Perceived obsolescence, organizational embeddedness, and turnover of it workers

Perceived Obsolescence, Organizational Embeddedness, and Turnover of IT Workers: An Empirical Study

Xiaoni Zhang Northern Kentucky University Sherry D. Ryan University of North Texas Victor R. Prybutok University of North Texas Leon Kappelman University of North Texas

Abstract

Technical skill currency is especially important to IT professionals, yet past research has not integrated perceived obsolescence into the nomological net with organizational embeddedness when investigating IT workers' turnover intentions. Responding to a call for additional research on IT worker turnover we utilize a sample of 1,777 IT workers in a very large and complex IT organization and find that organizational embeddedness explains significant incremental variance beyond the traditional turnover model. The findings presented here show that organizational embeddedness is an essential mediator between perceived skill obsolescence and IT voluntary turnover. Our data also show that the effect of perceived obsolescence on embeddedness is moderated by age. This work provides organizations with insights on why employees choose to stay in an organization, or leave it. We find that both fit and sacrifice dimensions of embeddedness mediate the relationship between perceived skill obsolescence and turnover intention, but that only the sacrifice dimension is a full mediator. These findings suggest that turnover intention is a multivariate issue and that future IT turnover research needs to include other variables such as perceived obsolescence and age. Furthermore there is a need for future research on the role of the specific dimensions of embeddedness. We suggest strategies to reduce voluntary turnover based on the predictive variables in the research model.

ACM Classifications: K.6.1 Project and People Management; K.7 The Computing Profession

Keywords: Information Technology Professionals; Voluntary Turnover; Turnover Intention; Organizational Embeddedness; Job Embeddedness; Job Satisfaction; Retention; Perceived Skill Obsolescence; IT Workforce.

Introduction In the last decade the demand for IT professionals has expanded and contracted. Nevertheless, high-technology companies and IT departments frequently suffer from high employee turnover issues and quite often find it hard to hire the talent that they need (Chien and Chen, 2008). The U.S. Bureau of Labor Statistics reports that from 2006-2018, technical consulting services and computer systems design and related services will be among the fastest growing occupations in wage and salary employment (Figueroa and Woods, 2007). Much research on turnover intention has been done in the IT and management literatures (for a meta analyses see Cotton and Tuttle, 1986; Griffeth et al.,

The DATA BASE for Advances in Information Systems 12 Volume 43, Number 4, November 2012

https://www.researchgate.net/publication/223190807_Data_mining_to_improve_personnel_selection_and_enhance_human_capital_A_case_study_in_high-technology_industry?el=1_x_8&enrichId=rgreq-7746708b-7f33-4148-a770-ae20c40c4d35&enrichSource=Y292ZXJQYWdlOzI2MjMxNTQ4NjtBUzoxMjAzMTgyMTEzOTk2ODFAMTQwNTY5NzUwNTUyMg==

2000; Joseph et al., 2007). Nevertheless, the variance accounted for by these models is relatively low, comprising about 15-34% of the variance in IT worker turnover intentions (e.g., Igbaria and Guimaraes, 1999; Thatcher et al., 2006). In 2001, Mitchell et al. introduced a new construct entitled “Job Embeddedness” that explained significant incremental variance over and above traditional turnover models using samples of grocery and hospital workers. According to Mitchell et al. (2001), the essential aspects of job embeddedness are “(1) the extent to which people have links to other people or activities, (2) the extent to which their jobs and communities are similar to or fit with the other aspects in their life spaces, and, (3) the ease with which links can be broken – what they would give up if they left, especially if they had to physically move to other cities or homes” (Mitchell et al., 2001, p, 1104). Joseph et al. (2007) called for more research on embeddedness and suggested that it is potentially important to IT professionals because this may be a significant factor in determining whether IT workers stay with their organization even though they are dissatisfied with their jobs. We argue that a factor that may influence job embeddedness, as well as directly effecting turnover in IT workers, is the perception of professional obsolescence. Joseph and Ang (2001) have identified the threat of rapid professional obsolescence as an important career challenge that faces IT professionals. IT professionals’ core competencies erode quickly, because unlike other professionals whose basic knowledge may be fairly stable, technological change can cause dynamic shifts. In fact, it is estimated that fifty percent of an IT professional’s knowledge and skills may be out-of-date within two years (Ang and Slaughter, 2000; Joseph and Ang, 2001). “Although most employees in this fast-changing business world need to keep their skills up-to-date, nowhere is this requirement as essential as among information technology workers,” (Standridge and Autrey, 2001, p. 3). While professional obsolescence is especially salient to IT professionals, here-to-date, it has not been incorporated in the nomological net with job embeddedness when investigating IT workers’ turnover intentions. Our paper significantly contributes to the IT discipline by extending the embeddedness theory to account for contextual influences – perceived skill obsolescence. Specifically, we theorize and test the relationships between perceived skill obsolescence, embeddedness, and turnover intention. Second, we theorize and empirically test how age interacts with other variables (specifically between perceived

obsolescence moderated by age and embeddedness and between perceived obsolescence moderated by age and turnover intention). Prior IT turnover studies have, in general, examined age as a control variable but do not theorize on its relationships with other variables related to turnover intention (Gallivan 2004; Moore 2000). We do so in this paper. And third, we compare the predictive validity of three models in predicting turnover: Model A with job alternative and job satisfaction as predictors; Model B with job alternative, job satisfaction, and embeddedness as predictors; Model C with job alternative, job satisfaction, embeddedness, and perceived skill obsolescence moderated by age as predictors.

Theoretical foundations and hypothesis development

The theoretical foundation for our research is built upon the job embeddedness and professional obsolescence literatures. Figure 1 shows our research model. In addition to embeddedness, perceived obsolescence, and age, we include job satisfaction and job alternatives as control variables from March and Simon’s (1958) traditional theory of voluntary turnover and test the efficacy of embeddedness in explaining additional variance of turnover intention beyond these traditional variables. In this section we discuss embeddedness and its subdimensions, embeddedness’ relationship to turnover intention, perceived obsolescence and its relationship to turnover intention, and perceived obsolescence and its relationship to embeddedness.

