysis because 7 subjects did not pass the manipulation check.

Since the research design for this research is quasi experimental, thus there is no requirement for the sampling process to be probabilistic. All subjects recruited for this research is participating voluntarily.

Subjects are recruited using non-probabilistic sampling from a single homogenous group. The group selected as the subject for this research is college students that have experience (past or present) as officers in an organization.

Data Collection

Data collection is conducted for one week between 19th and 25th of April 2009. Data collection is conducted by giving respondents a set of questionnaire consisting of 48 different scenarios, 6 respondent identification items, and 4 manipulation check items. Displayed below is the list of variables and the dimensions used in this research.

Table B.1 Dimension of Variables Used

Variables		Coding	Dimension
Code	Name
X1	Length of Relationship	1	Old
		0	Recent
X2a	Positive Valence	1	Existing
		0	Non Existing
X2b	Negative Valence	1	Existing
		0	Non Existing
X3	Favors Owed	1	Existing
		0	Non Existing
X4	Potential Favors	1	Existing
		0	Non Existing
X5	Source  of Power	1	Personal
		0	Formal

C. Data Analysis

Descriptive Data Analysis

In order to understand the data obtained before running the inferential statistics, we use descriptive statistics to capture the relationship between variables. The descriptive statistics used is cross tabulation. The results of the cross tabulations between each independent situational variable to the participant’s final decision, are shown in Appendix B.

Logistic Regression Analysis

This research uses two different data analysis method. The first method is the logistic regression method that is used to see the general decision structure by measuring the effect of each manipulated independent variable to the response probability of individual respondent for each scenario. Here is the result of the data analysis with Logistic Regression method using SPSS 15.0

Case Processing Summary

Unweighted Cases(a)		N	Percent
Selected Cases	Included in Analysis	1583	100.0
	Missing Cases	0	.0
	Total	1583	100.0

a  If weight is in effect, see classification table for the total number of cases.

Step 0: Model only consist of constant:

Classification Table (a,b)

	Observed		Predicted
			Comply		Percentage Correct
			No	Yes
Step 0	Comply	No	0	707	.0
		Yes	0	876	100.0
	Overall Percentage				55.3

a  Constant is included in the model.

b  The cut value is .500

Variables in the Equation

		B	S.E.	Wald	df	Sig.	Exp(B)
Step 0	Constant	.214	.051	17.973	1	.000	1.239

Thus the model’s predictive capability, with just the constant, is only 55.3% with a value that significantly does not equal zero.

Step 1: Enter Method, All Independent Variables Included:

Classification Table(a)

	Observed		Predicted
			COMPLY		Percentage Correct
			No	Yes
Step 1	COMPLY	No	508	199	71.9
		Yes	151	725	82.8
	Overall Percentage				77.9

a. The cut value is .500

Overall, the proposed model predicted 77.9% of the decisions correctly. Even though this value is not the same as R²; this value can be used as some measure for the model’s prediction power. Besides that, in comparison with the “step 0” model, this model has greater prediction power.

Omnibus Tests of Model Coefficients

		Chi-square	df	Sig.
Step 1	Step	691.289	7	.000
	Block	691.289	7	.000
	Model	691.289	7	.000

Model Summary

Step	Cox & Snell R Square	Nagelkerke R Square
1	.354	.474

The R square of the proposed model is 0.354 (Cox&Snell) and 0.474 (Nagelkerke). This means that the predictors used in the model can predict between 35.4% to 47.4% of the variability of response to a request.

Variables in the Equation

		B	S.E.	Wald	df	Sig.	Exp(B)
Step 1(a)	Length of Relationship	.928	.132	49.452	1	.000	2.528
	Positive Valence	1.165	.159	53.345	1	.000	3.205
	Negative Valence	-2.294	.163	198.892	1	.000	.101
	Favors Owed	.798	.131	37.126	1	.000	2.222
	Potential Favors	.500	.130	14.879	1	.000	1.648
	Source of Power	1.590	.139	130.692	1	.000	4.903
	Gender	.271	.129	4.406	1	.036	1.311
	Constant	-1.377	.179	58.901	1	.000	.252

a  Variable(s) entered on step 1: lama, pandpos, pandneg, favorowe, potentfav, source, gender.

The analysis above shows that all independent situational variables possess significant effect in explaining the individual decision in granting a favor. Therefore, we obtain the following equation:

Ln(Y)=-1.377+0.928X₁+1.165X_2a-2.294X_2b+0.798X₃+0.5X₄+1.59X₅+0.271X₆

Sig. 0.000     0.000         0.000           0.000       0.000     0.000    0.000     0.036

We would also obtain from the information regarding the Exp(B) that in situation where the all of the situational variable is absent (constant), people have more tendency to refuse the request since there are only 20.13% probability that the request be granted. The maximum probability of a request to be granted -when all positive variables are included- is 97.97%, while the minimum probability –when only the negative variable is included- is 2.48%.

If each variable is calculated individually, people with long standing relationship have 2.528 times more probability to be granted a favor than people with relatively new relationship and the probability of the favor approved increased by 18.83% if asked by people with longstanding relationship.

Meanwhile, people with positive relationship have 3.205 times more probability to be granted a favor than people with no positive relationship. On the contrary, people with negative relationship have 0.101 times less probability to be granted a favor than people with no negative relationship.

Regarding the reciprocity of favors, people with past favors due have 2.222 times more probability to be granted a favor than people with no past favors. In addition, people with the potential to repay favors have 1.648 times more probability to be granted a favor than people without potentials.

When the favor asked is concerning personal power, it has 4.903 times more probability to be granted than favors regarding formal power. Regarding gender, even though this variable is considered not significant, the result showed that males have a slight tendency to grant favors more, with 1.331 times or 6.73% more probability to grant favors than females do.

Log linear Model Analysis

The second method of analysis is the log linear model used in order to see the interaction effect between gender and the source of power to the other situational variable used in this research. Displayed below is the result of the log linear model analysis using Microsoft Excel. The interaction effect analyzed in this case is only the interaction between Favors Owed and Source of Power in affecting the Decision to Grant Favor. The Contingency and Marginal table for the Log linear model analysis is shown in appendix C.

Fitted Value			Model
Source	Agree	Favor	(S,A,F)	(SA,F)	(AF,S)	(AF,AS)	(SAF)
Personal	Yes	Yes	219.00	268.17	243.35	297.98	295	295.5
		No	218.72	267.83	194.38	238.02	241	241.5
	No	Yes	176.75	127.58	152.40	110.01	101	101.5
		No	176.53	127.42	200.87	144.99	154	154.5
Formal	Yes	Yes	219.28	170.11	243.65	189.02	192	192.5
		No	219.00	169.89	194.62	150.98	148	148.5
	No	Yes	176.97	226.14	152.60	194.99	204	204.5
		No	176.75	225.86	201.13	257.01	248	248.5
		G2	126.4	26.5	102.1	2.2092	0
		df	4	3	2	1	0
		p-value	0.0000	0.0000	0.0000	0.1372

Based on the fitted value result in the above, only the (AF,AS) model is calculated because it is the only significant model while the other models has p-value of less than 0.05. Since the (AF,AS)  model is significant, this means that there are interaction effect between favors owed and source of power in influencing the decision for granting a favor request.

Estimated Odds Ratio	Conditional Association		Marginal Association
Model	SA	FA	SA	FA
(S,A,F)	1.00	1.00	1.00	1.00
(SA,F)	2.79	1.00	2.79
(AF,S)	1.00	1.65		1.65
(AF, AS)	2.79	1.65	2.79	1.65
(SAF) Level 1	3.10	1.87	2.79	1.65
(SAF) Level 2	2.62	1.58

The analysis will only concern the (AF,AS) model since it is the only model significant. Thus, based on the estimated odds ratio table for the (AF,AS) model above, it can be concluded that:

1. The probability to grant requests from individuals with past favors due, when the request asked is concerning the use of personal power, is 1.87 times greater than individuals without past favors.

2. The probability to grant requests from individuals with past favors due, when the request asked is concerning the use of formal power, is only 1.58 times greater than individuals without past favors due.

3. The probability to grant request concerning personal power, when the person who asked have past favors due, is 3.10 times greater than request concerning formal power.

4. The probability to grant request concerning personal power, when the person who asked did not have past favors due, is only 2.62 times greater than request concerning formal power.

5. In total, the interaction effect between personal and favors due caused request from individuals with favors due and concerning personal power have 1.18 times greater probability to be granted.

Discussion and Conclusion

Regarding the findings, there are several important implications to note. The first one is that social/network power has greater effect on personal-based power has than formal-based. The use of formal power has more strict rules and regulation as well as social and ethical norms that might limit the actions of individuals with formal power. Personal-based power is less regulated and individuals can use theirs freely with less limitation from social and ethical norms.

The second implication is that negative relationship has greater impact than positive relationship. This finding is generally consistent with the tendency of humans to give greater reaction to negative actions than positive actions.

The third implication is that the reciprocation of favors, both past and future, plays important roles in influencing the decision. Even though this interplay of favors is still moderated by the source of power concerned, it is still significant even in favor requests concerning formal power that is limited by rules and regulations.

The last implication is that gender differences have only slight influence in regarding the effect of social/network power. Of course, the gender of the requester is not supplied in the scenario, thus the effect gender interaction in this case is still open for further investigation.

From the results above, it is most likely that social/network power can significantly explain the tendency of individuals to grant favors regarding their powerbase. However, it is imperative to note the limitation of this research in which policy capturing is a quasi experimental approach. Experimental and quasi experimental approaches tend to focus on high internal validity in sacrifice of external validity. With so many external factors controlled, it is possible that there are other important factors that is not analyzed in this research. One important factors not included in this research is the effect of culture and values. This study is conducted within oriental culture environment that scored high on collectivism. A replication in low collectivism culture environment might yield different result, thus it is interesting to contrast the difference between these cultures.

List of Reference

Abdulkadiroglu, A. & Bagwell, K. (2005) Trust, reciprocity and favors in cooperative relationships. Discussion Papers 0405-22, Columbia University, Department of Economics.

Cornwell, E.Y dan Cornwell, B. (2008) Access to Expertise as a Form of

Social Capital: An Examination of Race- and Class-Based Disparities in Network Ties to Experts. Sociological Perspectives, Vol. 51, No. 4, pp. 853–876, ISSN 0731-1214, electronic ISSN 1533-8673.