Embeddedness

Employee retention issues are not simplistic. While traditional turnover models (e.g., March and Simon, 1958) address why some people choose to quit an organization, they do not specifically address why some employees are reluctant to leave (Mitchell et al., 2001). Drawing upon research in the psychological attachment literature, Mitchell et al. (2001) proposed a new construct called “Job Embeddedness” that explains why employees remain in an organization. Applying these concepts within an organizational context, Mitchell et al. (2001) conceptualized these connections as forming a web. The greater the number of connections, the more dense the web becomes. The thicker the web, the more an employee becomes “embedded” or stuck in the web, and the harder it becomes for the individual to detach or leave.


https://www.researchgate.net/publication/211388636_Turnover_of_information_technology_professionals_A_narrative_review_meta-analytic_structural_equation_modeling_and_model_development?el=1_x_8&enrichId=rgreq-7746708b-7f33-4148-a770-ae20c40c4d35&enrichSource=Y292ZXJQYWdlOzI2MjMxNTQ4NjtBUzoxMjAzMTgyMTEzOTk2ODFAMTQwNTY5NzUwNTUyMg==

https://www.researchgate.net/publication/211391427_Why_People_Stay_Using_Job_Embeddedness_to_Predict_Voluntary_Turnover?el=1_x_8&enrichId=rgreq-7746708b-7f33-4148-a770-ae20c40c4d35&enrichSource=Y292ZXJQYWdlOzI2MjMxNTQ4NjtBUzoxMjAzMTgyMTEzOTk2ODFAMTQwNTY5NzUwNTUyMg==


https://www.researchgate.net/publication/230087873_Rapid_Skill_Obsolescence_in_an_IT_Company_A_Case_Study_of_Acxiom_Corporation?el=1_x_8&enrichId=rgreq-7746708b-7f33-4148-a770-ae20c40c4d35&enrichSource=Y292ZXJQYWdlOzI2MjMxNTQ4NjtBUzoxMjAzMTgyMTEzOTk2ODFAMTQwNTY5NzUwNTUyMg==

Figure 1. Research Model

Mitchell et al. (2001) identified three critical aspects to embeddedness, each of which has both organization-related (on-the-job) and community-related (off-the-job) facets: Fit, Sacrifice, and Links. Each of these is described below. Fit Fit is the degree to which the jobs and communities of people are similar to or fit with the other aspects of their life spaces. While community-related fit includes enjoying leisure activities that are offered and loving the community in which they live, our focus is organization fit. Organization-related fit is essentially an individual’s cognitive assessment of compatibility with his or her organization. The greater the compatibility of an employee’s career goals, personal values, knowledge, skills and abilities with the requirements of the job and culture of the organization, the greater the fit and thus, the greater the connection to the organization. If the individual’s overarching approach to work and the organization are not congruent, then the fit with the organization is diminished. Conversely, the greater the consistency, the more strands or connections the employee has to the organization.

Sacrifice The second dimension of embeddedness, sacrifice, identifies an individual’s cognitive assessment of financial and psychological benefits that may be forfeited upon leaving his or her job. Community-related sacrifices might include leaving a safe neighbor or moving from a town in which the individual is respected. Organization-related sacrifices include tangible items such as loss of health benefits or vacation time as well as psychological sacrifices such as leaving an interesting project or seniority. The importance of the entities being given up likely varies between individuals depending upon their life situation, overall goals, or personal preferences. For example, one employee may cherish the short commute while another employee relishes the substantial benefits package that awaits them upon retirement. Shaw, Delery, Jenkins, and Gupta (1998) suggest that many of the “perks” that would be sacrificed serve as anchors to retain employees in their current organizations, and also serve as ways to preserve their overall self-interest. A few important examples of this type of self-interest preservation facilitated through remaining with an organization are job stability, advancement possibilities, and tenure.

Job Embeddedness

Organizational Fit

Organizational Sacrifice

Organizational Links

Perceived Obsolescence

Turnover Intention

H1

H3

H2

Job Satisfaction Job Alternatives

Age



Links Links are the degree to which people have connections to other people or activities. Organization-related links are the psychological or social connections that the employee has with the organization and other employees within the organization such as being involved in committees or work teams. Prior research (Lee and Mowday, 1987, Mitchell et al., 2001) investigated how employee embeddedness affects the desire to leave by examining the employees’ perceptions of their fit in their jobs and organizations or the number of links the employees have with other individuals within their work environments. Lee et al. (2004) theorized that employees with a greater number of links to coworkers are more motivated than those employees with fewer links and that embeddedness was a significant predictor of both in-role and extra-role job performance. They further found that the more individuals are socially entangled in an organization, the more likely they care and are involved in contextual performance. Embedded employees tend to have larger social networks and get help from their network to perform better in their jobs (Settoon et al., 1996). Employees become embedded through extensive and extra-role services such as working with colleagues and teams (Hom et al., 2009). In addition, Hom et al. (2009) found that such workplace links improve job attitudes. Unlike the attitudinal perceptions that exist in the organization-related fit dimension, organization-related links as defined by Mitchell et al. (2001) are discernible connections between the employee and other facets of the organization, that is, they are objective factors. For example, the number of employees one interacts with on a regular basis is easily identified and can be observed. It is these interactions and connections with the individual, team, department, or organization that bind the employee closer and stronger than those with little or no such links within the enterprise and increase the likelihood of the employee’s retention. Community-related links are established in the community as a result of being employed by their current organization in that particular location, which might include owning a home or having roots in the community through membership in religious or civic organizations. Cohen (1995) found that hobbies and church-related activities were likely to engender commitment. This suggests that individuals have links among different facets of their lives.