Daniel J. McAllister (1995)Affect- and Cognition-Based Trust as Foundations for Interpersonal Cooperation in Organizations. The Academy of Management Journal, Vol. 38, No. 1 (Feb., 1995), pp. 24-59

Kline, T.J. & Sulsky, L.M. (1995) A policy-capturing approach to individual decision-making: a demonstration using professors’ judgments of the acceptability of psychology graduate school applicants. Canadian Journal of Behavioural Science. Ottawa: Oct 1995. Vol. 27, Iss.4; pg. 393

Labianca, G. & Brass, D.J. (2003) Exploring The Social Ledger: Negative Relationships And Negative Asymmetry In Social Networks In Organizations. Academy of Management Review.

Portes, A. (2000) The Two Meanings of Social Capital. Sociological Forum, Vol. 15, No. 1 (Mar., 2000), pp. 1-12. Springer, Diakses melalui: http://www.jstor.org/stable/3070334 pada: 15 Februari 2009

Robbins, S.P. dan Judge, T.A. (2007). Organizational Behavior. Upper Saddle River, N.J: Pearson Prentice Hall.

Weber, M. (1962) Basic Concepts in Sociology. Translated by H. P. Secher. The Citadel Press. Diakses dalam bentuk html melalui: http://www.ne.jp/asahi/moriyuki/abukuma/weber/method/basic/basic_concept_frame.html

Salehudin, I. (2009) The 6^th Power: Social Network Power, Pengembangan konsep Social Capital pada konteks Individu. Manajemen Usahawan Indonesia. No. 03/TH. XXXVIII 2009. ISSN: 0302-9859

Waller, M.A. & Novack, R.A. (1995) Using Policy Capturing To Identify The Effects Of External Consistency On

Logistics Managers’ Performance. Transportation Journal. Lock Haven: Spring 1995. Vol. 34, Iss. 3; pg. 45.

Appendix A: Sample Question

n
If someone you are acquainted (from way back /recently), with (positive/no particular/negative) view of the person, while you (owed/did not owed) favors and he (have/does not have) potential in repaying favors in the future, asks you of something outside your job description that uses your (personal capacity/organizational authority), will you grant his request?
Yes	No

Appendix B: Descriptive Cross tabulation

Table C.1 Cross tabulation between Length of Relationship and Decision to Comply

		Comply		Total Response
		NO	YES
Length of Relationship	Recently	410	381	791
	Old	297	495	792	Chi Square	Sig.
Total		707	876	1583	32.8958	0.0000

Table C.2 Cross tabulation between Positive Valence and Decision to Comply

		Comply		Total Response
		NO	YES
Positive Valence	NO	612	443	1055
	YES	95	433	528	Chi Square	Sig.
Total		707	876	1583	227.9996	0.0000

Table C.3 Cross tabulation between Negative Valence and Decision to Comply

		Comply		Total Response
		NO	YES
Negative Valence	NO	289	766	1055
	YES	418	110	528	Chi Square	Sig.
Total		707	876	1583	381.6414	0.0000

Table C.4 Cross tabulation between Favors Owed and Decision to Comply

		Comply		Total Response
		NO	YES
Favors Owed	NO	402	389	791
	YES	305	487	792	Chi Square	Sig.
Total		707	876	1583	24.2712	0.0000

Table C.5 Cross tabulation between Potential Favors and Decision to Comply

		Comply		Total Response
		NO	YES
Potential Favors	NO	384	408	792
	YES	323	468	791	Chi Square	Sig.
Total		707	876	1583	9.3720	0.0022

Table C.6 Cross tabulation between Source of Power and Decision to Comply

		Comply		Total Response
		NO	YES
Source  of Power	Formal	452	340	792
	Personal	255	536	791	Chi Square	Sig.
Total		707	876	1583	98.7458	0.0000

Table C.7 Cross tabulation between Gender and Decision to Comply

		Comply		Total Response
		NO	YES
Gender	Female	359	408	767
	Male	348	468	816	Chi Square	Sig.
Total		707	876	1583	2.7666	0.0962

Appendix C: Log linear Model Analysis

Table C.8 Contingency Table for 3 Variables (Source-Agree-Favor)

		Favor
Source	Agree	Yes	No	Total
Personal	Yes	295	241	536
	No	101	154	255
Formal	Yes	192	148	340
	No	204	248	452
	Total	792	791	1583

Table C.9 Marginal Table (Source-Favor)

	Favor
Source	Yes	No	Total
Personal	396	395	791
Formal	396	396	792
Total	792	791	1583

Table C.10 Marginal Table (Agree-Favor)

	Favor
Agree	Yes	No	Total
Yes	487	389	876
No	305	402	707
Total	792	791	1583

Table C.11 Marginal Table (Agree-Source)

	Source
Agree	Personal	Formal	Total
Yes	536	340	876
No	255	452	707
Total	791	792	1583

Please Cite: Salehudin, I. and Mukhlish, B.M. (2008) Application of Planned Behavior Framework in Understanding Factors Influencing Intention to Leave among Alumnae of the Faculty of Economics University of Indonesia Year 2000-2003. Proceeding of 3^rd International Conference on Business and Management Research (ICBMR), Bali-Indonesia.

Application of Planned Behavior Framework in Understanding Factors Influencing Intention to Leave among Alumnae of the Faculty of Economics University of Indonesia Year 2000-2003

Imam Salehudin, SE

University of Indonesia

gsimam@gmail.com

Basuki Muhammad Mukhlish, SE

University of Indonesia

basukimukhlish@gmail.com

Abstract

Employee’s turnover or job-switching behavior has always been a major concern of every human resources manager in relation to human capital investments. Many researches have been done to analyze how and why people switched jobs. Although most of these researches use attitudinal approaches that linked attitude directly to behavior, Chandrashekaran et al. (2000) and Mitchell-Sablynski et al. (2001) has used intention as intervening variable in their job switching model.

This research uses Theory of Planned Behavior from Ajzen (2004) as framework to model intentions to leave among fresh graduates. Theory of Planned Behavior uses Attitude, Subjective Norms and Perceived Behavioral Control to predict intention which leads to behavior. Psychological contract is used as one attitudinal variable based on research done by Chay and Aryee (1994), while job-embeddedness is also used as the second attitudinal variable based on research done by Mitchell-Sablynski et al. (2001). Ease of Movement is used as behavioral control variable, both actual and perceived, based on separate researches done by Spencer and Steers (1980), Trevor (2001), and Malcolmson et. al (2005). While subjective norms is developed using the questions used by Ajzen (2004).

The respondent used in this research is 129 graduates randomly sampled from 1105 alumnus of the Faculty of Economics University of Indonesia year 2000-2003 using clusters sampling method. Structural Equation Modeling process is then used to test whether the data obtained from the survey supports the model proposed.

The result obtained from this research is that there is a significant negative relationship between both attitude construct and the intention to leave, in which higher Job Embeddedness and Psychological Contract would decrease the intention to leave. However, this research found that the relationship between Subjective Norms and Perceived Movement Capital to the Intention to Leave is not significant based on the data used in this research. Further research is recommended to confirm about the relationship between the Intention to Leave and the actual behavior of Job Switching, and to analyze the influence of Actual Movement Capital toward Perceived Movement Capital and the actual Job Switching Behavior.

Keywords: Planned Behavior, Intent to Leave, Fresh Graduate, Job-Embeddedness, Psychological Contract, Ease of Movement.

Backgrounds

Normally, employers would not want to see their employee quits their job, especially if the one quitting is a talented new prospect and she plans to move to the competing company. Most company sees employee turnover as a loss since most do spend a lot of money in order to attract, develop, maintain and retain their employees. Some would see this spending as a loss of investment since they thought of their expenditure as investments in their human capital. Like any other investments, companies would expect to get a corresponding return from that investment. Employee turnover would mean that the investments made for that employee stops generating returns. It could also prove crippling for the company, especially it the employee leaving is a key person and she switched to the company’s competitor.

Thus, minimizing employee turnover has always been one of the key performance indicators for human resources managers in most company. To try minimizing employee turnover, managers must understand it first. Although there has been plenty of researches done to understand the concept of employee turnover, there is still much to be learned and understood about why and how employees decide to quit their jobs.

Literature Review and Hypothesis

Various researches have been done to understand more about employee turnover and how it happens. Each of these researches have contributed toward understanding more about why and how employees decided to quit their jobs, so it is important to use this accumulated knowledge as the basis of this research.

Researches using attitudinal variables most often viewed the relationship between attitude and behavior as a direct relationship, in which attitude would correspond directly toward behavior. Most of these researches also used Job Satisfaction as their main attitudinal variable, such as Blumberg (1980), Hom and Kinicki (2001), and Trevor (2001). Although the result was most often mixed, Hom and Kinicki (2001) found that the attitudinal variables they used do not have direct relationship toward the actual Job Switching behavior.

Some researches have used intention as intervening variable in their job switching model. Chandrashekaran et al. (2000) found that salesman with higher intention to leave will quit their jobs faster, while Sablynski et al. (2001) found that attitudinal variables such as Job Satisfaction and Job Embeddedness would influences employee’s intention to leave, which in turn would influence the actual voluntary turnover. Thus, it can be concluded that intention to leave can be used as an intervening variable between attitude and behavior.

This research uses Theory of Planned Behavior from Ajzen (2004) as framework to model intentions to leave among fresh graduates. Theory of Planned Behavior has been used frequently to model behaviors in several fields of social science, such as marketing and psychology. The theory uses Attitude, Subjective Norms and Perceived Behavioral Control to predict intention which will leads to behavior.

In this research, there are two attitudinal variables used. Psychological contract is used as one attitudinal variable based on research done by Chay and Aryee (1994), while job-embeddedness is also used as the second attitudinal variable based on research done by Mitchell-Sablynski et al. (2001). Subjective Norms is developed from the concept used in the original model by Ajzen (2002). He proposed that Subjective Norms refers to an individual’s perceptions of other people’s opinions on whether or not he or she should perform a particular behavior, while perceived behavioral control refers to an individual’s perceptions of the presence or absence of the requisite resources or opportunities necessary for performing a behavior. Perceived Movement Capital is used as perceived behavioral control variable, based on separate researches done by Spencer and Steers (1980), Trevor (2001), and Malcolmson et. al (2005). Therefore, the following hypotheses are suggested based on previous researches above:

H1: Job Embeddedness has negative effect on Intention to Leave, thus alumnae with higher Job Embeddedness will experience less Intention to Leave.