Embeddedness and Its Relationship to Turnover Intention The Embeddedness construct helps to explain why people stay in their jobs and thus reduces voluntary turnover intention (Harman et al., 2007). Although Mitchell et al. (2001) purported that the three embeddedness dimensions have both organizational and community aspects, because of labor laws, union issues, and organizational policies, the current study did not include the non-work related (community aspects) of job embeddedness. Participation in community groups such as political, religious, or other organizations is not a part of the terms and conditions of employment that are under the organization’s direct control, therefore were excluded. Other previous research (Hom et al., 2009; Sekiguchi et al., 2008) has also taken this approach. Holtom et al. (2006) used examples from the Fortune 100 Best Companies to Work For to describe how embeddedness can be used to build human and social capital by increasing employee retention. Operationalizing embeddedness by averaging preexisting items found on the Graduate Management Admissions Council Wave III survey rather than Mitchell et al.’s (2001) scale, Holtom and Inderrieden (2006) showed that embeddedness predicts turnover intention when gender and job satisfaction is controlled. Mallol et al. (2007) found that on-the-job embeddedness is a predictor of turnover intention when controlling job satisfaction and organizational commitment for respondents working in culturally diverse companies. Felps et al. (2009) studied two samples, a large recreation and hospitality organization and a retail bank in the U.S. Midwest, and found that job embeddedness explained variance in individual voluntary turnover beyond that explained by other individual and group-level predictors. A few studies examined the embeddedness construct and concluded that embeddedness makes a unique contribution in predicting voluntary turnover (Cohen, 2006; Dess and Shaw, 2001; Holtom et al., 2006; Lee et al., 2004). Joseph et al. (2007) propose to include embeddedness as a factor in their contextual IT turnover model. Embeddedness has been investigated in the management literature and the negative relationship was validated between embeddedness and turnover intention with several different types of workers (e.g., grocery employees, nurses, admittance clerks). The literature suggests that there may be differences in the characteristics of IT workers and other types of workers that could affect employee embeddedness. First, a number of researchers have investigated the


personality types of IT workers. Using the Myers-Brigs Type Indicator (MBTI), a psychometric instrument that describes different personality types, past researchers have found that computer professionals, especially the more technically oriented ones, often have a personality type ISTJ: being Introverted (I), preferring to work alone and being less oriented toward social interaction, Sensing (S), preferring to seek detailed information and actual facts, Thinking (T), making decisions based on logic and objective consideration, and Judging (J), being organized and establishing deadlines they expect others to follow (Bush and Schkade, 1985; Capretz, 2003). Thus, while this personality type is characterized by being responsible, dependable, and persevering, they are also more at ease with working with facts and logic than socially interacting with people (Gorla and Lam, 2004; Hunter, 2009). Good communication skills are necessary for effective teamwork and successful interaction with users, yet ISTJ personality types are often poor communicators, thus affecting their team skills. As a result, there has been a great deal of emphasis in IS curricula stressing communication and team skills (Topi et al., 2009). One of the subdimensions of embeddedness, links, concerns the connections that individuals have with others, both formal and informal. Those that are less comfortable with social interaction are less likely to have robust links within the organization. A few researchers have suggested that IT workers feel a low sense of belonging to their organizations (Lim, 2007; Whalley and Barley, 1997). This may impact embeddedness because embeddedness concerns the connectedness that one perceives with the organization. Moreover, IT workers differ from other types of workers in that they perform significantly fewer Organizational Citizenship Behaviors (OCB), such as helping a colleague even when there is no explicit remuneration for doing so (Moore and Love, 2005). Organizational Citizenship Behaviors have been defined as "individual behaviors that are discretionary, not directly or explicitly recognized by the formal reward system" (Organ, 1988, p. 4). They represent ‘extra effort’ by employees. By performing these extra-role behaviors, connections form between individuals, thus, resulting in greater embeddedness. To our knowledge, the relationship between embeddedness and turnover intention has not yet been empirically tested with a sample of IT workers. Because of the possible differences between IT and non-IT workers as well as the high organizational costs related to IT turnover, we argue that it is useful to verify the embeddedness to turnover intention

relationship with an IT worker sample. We posit that the negative relationship found in prior studies will be confirmed in an IT worker sample. Therefore, we hypothesize:

H1: There is a negative relationship between embeddedness and turnover intention. H1a: Embeddedness will explain significant incremental variance in turnover intention above the traditional model variables (job satisfaction and job alternatives).

Perceived Obsolescence and Its Relationship to Turnover Intention An important area of research is how technological trends affect turnover decisions of IT professionals. Obsolescence has been defined as "the degree to which professionals lack the up-to-date knowledge and skills necessary to maintain effective performance in either their current or future work roles” (Ang and Slaughter, 2000, p. 23). Technological change causes formal education and previously acquired skills to quickly become obsolete (Banel and Sicherman, 1998). Thus, the technological environment erodes IT skills currently held and demands that new skills be attained. When IT employees feel that their skills are not up-to-date but value a career of achievement, they are more likely to consider job opportunities that will allow them to gain new skills (Kaplan and Lerouge, 2007). IT professionals need current IT knowledge and skills because it impacts their employability, compensation, and career advancement possibilities (Joseph et al., 2010). The lack of up-to-date skills reduces an IT workers’ value in a competitive external labor market. Coombs (2009) found that technical development, which included opportunity to use new technology and opportunity to be trained to use new technology, were important behavioral beliefs that explained why IT workers decided not to leave an organization. Similarly, The Gartner IT Market Compensation survey (Business Communications Review, 2007) shows lack of career development opportunities, and thereby increasing obsolescence, as one of the more important reasons for turnover. Thus, these studies suggest that there is a connection between perceived obsolescence and voluntary turnover. We argue that when technical IT professionals see themselves becoming obsolete if they stay with their current organization, they are more likely to turnover; however, age weakens this relationship. Expectancy theory has demonstrated that


individuals calculate the subjective expected value of anticipated gains and losses from any potential career-related choice (Steiner and Farr, 1986). Individuals’ expectancies and valences tend to decrease with age (Kopelman, 1977). The calculation partially discounts outcomes which are further in the future because they have lower probabilities. Likewise, Human Capital Theory proposes that employees may differ in their desire to make changes as a function of the anticipated payoff for their efforts (Becker, 1993). Individuals who are older may be less willing to leave an organization because of a shorter stream of payoffs. Consequently, older IT workers whose skills are beginning to obsolesce may not be as likely to give up their current job. This is consistent with the meta-analysis finding that shows a medium effect size between voluntary turnover and age, with older workers less likely to turnover (Ng and Feldman, 2009). Consistent with this body of literature we developed the following hypothesis:

H2: The positive relationship between perceived obsolescence and turnover intention will be moderated by age such that the higher the age the weaker the effect.

Perceived Obsolescence and Its Relationship to Embeddedness Embeddedness is conceptualized as a web which entangles an employee so that he or she cannot get out (Mitchell et al., 2001). We argue that perceived obsolescence should be incorporated in the nomological net with embeddedness because of its salience for IT workers. We propose that while the relationship between perceived obsolescence and embeddedness is negative, it is moderated (weakened) by age. As technical IT workers’ skills become obsolete, we argue that they may feel less able to help their company take advantage of swiftly changing technologies and, a result, may feel anxious and unfulfilled in achieving their professional ambitions. As a response, their perceived fit and embeddedness within their organization goes down (becomes more negative). In addition, as IT workers’ skills become increasingly obsolete, employees may have fewer opportunities for skill-related promotions or pay raises. Without these opportunities they have little to sacrifice if they left the company.