Job Embeddedness can be broken down into three components, which is Fit, Links, and Sacrifice. Job Fit is the measurement on how strong employee feels that he belong to his job and organization. Job Links is the measurement on how much is the connection and interdependence between an employee and his work environments. Job Sacrifice is how big the sacrifice for the employee to leave his job and/or organization is.

The concept of Job Embeddedness is an alternative approach to the commonly used Job Satisfaction. Sablynski et. al. (2001) even postulates further that Job Embeddedness is better predictor of employee turnover, absenteeism, and job performance than Job Satisfaction. Strong Job Embeddedness would reduce employee’s intention to leave the same way strong Job Satisfaction would. However, weak Job Embeddedness would not encourage employees to quit like strong Job Dissatisfaction would. Weak Job Embeddedness would only means that the employee will be more susceptible to shocks and dissatisfactions.

H2: Relational Psychological Contract has negative effect on Intention to Leave, thus alumnae with Relational Psychological Contract will experience less Intention to Leave than alumnae with Transactional Psychological Contract.

Rousseau (1989) defined Psychological Contract as the employee’s perception of the reciprocal obligations existing with their employer; as such, the employee has beliefs regarding the organization’s obligations to them as well as their own obligations to the organization. Two dominant Psychological Contracts identified by past researches are relational and transactional contracts. MacNeil (1985) stated that a relational contract characterizes traditional employment relationship in which employees expects long-term relationships, experiences both monetizable and socio-emotional elements, has broad scope and is based on the values of good faith and fair dealing. In contrast, Rousseau and Parks (1993) stated that a transactional contract is characterized by short-term, purely monetizable agreements with limited involvement of each party in the lives and activities of the other, and has a narrow focus but a high degree of specificity.

Employees with transactional Psychological Contract would feel no compunction for quitting their jobs and pursue their career elsewhere, if a better career opportunity is present elsewhere. In the other hand, Chay and Aryee (1994) postulates that employees with relational Psychological Contract would not seek such opportunity and would even hesitate should any such opportunity is suddenly presented to them. Thus, employees with relational Psychological Contract should experience less intention to leave than employees with transactional Psychological Contract.

H3: Subjective Norms has negative effect on Intention to Leave, thus alumnae with stronger Subjective Norms would experience less Intention to Leave than alumnae with weaker Subjective Norms.

Ajzen (2002) stated that subjective norms refer to an individual’s perceptions of other people’s opinions on whether or not he or she should perform a particular behavior. Subjective Norms consists of Normative Believe and Motivation to Comply, and calculated my multiplying the two. Normative Believe is the perception of alumnus of whether their friends and family would disapprove of their action if they choose to quit, while Motivation to Comply is the extent to which the alumnus will act in regard of their disapprovals. Alumnae with higher Subjective Norms would then experience less Intention to Leave, since they would see that their friends and family would disapprove and would not act regardless of their disapproval.

H4: Perceived Movement Capital has positive effect on Intention to Leave, thus alumnae that perceive that they have better Movement Capital would experience more Intention to Leave than alumnae that perceive that they have less.

The concept of Ease of Movement has been found to be a contributing element to employee turnover. It consists of both individual and market-driven determinants. Movement Capital is the individual element of Ease of Movement. The underlying concept is that employees with better credentials will be more attractive in the general job market, thus have more opportunity to leave. Thus, dissatisfaction levels will matter more for voluntary turnover when people are better able to secure alternative employment by signaling the market that they are worth hiring. When they have little with which to signal competence, dissatisfaction will be less likely to matter since viable alternatives will be less likely to exist (Trevor, 2001). This research uses perceived Movement Capital rather than actual Movement Capital to predict alumnae’s Intention to Leave. Each alumnus would perceive their Movement Capital differently and is more likely to act on what they perceive and believe than what the actual reality is. Actual Movement Capital, however, does influence Perceived Movement Capital and the actual turnover behavior.

Five indicators of movement capital are used in this research. These indicators are obtained through exploratory Focus Group Discussion held at the beginning of this research. First is the Alma Mater or the origin of the Alumnae, which in this case is the Faculty of Economics University of Indonesia. This indicator is an indigenous indicator as education institution in Indonesia is divided into several tiers, in which graduates from top tier institutions should get better reception than graduates from the low tier institutions. Faculty of Economics University of Indonesia is considered a top tier educational institution. The second indicator is GPA. This is also an indigenous indicator as companies in Indonesia uses GPA as minimum requirements and screening criterion for fresh graduate employee recruitment. The third indicator is level of education, which is a classic movement capital indicator as employers would require certain level of education for a certain level of jobs and infer cognitive ability from it, thus alumnus with higher education would perceive themselves as more desirable to the employers than alumnus with lower education. The fourth indicator is work related skill certifications which related to perceived Movement Capital much the same as level of education. The last indicator is Job Experiences, which is the variety and extent of experiences in working in different positions. Employers require a certain amount of experience for certain levels of jobs and experience does improve job performance, thus alumnus with more experience would perceive themselves as more desirable to the employers than alumnus with little job experience. Overall, employees high in these elements of movement capital, should, by virtue of perceived opportunity, be more ready to act upon their dissatisfaction by leaving. Comprehensive model of the hypotheses above can be seen in the figure below.