While we propose that the relationship between perceived obsolescence and embeddedness is negative, previous research has found many organization-based attitudes change as employees age. For example, Ng and Feldman (2010) found there is a relationship between age and person-organization fit, organizational identification, affective commitment, and normative commitment, which are related to the fit dimension of embeddedness. Other studies have also found a relationship between age or tenure and person-organization fit in terms of personal and organizational values congruency (e.g., Posner, 2010; Bretz, and Judge, 1994). Thus, as employees age, the negative relationship between perceived obsolescence and embeddedness is weakened. Based upon “sunk cost theory” (Staw and Hong, 1995), we also argue that there may be an interactive relationship between age and perceived obsolescence, leading to embeddedness. Sunk cost theory suggests that individuals consider the amount of time and effort already invested when making decisions (Staw and Hong, 1995). Individuals who are older often have a greater investment than those who are younger in terms of having more experience with older technologies used at their particular organization. These “sunk costs” in older technologies may be perceived as strands that embed employees in their organizations. We therefore hypothesize:

H3: The influence of perceived obsolescence in decreasing embeddedness will be moderated by age such that the higher the age the weaker the effect.

Methods

Survey Development and Pilot Study

The research team worked with a U.S. Department of Veteran Affairs (VA) taskforce comprised of 27 employees from across the 230,000-employee organization in developing, reviewing, and piloting a survey to test our proposed model. The survey went through theoretical development, review with the VA taskforce, and a thorough piloting process. The research team searched the academic literature and used or modified preexisting scales for inclusion in the survey. An initial set of survey questions that measured the research constructs and collected data on demographic characteristics, training, and career aspirations was presented to the VA taskforce. The research team and the VA taskforce met both in person and through teleconferencing for a period of months in an effort to develop an online survey able to both gather the required information


and satisfy the legal and regulatory concerns of all parties, including five labor unions. During this time, demographic questions specific to the VA and other government institutions (such as grade level and job classification) were added. Questions were modified for the organizational context (e.g., It is likely that I will be working at VA this time next year). Once the initial survey was developed, it was put on a secured Internet site at the research team’s university. The VA taskforce and their designees, 38 in total, completed the survey. Written and oral comments were provided through email and teleconferencing, respectively. The piloting process not only tested the appearance of the survey and ease of use, but also allowed the researchers to test the database responsible for holding the responses. The database was accessible only by the research team and not by the VA or members of the taskforce. Modifications were made to resolve technical problems and after two additional rounds of pilot testing with the 27 member taskforce, the final survey was established. Survey Administration VA supervisors forwarded an email that they received from the taskforce to the IT workers in their departments, along with a letter from the Assistant Secretary for Information and Technology (i.e., the CIO) asking that all IT workers complete the survey. Each email contained the URL of the survey website, instructions for logging in and completing the survey, and the phone and email contact information of the research team and help desk in case of emergencies or questions. Each potential respondent was informed that all individual responses would be kept anonymous and that only aggregate data would be presented in results and reports. The target population was 3,579 non-clerical and non-managerial IT workers across the two major VA divisions – health and benefits. As a result of the data cleaning process (employees who entered invalid emails for ID purposes were eliminated as well as participants providing erroneous data such as specifying opposite attitudes on reversed items), the final number of usable respondents was 1,777, resulting in a usable response rate of 49.7%. Measures Previously validated measures were used whenever possible. Table 2 contains weights information on embeddedness. The full set of items for turnover intention, job satisfaction, job alternatives, and perceived obsolescence are showed in Table 3.

The items retained after item purification are shown in bold and shaded in gray. Items from the Jackson and Turner (1987) Turnover Intention Scale, also used by Moore (2000), were used on the survey to assess the employee’s attitudes towards leaving the organization within the next calendar year. Turnover intention was operationalized as the likelihood that a person will seek employment elsewhere rather than remaining in his/her present job. To assess overall job satisfaction, we used an abbreviated version of the Cammann, Fichman, Jenkins, and Klesh (1981) Michigan Organizational Assessment Questionnaire Subscale. Job satisfaction is defined as the degree to which an individual has a positive emotional state or feeling resulting from the appraisal of his/her job or job experience. Availability of perceived job alternatives is a subjective scale based on the perceived ease by which one can find another job. This scale is subjective because it is not possible to identify accurately the exact number of jobs available. Two items used for this scale were adapted from Lee and Mowday (1987). Perceived Obsolescence assesses the degree to which employees assess their skills as becoming outdated. Items, taken from Ryan and Harrison (2001), were measured on a 7-point Likert scale ranging from Strongly Agree to Strongly Disagree. The measures for all three embeddedness’ dimensions were taken from Mitchell et al. (2001). The first dimension, organizational fit, is the perceived compatibility or comfort of an employee with an organization and with his/her environment. Organizational fit encompasses the ideas related to how well the employee perceives their personal values, goals, and needs to be in harmony with that of the organization. The second dimension of organizational embeddedness is organizational sacrifice, which is the perceived cost of material or psychological benefits that potentially would be forfeit if the employee left the organization. This facet of organizational embeddedness is assessed with questions such as “I would sacrifice a lot if I left this job,” “I am well compensated for my level of performance,” and “I have a lot of freedom on this job to decide how to pursue my goals” (Mitchell et al. 2001).


The last dimension of the organizational embeddedness scale is organizational links. This component of the Organizational Embeddedness Scale assesses the quantifiable links that the employee has with the organization, other employees within the organization, and links established in the community as a by-product of being employed by their current organization. Results We used PLS as our statistical analysis tool for several reasons. First, the objective of PLS is to maximize the variances explained (Barclay et al., 1995). PLS trades parameter estimation efficiency for prediction accuracy and simplicity, and has fewer assumptions than many other statistical methodologies. Second, Falk and Miller (1992) state that it is also appropriate when cross-sectional survey data are used. Third, a few of our constructs contain two items. However, covariance based SEM (CBSEM) such as LISREL requires a minimum of three items (Kenny et al., 1985). Fourth, embeddedness is a formative second order construct that consists of three formative first order constructs. Though the inclusion of formative measures in CBSEM such as LISREL has been reported (MacCallum and Browne, 1993; Jöreskog and Sörbom, 1996) researchers find that identification problems with such models are common. In contrast, similar problems do not arise with PLS path modeling. Fifth, PLS path modeling is methodologically superior to CBSEM whenever improper or nonconvergent results are likely (Krijnen et al., 1998). In addition, our model is complex, containing both reflective and formative constructs as well as higher order constructs. Wold (1985) notes that PLS path models are appropriate when models are complex with many latent and manifest variables without resulting in estimation problems, and in these cases, PLS is better than other approaches. Therefore, PLS is an appropriate analytical tool. Demographics The average number of years of IT experience for the VA IT worker is 12.3 with a standard deviation of 8.92. The VA IT workforce is older than many IT shops with 60.3% of the workers above the age of 46 and many of them having sufficient longevity that they are eligible for retirement within 10 years. Males make up nearly 59% of the VA IT workforce. The VA is a well-educated organization, with 43.6% of its IT workforce holding a college degree and another 25.6% having some graduate school

experience or holding a graduate or terminal degree. Table 1 shows the job title of the respondents and the percentage composition of the sample.

Table 1. Job Titles

Title Percent

Computer Software Engineer – Applications 65.0%

Computer Software Engineer – Operating Sys 5.9%

Customer Support 5.6%

Systems Analyst 1.1%

Data Base Administrator – Data Mgt. 10.2%

Network & Comp. Administrator 4.6%

Network & Data Analyst 1.5%

Other IT specialist 6.1% Assessment of Measurement Model

According to Jarvis et al. (2003), the choice between the formative and reflective constructs is theory driven. Prior works (Diamantopoulos and Winklhofer, 2001; Jarvis et al., 2003) summarize the differences between reflective and formative indicators. These include 1) the direction of causality is from items to construct for formative indicators whereas the direction of causality is reversed for constructs with reflective indicators; 2) indicators are not necessarily similar in content for formative indicators whereas indicators are interchangeable for reflective indicators; 3) indicators may not co-vary for formative indicators whereas indicators are expected to co-vary for reflective indicators; and 4) formative indicators may not have the same antecedents and consequences whereas reflective indicators have the same antecedents and consequences. Consistent with these issues, the definition of reflective constructs, and other relevant prior literature (Cammann et al., 1981; Jackson and Turner, 1987; Lee and Mowday, 1987; Moore, 2000) turnover intention, perceived job alternatives, job satisfaction, and perceived obsolescence are modeled as constructs with reflective indicators. Because our model contains both reflective and formative constructs, we need to use different approaches to validate these two types of constructs. In the following section, we validate the formative construct embeddedness first, followed by the four reflective constructs. We followed the guidelines suggested by Diamantopoulos and Winklhofer (2001) to validate the formative construct.


Validity for Formative Construct - Embeddedness Content Validity Consistent with Jarvis et al.’s (2003) criteria as mentioned above, Mitchell et al. (2001) theorized that embeddedness is a formative construct consisting of fit, sacrifice, and link. That is, the causal direction is such that the dimensions of fit, sacrifice, and link form the second order construct embeddedness. According to Roberts and Thatcher (2009), the formative nature of a construct meets Jarvis et al.’s (2003) first criterion if the formative construct increases in value as one or more of the indicators increases in value. This is the case in our data. For example, as fit increases in value embeddedness also increases. Jarvis et al.’s (2003) second criterion deals with the interchangeability of constructs. Formative indicators may not be interchangeable and will often have different themes. In our data, fit, sacrifice, and link cover different aspects of embeddedness. For example, fit items, “My values are compatible with the organization's values” and “I can reach my professional goals working for this organization,” are different in capturing the theoretical portion of embeddedness than are those measuring the sacrifice and link dimensions. Jarvis et al.’s (2003) third criterion states that covariation of indicators is not necessary. In our case, items used to measure fit such as, “My coworkers are similar to me” may not covary with items to measure sacrifice, such as “My promotional opportunities are excellent here.” The fourth criterion deals with whether or not the indicators have the same antecedents and consequences. According to Roberts and Thatcher (2009, p. 13), “formative constructs are composites made up of indicators that may be very different.” While we do not analyze the antecedents in this section, our post hoc analysis shows that an important antecedent of fit and sacrifice, perceived obsolescence x age, is not an antecedent of link. Weights are important in evaluating the relative contribution of each measure for the formative constructs. We used PLSGraph 3.0 to perform the analysis. Table 2 shows weights, standard errors, and t statistics for all of the embeddedness items. As shown in Table 2, some items have low weights and low t-values. To purify items for the formative construct, care must be taken in removing items because deleting measures changes the definition of the formative construct (Jarvis et al., 2003). Sac5 had the lowest t-value. After examining the content validity of sac3 (The benefits are good on this job),

sac4 (The retirement benefits provided by this organization are excellent), sac5 (The health-care benefits provided by this organization are excellent), and sac9 (The perks on this job are outstanding), we deemed that sac3 was sufficient by this to represent the themes for sac4, sac5, and sac9. Therefore, these were removed based on t-values and content examination. We performed similar steps to purify the items for fit and link. We carefully examined all the items and felt that each item has its unique perspective. Nunnally and Bernstein state that “the breadth of definition is extremely important to causal indicators” (Nunnally and Bernstein, 1994, p. 484). We decided that no further item deletion was needed. In Table 2, the items that remain after purification are in bold and shaded in gray. Multicollinearity

One issue that needs to be considered for formative constructs is multicollinearity because stability of the coefficients is affected by the correlations among the indicators (Bollen, 1989). Though the conventional suggested VIF value is 10 (Kleinbaum et al., 1988), Hensler, Ringle, and Sinkovics recently (2009, p. 302) recommended that “any VIF substantially greater than 1 indicates multicollinearity.” We checked VIF values for all of the items measuring embeddedness and found that the VIF values range from 1.01 and 1.67 and this range of VIF is not considered substantially greater than 1. Therefore, multicollinearity is not a concern. Non-response Bias We followed the procedures of Armstrong and Overton (1977) and used ANOVA to analyze the non-response bias. Our test results show that there are no significant differences on the study variables and demographic variables except the age group. However that difference reflects age using 11 categories with 6 indicating ages 41-45 and 7 showing ages 46-50. The mean age group in the early response is 6.72 and the mean age in the late response is 6.5. Consistent with the groupings the mean age category for the early and late responses could be converted to the ages of 44.6 and 43.5 respectively. However, both the mean category of the early and late respondents are within the same group 6 classification (41-45) and one year difference in 40 plus years is of little practical significance in the workforce.