Figure 1: Hypothesized Model of Intention to Leave

Methodology

Population and Sample

The population used for this research is the alumnae of the Faculty of Economics University of Indonesia year 2000-2003 who is currently working in an organization. Based on previous research by Veiga (1983), it is decided that the population would be limited to fresh graduates to contain the difference of the career dynamics inherent within a specific career stage. This would simplify the sampling process and reduce biases from unidentified variables.

The respondent used in this research is sampled using clustered random sampling. To perform the sampling, the researchers first obtained a list of alumnae from the Faculty of Economics University of Indonesia. Then, the alumnae are categorized by Year and Department. The respondent is the randomly chosen from each category in proportion to the size of each category. 430 alumnae are randomly chosen out of 1105 alumnae listed.

After choosing the respondent, the surveyor would then call each respondent to ask whether they are currently working and willing to answer the questionnaire via email. Out of 430 alumnae sampled, only 129 responded. Thus the respond rate for the survey is 30%. The respond rate is somewhat lower than our expectation because some of the contact information seems to be outdated, thus reducing the effective number of respondent.

Measures

There are a total of 40 indicators used in the questionnaire. Job Embeddedness is measured using 17 items in three different constructs, which is Job Fit (6 items), Job Links (4 items) and Job Sacrifice (7 items). Psychological Construct is measured using 3 items. Subjective Norms is calculated into a single measure using two items. Perceived Movement Capital is measured using 15 items divided into five different construct, which is Alma Mater, GPA, Level of Education, Certification of Skills, and Job Experience. The items are composed using 5 point Likert scale.

The reliability measurement obtained using Cronbach’s Alpha for the items are 0.805, which is good enough to be used for further analusis. The complete breakdown of the variables used in this research can be seen below in table 1.

No.	Variables	Symbol	Constructs	Symbol	Indicators
1.	Job Embeddedness	ξ1	Job Fit	X1	6 items
			Job Links	X2	4 items
			Job Sacrifice	X3	7 items
2.	Psychological Contract	ξ 2	Psychological Contract	X4-X6	3 items
3.	Subjective Norms	ξ 3	Subjective Norms	X7	2 items
4.	Perceived Movement Capital	ξ4	Alma Mater	X8	3 items
			GPA	X9	3 items
			Level of Education	X10	3 items
			Certification of Skills	X11	3 items
			Job Experiences	X12	3 items
5.	Intention to Leave	η1	Intention to Leave	Y1-Y3	3 items

Table 1: Variables and Constructs

Analyses

The method of analysis used in this research is Structural Equation Modeling with Lisrel 8.51 Full Version. The measurement model must be estimated first before the structural model can estimated using the data obtained in the survey. The mathematical equation below is derived from the constructs above to be used in the estimation of the measurement model in this research.

Equation 1: Measurement Model Equation

Figure 2 & 3: Standardized Loading Factor and t-value for the Measurement Model

The purpose of the measurement model is to identify invalid items and unreliable construct. As the rule of thumb, items with t-value less than 1.96 or standardized loading factor of less than 0.3 is considered invalid. Unreliable construct is identified by calculating construct reliability (CR) and variance extracted (VE). CR is calculated by dividing the square of total loading factors with the sum of it and the total error in that construct. VE is calculated by dividing the sum of squared loading factor with the sum of it and the total error in that construct. If CR is less than 0.7 or VE is less than 0.5 than the construct is considered unreliable and must be dropped.

Thus, from the measurement model, it is found that 2 item must be dropped because their Standardized Loading Factor is less than 0.5. Item PSYC03 is still retained because its loading factor is more than 0.3 and it is based on strong results in previous researches. Complete detail on the analysis can be seen in the table below,

No.	Variables	T-Value	SLF	Error	CR	VE	Analysis	Treatment
1	JFIT	N/A*	0.83	0.3	0.84	0.72	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
2	JLINK	-3.19	-0.29	0.91	N/A	N/A	SLF<0.5	Not Valid, drop item
3	JSAC	12.34	0.86	0.26	0.84	0.72	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
4	PSYC01	N/A*	0.96	0.09	0.83	0.73	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
5	PSYC02	2.07	0.19	0.96	N/A	N/A	SLF<0.5	Not Valid, drop item
6	PSYC03	4.69	0.45	0.32	0.83	0.73	SLF>0.3, (Igbaria et. al, 1997)	Can be Accepted, Keep item
7	SNORM	N/A*	1.15	-0.31	1.31	0.57	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
8	PMC1	N/A*	0.98	0.05	0.87	0.59	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
9	PMC2	10.96	0.73	0.47	0.87	0.59	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
10	PMC3	16.27	0.87	0.24	0.87	0.59	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
11	PMC4	6.62	0.52	0.73	0.87	0.59	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
12	PMC5	9.29	0.66	0.56	0.87	0.59	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
13	INTLE01	N/A*	0.82	0.32	0.89	0.72	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
14	INTLE02	13.5	0.86	0.27	0.89	0.72	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item
15	INTLE03	13.77	0.87	0.25	0.89	0.72	SLF>0.5, T-Value>1.96, CR>0.7, VE>0.5	Keep item

Table 2: Analysis on the result of the Measurement Model

The mathematical equation below estimates the relationships tested in the structural model based on the relationship mentioned in the research Hypotheses

Equation 2: Structural Model Equation

After estimating the measurement model and eliminating invalid items, the structural model can then be estimated. The estimation is done using the Maximum Likelihood method. Several error covariances are added based on the modification index, but is limited between items within a single construct. Suggestions of error covariance between items in different construct are disregarded.   The result obtained can be seen in figure 4 and 5.