Table 2. Weights for Embeddedness

Item Weights Std. Error T-Statistic

Organizational Fit (fit)

I like the members of my work group. fit1 0.08 0.05 1.54

My coworkers are similar to me. fit2 0.06 0.04 1.81

My job utilizes my skills and talents well. fit3 -0.03 0.04 0.77

I feel like I am a good match for this company. fit4 0.40 0.06 6.88

I fit with the company's culture. fit5 0.07 0.06 1.15

I like the authority and responsibility I have at this company. fit6 0.38 0.05 8.26

My values are compatible with the organization's values. fit7rev 0.15 0.06 2.67

I can reach my professional goals working for this organization. fit8 0.19 0.05 3.86

I feel good about my professional growth and development. fit9 0.21 0.05 4.50

Organizational Sacrifice (sac)

The prospects for continuing employment with this company are excellent. sac1 0.72 0.04 18.06

I would not sacrifice a lot if I left this job. sac2(r) 0.24 0.05 4.87

The benefits are good on this job. sac3 0.13 0.05 2.80

The retirement benefits provided by this organization are excellent. sac4 0.02 0.04 0.50

The health-care benefits provided by this organization are excellent. sac5 0.00 0.04 0.18

I feel that people at work respect me a great deal. sac6 0.13 0.05 2.84

I am well compensated for my level of performance. sac7 0.09 0.05 2.05

I have a lot of freedom on this job to decide how to pursue my goals. sac8 0.15 0.06 2.44

The perks on this job are outstanding. sac9 0.06 0.05 1.27

My promotional opportunities are excellent here. sac10 0.14 0.05 2.92

Organizational Links (link)

How many years have you been in your current position? link1 -0.44 0.14 3.48

How long have you been with VA? link2 1.12 0.13 9.30

How many years of federal service do you have? link3 -0.30 0.18 1.80

How many coworkers do you interact with regularly? link4 0.11 0.11 0.62

How many coworkers are highly dependent on you? link5 -0.16 0.15 0.83

How many work teams are you on? link6 -0.28 0.17 1.94

How many work committees are you on? link7 -0.10 0.09 0.88

Nomological Validity To validate a formative construct, nomological validity is required. Nomological validity is when a construct behaves as it should within a system of related constructs called a nomological net (Baggozi, 1994). To establish nomological validity one needs to link a formative construct to theoretical relationships that exist and to other reflective constructs (Diamantopoulos and Winklhofer, 2001; Petter, Straub, and Rai 2007; Hensler, Ringle, and

Sinkovics 2009; and Roberts and Thatcher 2009). First, our structural model is built upon theory developed from prior studies (see Joseph et al., 2007). Second, Henseler et al. (2009) suggest the relationship between the formative construct and the other construct should be strong and significant. In our model, intention is a reflective construct and we test the relationship between embeddedness and intention. The path between embeddedness and intention is significant at the 0.01 level and the path


coefficient is -0.44. Therefore, nomological validity is established. Validity for Reflective Constructs Convergent Validity Convergent validity is established because all of the loadings are statistically significant and greater than 0.7 (Chin et al., 1998). Following this rule, we removed items with loadings less than 0.7. Table 3 shows descriptive statistics, composite reliability, loadings and average variance extracted for reflective constructs: perceived job alternatives, job satisfaction, turnover intentions, and perceived obsolescence. Composite reliability is in the range of 0.80 and 0.91, greater than the suggested 0.7 (Fornell and Larcker, 1981). The items that remain after purification are in bold and shaded in gray. In addition to loadings and composite reliability, we also checked average variance extracted. The average variance extracted (AVE) for these four constructs is between 0.58 and 0.83, higher than the

recommended 0.5 (Fornell and Larcker, 1981). Thus, convergent validity is established. Discriminant Validity The first way to evaluate discriminant validity is to compare the AVE values with the square of the correlations among the latent variables. The AVE values should be higher (Chin et al., 1998). This is the case in our data. Another approach to evaluate discriminant validity is to examine item to construct relationships to ensure that items correlate higher to the construct they intend to measure than to other constructs (Campbell and Fiske, 1959). Table 4 shows the correlation among constructs. The bold number on the diagonal is the square root of average variance extracted for the reflective constructs. The AVEs are much higher than the correlations among other constructs. In addition, Table 5 shows item to construct correlation. Each indicator correlates more highly with its own underlying construct than with other constructs.

Table 3. Mean, Standard Deviation, Loadings, Average Variance Extracted, and Composite Reliability

Item Mean Std Dev.

Loadings AVE CR

Turnover Intention It is likely that I will be working at VA this time next year (R).

2.56 1.09 0.86 0.76 0.86

I will be working at VA five years from now (R).

3.55 1.54 0.88

It is likely that I will take steps during the next year to secure a job at a different organization.

3.36 1.88 0.67

I will probably look for a job at another organization in the next year.

3.07 1.77 0.62

Job Alternatives There is a good probability that I could find an acceptable alternative to my job outside of VA.

4.47 1.54 0.92 0.76 0.86

If I searched for an alternative job, within a year I could find an acceptable job.

4.78 1.43 0.81

Job Satisfaction All in all I am satisfied with my job. 4.72 1.39 0.93 0.83 0.91

In general, I like working here. 5.09 1.13 0.89

In general I don’t like my job (R). 5.66 1.66 0.49

Perceived Obsolescence I feel like some of my skills are becoming obsolete.

3.66 1.64 0.92 0.59 0.81

I do not feel as if my current job skills are becoming outdated (R).

4.38 1.77 0.81

There are lots of changes to this job that I wish I knew more about so that I could do better.

3.51 1.60 0.92

My skills seem to be lagging behind the pace of change in this kind of job.