Figure 4: T-Values for the Structural Model

Figure 5: Standardized Loading Factor for the Structural Model

Results

Goodness of Fit

Before doing the hypotheses testing, it is imperative to see whether the model used in this research is acceptable. There is several goodness of fit measurement that was established to measure whether the model tested is adequate for further analyses. It was also established that there is no single absolute measurement and satisfying the entire range of goodness of fit measurements would be difficult, thus satisfying the majority of the measurements would be enough to verify the acceptability of the model.

The result of goodness of fit measurements shows that the majority of measurement shows good model fitness while the rest shows only marginal model fitness. Thus, it can be concluded from the goodness of fit measurements above that the model used in this research is acceptable and further analysis can be done to test the hypotheses of this research on the results obtained from the model.

No.	Measurement	Result	Standard	Fitness
1	Chi-Square= df= N=	62.96 60 129	Chi-Square/df=<2	Good Fit
2	P-value	0.37201	>0.05	Good Fit
3	RMSEA	0.020	<0.05	Close Fit
4	P-Value for Test of Close Fit	0.88	>0.05	Good Fit
5	ECVI	S=1.42 M=0.98 I=6.76	M is small and closer to S	Good Fit
6	AIC	S=182.00 M=124.96 I=865.22	M is small and closer to S	Good Fit
7	CAIC	S=533.24 M=244.61 I=915.39	M is small and closer to S	Good Fit
8	NNI NNFI CFI IFI RFI	0.91 0.97 0.98 0.98 0.88	>0.90 >0.90 >0.90 >0.90 >0.90	Good Fit Good Fit Good Fit Good Fit Marginal Fit
9	CN	152.10	>200	Marginal Fit
10	RMR	0.077	=<0.05	Marginal Fit
11	GFI	0.93	>0.9	Good Fit
12	AGFI	0.89	>0.9	Marginal Fit

Table 3: Measurements of Model Goodness of Fit

Hypotheses Test

The hypothesis test is done by analyzing the t-value and the sign of the loading factor of each path. Using the values obtained from the structural model, it can be concluded that Hypothesis 1 and 2 can be accepted because both path has t-value of more than 1.96. Both also have negative sign that confirms the hypothesis that both attitudes have negative effect toward Intention to Leave. Hypothesis 3 and 4 cannot be accepted because the t-value of their path is less than 1.96. In this case, the data does not support the hypothesis. The detailed hypothesis test can be seen below in table 4.

Hypotheses	Path	T-Value	SLF	Analyses
H1	JEMB–>INTLE	-5.22	-0.45	Significant, Negative SLF. Data supports the Hypothesis
H2	PSYCO–>INTLE	-4.27	-0.30	Significant, Negative SLF.Data supports the Hypothesis
H3	SNORM–>INTLE	1.20	0.08	Not Significant, Data does not support the Hypothesis
H4	PMC–>INTLE	0.97	0.02	Not Significant, Data does not support the Hypothesis

Table 4: Hypotheses Test

Discussion

Job Embeddedness and Psychological Contract

Statistical results show that there is significant negative relationship between the two attitude variables and the Intention to Leave. These findings support the first two research hypothesis. The first finding is that Job Embeddedness, which is how an employee feels attached about his job, does reduces employee’s intention to leave. This finding is consistent with an earlier research finding by Sablynski et. al. (2001), that employee with higher feeling of attachment to their job would exhibit less intention to leave than employee with less job embeddedness.

However, only two out of three construct is usable in this research since one construct, Job Links, is found not valid. The two remaining construct, Job Fit and Job Sacrifice, proved valid and reliable.  Job Links is found not valid because it has a standardized loading factor of less than 0.5, which is unacceptable. One plausible explanation for this is because that the population sampled in this research is fresh graduates, which graduated from college in less than five years ago and started working in their current company invariably less. Within this short span of employment, the respondent most probably does not have enough time to build sufficient Job Links to influence their Intention to Leave. In the other hand, Job Fit and Job Sacrifices does not require significant time to develop and influence employee intentions to leave.

The second finding is that tendency toward Relational Psychological Contract would have negative influence on an employee’s intention to leave. This finding is consistent with earlier finding of previous research by Chay and Aryee (1994). This can be explained that employee with a tendency toward Relational Psychological Contract would think of their employment relationship with their employes as a long term relationship. In viewing this, employees would not be actively seeking new jobs opportunities and might hesitate if offered one. All in all, the findings above confirms our first intuition that employee attitudes does plays a prominent role in determining whether an employee plans to leave his/her job or not.