4.28 1.49 0.49


Table 4. Construct Correlation Table

jobsat jobalt ti fit sac link embed po age age x po

jobsat 0.91

jobalt -0.08 0.87

ti -0.49** 0.13 0.87

fit 0.70** -0.05 -0.47** NA

sac 0.55** -0.10* -0.54** 0.61** NA

link 0.07 -0.09 -0.15* 0.09 0.12* NA

embed 0.62** -0.11* -0.58** 0.75** 0.96** 0.28** NA

po -0.25* 0.01 0.15* -0.35** -0.23** 0.02 -0.25** 0.77

age -0.09* 0.10* 0.08 -0.10* -0.09* 0.15** -0.07* 0.07* NA

age x po -0.20* 0.07 0.15* -0.27** -0.19** 0.11* -0.19** 0.65** 0.74** NA * The bold number on the diagonal is the square root of AVE for the reflective constructs. **: correlations are significant at 1% level. *: correlations are significant at 5% level.

Table 5. Item to Construct Correlation

jobalt jobsat obsol intention

jobalt1 0.94 -0.07 0.02 -0.04

jobalt2 0.81 -0.05 0.05 -0.03

jobsat1 -0.14 0.93 -0.20 -0.21

jobsat2 -0.05 0.88 -0.18 -0.17

perobsol1 0.04 -0.15 0.72 -0.06

perobsol2 0.00 -0.14 0.83 0.11

perobsol3 0.05 -0.19 0.74 0.01

intention1 0.10 -0.39 0.09 0.84

intention2 0.12 -0.36 0.09 0.87

Table 6. First Order Constructs Construct Coefficient Significance levelFit 0.35 0.001 Sacrifice 0.72 0.001 Link 0.07 0.05 Common Method Bias We used Harman's one-factor test (Harman, 1976) to check for common method bias on the four reflective constructs: job alternative, job satisfaction, intention, and perceived obsolescence. All the variables were entered into an exploratory factor analysis, using unrotated principal components factor analysis and principal component analysis with varimax rotation to determine the number of factors that are necessary to account for the variance in the variables. According to Podsakoff et

al. (2003), if common method variance is present, either (a) a single factor emerges from the factor analysis, or (b) one general factor accounts for the majority of the covariance among the variables. Our results show four factors account for 68% of the variance in the data. The first factor accounts for 22.8% of the variance, not a majority. Therefore, the risk of common method bias in this sample is considered minimal. Results of the Structural Model Figure 2 shows the results of PLS analysis. The path coefficient is -0.44 for the path between embeddedness and intention, which is significant at the 0.01 level. The R-square for our model is 37.3% when embeddedness and perceived obsolescence moderated by age are used to predict turnover intention, with job satisfaction and job alternatives used as control variables. This is higher than other research investigating IT turnover intention (e.g., 29% with organizational commitment, perceived job alternatives, and gender as significantly related to turnover intention (Thatcher et al., 2002/2003); 34% with affective organizational commitment and job satisfaction as direct antecedents to turnover intention (Thatcher et al., 2006); and 28% with job satisfaction, fairness of rewards, and perceived workload significantly predicting turnover intention (Rutner et al., 2008). The first-order constructs organizational fit, sacrifice, and links are significant in forming the second-order construct – embeddedness.


Figure 2. PLS Results (significance level of t statistics: *=0.1; **=0.5; ***=0.01)

In order to analyze further the predictive power of embeddedness and perceived obsolescence over traditional turnover intention variables, we ran several models. Model A has only the traditional variables of job satisfaction and perceived job alternatives as independent variables to predict turnover intention, and resulted in an R-square of 25.2%. Model B extends Model A by adding the independent variable of job embeddedness to predict turnover intention, and resulted in an R-square of 37%. Model C extends Model B by including perceived obsolescence and the interaction of perceived obsolescence with age as predictors, achieving the 37.3% R-square.

Each construct’s effect size (f2) is calculated using the formula:

)( 22partialfull RR / )1( 2

fullR (Mathieson et al., 2001, Chin et al., 2003)

Cohen (1988) suggests that an effect size of .02 is small, .15 is medium, and .35 is large. Our calculation of the effect size is 0.19. This support H1a, which states that embeddedness will explain significant incremental variance in turnover intention above the traditional model variables (job satisfaction and job alternatives). It does not, however, support H2, which is discussed further below.

Tests of Moderators We have two moderators in our model, where age moderates the relationship between perceived obsolescence to turnover intention and age moderates the relationship between perceived obsolescence to embeddedness. We followed the suggestions by Chin et al. (2003) to create a moderator age x perceived obsolescence by using the indictor of age to multiply each indicator of perceived obsolescence.

Table 7. Comparison of Models Models Predictor

Variables R Square

Coefficients

Model A

Job Sat and Job Alts

25.2% All paths are significant.

Model B

Job Sat, Job Alts and JE

37% All paths are significant at 1% level

Model C

Job Sat, Job Alts, JE, PO, and PO x Age

37.3% All paths are significant except the paths from PO to TOI, and PO x Age to TOI

PLS was used to test the model and PLS uses the standardized scores when computing the results.

Job Embeddedness

Organizational Fit

Organizational Sacrifice

Organizational Links

Perceived Obsolescence

Turnover Intention R2=37.3

H1

H3

H2

Job Satisfaction Job Alternatives

Age

-0.22***

-0.44***

0.055

-0.02** 0.01**

-0.002

0.02**

-0.03

-0.07***


We ran the analysis with separate tests for the two endogenous variables and simultaneous testing on the two endogenous variables but the results remained the same. H2, which states that the negative influence of perceived obsolescence on turnover intention will be moderated by age such that the higher the age the weaker the effect was not supported (t value: 1.16). H3, which states that the influence of perceived obsolescence on embeddedness will be moderated by age such that the higher the age the weaker the effect was supported (t value: 4.21). Figure 3 shows a graph of the interactive effect of perceived obsolescence and age on embeddedness. Post Hoc Analysis Because perceived skill obsolescence is an antecedent to embeddedness, and prior studies suggest that perceived skill obsolescence is related to the fit and sacrifice dimensions of embeddedness (Tsai et al., 2007), we disaggregated the dimensions of embeddedness in our model to examine the impact of perceived skill obsolescence on fit and sacrifice, as well as turnover intention. We first examined a model with perceived skill obsolescence as a predictor and fit and sacrifice as dependent variables. The result shows that all the paths are

significant with skill obsolescence explaining 19.2% of variance in fit and 8.5% in sacrifice. The path coefficients are -0.44 for the path between perceived skill obsolescence and fit, and -0.29 for the path between perceived skill obsolescence and sacrifice. We then examined a model containing perceived skill obsolescence, fit, sacrifice, and turnover intention. In this model, perceived skill obsolescence is an antecedent to fit, sacrifice, and intention. Fit and sacrifice mediate the relationships between perceived skill obsolescence and turnover intention. Our results show that all the paths are significant at 0.01 level except the path from perceived skill obsolescence to intention. Perceived skill obsolescence explains 15.8% of variance in fit and 5.8% of variance in sacrifice. Fit, sacrifice, and perceived skill obsolescence explain 31.7% of variance in turnover intention. Lastly, we added job satisfaction and job alternatives as predictors for turnover intention to our second model. Our results show that all the paths are significant at 0.01 level except two: the path between fit to turnover intention and the path between perceived skill obsolescence to turnover intention. We performed Sobel test on the mediation effect and concluded that sacrifice fully mediates the relationship between perceived skill obsolescence and turnover intention.