Subjective Norms and Perceived Movement Capital

Statistical results show that the data used in this research does not supports the other two hypotheses. The first unsupported hypothesis is the relationship between Subjective Norms and Intention to Leave. It is found that the relationship between Subjective Norms and Employee Intention to Leave is statistically not significant and considered inconclusive. This finding contradicts the Planned Behavior Theory which suggested that Subjective Norms influences intention. This finding might be caused by the differential of compliance toward norms between one employee and the other. If this differential is indeed significant, further adaptation of the current questionnaire might be required to incorporate this differential within the questionnaire.

The second unsupported hypothesis is the relationship between Perceived Movement Capital and Intention to Leave. Although it is possible that higher perceived movement capital simply does not influences intention to leave, this may as well be cause by divergence of relationships between each of its constructs and employee’s Intention to Leave. Although there is no significant relationship between it and Intention to Leave, there are still possibilities that some of its component might have some influence toward Intention to Leave. Perceived Movement Capital consists of five different construct. Further breakdown of the relationship between Perceived Movement Capital and Intention to Leave is required to determine if any construct did have some relationship with Intention to Leave.

Future Research Directions and Limitation

This research has several limitations that can be explored in further research. The first is that this research limited its sample only to include fresh graduate as respondent. While this decision surely would exclude several biases and unnecessary variables, it would also reduce the validity of several job-related construct that requires sufficient time to develop. Some examples of this construct would be Job Links and Experience. Further research should be done, that includes respondents from other career stages. Hopefully, further research could identify and compare the differences of career dynamics between employees from different career stages.

The second limitation is that this research also limits the sample to the graduates of a specific faculty in a specific university. Further research should be done to include graduates from other University as well as other faculties. In this way, there would be a control groups to provide comparison of the statistical results with.

This research topic still has much room for expansion. There are still many aspects and angles of employee turnover that have not been researched. The future directions for this research are to include several variables such as actual movement capital and the actual turnover behavior. In this way, researcher can confirm whether employee intentions do indeed leads to employee behavior.

Conclusion and Suggestions

The conclusion of this research is that Employee Attitude, in this case Job Embeddedness and Psychological Contract, is confirmed as the dominant factor that affects Employee’s Intention to Leave. In this way, companies trying to reduce employee turnover can start by influencing their employee’s attitude. Employers can give better orientation and socialization of company’s culture and values to facilitate adaptation and increase Job Fit. Employers can also increase employee benefit or giving them more freedom to do their work to increase Job Sacrifice. Employers should also screens candidates in their employee selections to filter out candidates with transactional psychological contract and establish relational psychological contract with their employees through consistent human research policies and intensive communication with their employees.

As for educational institutions, this research confirms that attitude is also an important output of their service. Although further research would be requires, but it is safe to say that to some extent how fresh graduates develop their workplace attitude is largely influenced by their previous attitude in college. Thus, attitude should be stated explicitly in the curriculum as one of the expected output.

Bibliography

1)      Ajzen, Icek. Constructing a TPB Questionnaire: Conceptual and Methodological Considerations (2002).

2)      Ajzen, Icek. Behavioral Interventions Based on the Theory of Planned Behavior (2004).

3)      Chandrashekaran, Murali; McNeilly, Murali; Russ, Frederick A. and Marinova, Detelina. From Uncertain Intentions to Actual Behavior: A Threshold Model of Whether and When Salespeople Quit. Journal of Marketing Research, Vol. 37, No. 4. (Nov., 2000), pp. 463-479.

4)      Charlie O. Trevor. Interactions among Actual Ease-of-Movement Determinants and Job Satisfaction in the Prediction of Voluntary Turnover. The Academy of Management Journal, Vol. 44, No. 4. (Aug., 2001), pp. 621-638.

5)      Chay, Yue-Wah and Aryee, Samuel. Potential Moderating Influence of Career Growth Opportunities on Careerist Orientation and Work Attitudes: Evidence of the Protean Career Era in Singapore. Journal of Organizational Behavior, Vol. 20, No. 5. (Sep., 1999), pp. 613-623.

6)      MacNeil, I. R.. ‘Relational contract: what we do and do not know’, Wisconsin Law Review, 483-525. (1985)

7)      Malcolmson, Jo; Brown, Paula and Diamond, Holly. Key drivers of retention in the Royal Navy and Royal Marines (2005).

8)      Mitchell, Terence R.; Holtom, Brooks C.; Lee, Thomas W.; Sablynski, Chris J. and Erez, Miriam. Why People Stay: Using Job Embeddedness to Predict Voluntary Turnover. The Academy of Management Journal, Vol. 44, No. 6. (Dec., 2001), pp. 1102-1121.

9)      Rousseau, D. M. Psychological and implied contracts in organizations’, Employee Rights and Responsibilities Journal, 2, 121-139. (1989)

10)  Rousseau, D. M. and Parks, J. M. ‘The contracts of individuals and organizations’, In: Cummings, L. L. and Staw, B. M. (Eds.), Research in Organizational Behavior, Vol. 15, pp. 1-43, JAI Press, Greenwich, CT. (1993).

11)  Spencer, Daniel G. and Steers, Richard M. The Influence of Personal Factors and Perceived Work Experiences on Employee Turnover and Absenteeism. The Academy of Management Journal, Vol. 23, No. 3. (Sep., 1980), pp. 567-572.

12)  Veiga, John F. Mobility Influences during Managerial Career Stages. The Academy of Management Journal, Vol. 26, No. 1. (Mar., 1983), pp. 64-85.