Figure 3. Interactive Effect of Perceived Obsolescence and Age (Low, Medium, High) on Embeddedness


Perceived skill obsolescence explains 15.7% of variance in fit and 5.3% of variance in sacrifice. The five predictors (job satisfaction, job alternative, fit, sacrifice, and perceived skill obsolescence) for turnover intention collectively account for 35.5% of variance in turnover intention. Discussion: Implications, Limitations, and Conclusions IT worker voluntary turnover is expensive for organizations and therefore investigating ways to minimize voluntary turnover is useful and important. Past research has shown that the job embeddedness theory (Mitchell et al., 2001) explains additional incremental variance in turnover intention of various types of workers beyond the traditional turnover model consisting of job satisfaction and job alternatives. Our research contributes to the IT discipline by extending the embeddedness theory to account for contextual influences. The threat of professional obsolescence is especially relevant to IT professionals, but has not been included in the nomological net with job embeddedness when examining IT workers’ turnover intentions. We proposed a theoretical model for predicting turnover intentions that incorporated embeddedness and perceived obsolescence moderated by age. Two out of three of our main hypotheses were supported. First, our findings confirm the efficacy of embeddedness with a sample of IT workers and further extend the generalizability of the theory. Progressive model testing demonstrates that embeddedness explains incremental variance above traditional turnover variables. Our data also support the hypothesis that the influence of perceived obsolescence in decreasing embeddedness is moderated by age such that the higher the age the weaker the effect. However, in terms of the effect of perceived obsolescence moderated by age on turnover intention, we did not find a significant relationship. In our post hoc analysis, we found that perceived obsolescence was significantly related to two of the embeddedness dimensions: organizational fit and organizational sacrifice. Organizational fit is an individual’s cognitive evaluation of how his or her goals, values, skills, abilities, and present and future plans are consistent with that of his or her job and organization. Because of the rapid rate of technological change, most companies are forced to consider new and different technologies to stay competitive. According to a study of 70 IS/IT managers, the importance of most IT skills is increasing over time (Lee and Mirchandani, 2010).

Currently, skills such as Web 2.0, Security, Business Intelligence, Collaboration, and Architecture are in high demand (Collett, 2011). IT workers need to understand the latest technologies and recognize how they can be leveraged to support their organization (Tsai et al., 2007). It is important to note that our sample consisted of technical (non-managerial) IT professionals as indicated in Table 1. While some IT managers, because of the requirements of their positions, only need a high-level understanding of technology, technical IT workers do feel these pressures (Tsai et al., 2007). Other research suggests that valued IT professionals without technical skill currency may be moved to other positions and contribute to their organizations in different but important ways (Reich and Kaarst-Brown 1999; Reich and Kaarst-Brown, 2003). Therefore, perceptions of organization-related fit most likely depend on the technical orientation of the job position and the skills that the individuals possess. We recommend that future research compare and contrast embeddedness with those in technical job positions versus those in non-technical jobs. The second dimension that was important when considering the effect of perceived obsolescence on embeddedness is what the IT worker would sacrifice if he or she left the organization. While some of the sacrifices one makes in changing jobs has nothing to do with perceived obsolescence, others, we argue, do. As IT workers’ skills become increasingly obsolete, he or she may be less likely to obtain pay raises for technical merit or gain opportunities for skill-related promotions. IT professionals who do not acquire new skills often find that their salaries remain stagnate (Perelman, 2008). Superior skills equates to superior wages (Becker, 1993). Some organizations even require their technical professionals to acquire additional technical skills every year to be eligible for promotion (Mint, 2010). Studies have also shown those developers value technical skills in their colleagues. Hall et al. (2007) confirmed that technical competence was highly regarded and respected among software developers. This demonstrates that when skills become obsolete, technical professionals may have difficulty garnering a good deal of respect from other co-workers. With a lowered probability of obtaining these valued outcomes, the sacrifice of leaving becomes less. The post hoc analysis also showed that sacrifice fully mediates the relationship between perceived obsolescence and turnover intention but that the path where fit mediates the relationship between perceived obsolescence and turnover intention is not significant in the presence of job satisfaction and job alternatives. These findings suggest that turnover


intention involves a complex multivariate model and that future IT turnover research models that examine the relationship between turnover and embeddedness also need to include other variables such as perceived obsolescence, job satisfaction, job alternatives, and age. Embeddedness is a formative construct. This construct structure, in combination with the theory posited by Mitchell (2001), suggests that all three embeddedness dimensions are essential despite their differential individual contribution as mediators in our post hoc analysis. Consistent with including all three dimensions of embeddedness we found that the addition of the links dimension increased the model R-square from 35.5% to 37.3%. Implications for Practice Embeddedness is a construct developed in the organizational attachment literature that reveals the forces constraining employees from leaving their current jobs. Employees with many connections are enmeshed in the organizational web, and therefore are less likely to leave their current employment. Because embeddedness has been described as “a web or net in which an individual can become stuck” (Mitchell et al., 2001, p. 1104), it may have negative connotations. Our results imply that there are both positive and negative aspects to embeddedness. On the positive side, by more thoroughly understanding why employees choose to remain with an organization, more effective retention policies are potentially developed. Our empirical results also show that the negative influence of perceived obsolescence on embeddedness is moderated by age such that the higher the age the weaker the effect. IT workers perceiving skill obsolescence may feel excluded from exciting, high visibility projects, and this may lessen the connection they feel with the company. Organizations should structure their IT workforce practices to avoid skill obsolescence. As employees age, they may have a significant amount of sunk cost invested in having learned older technologies. One particularly important aspect of the cost and benefit evaluation of retooling for IT workers may be the loss of expertise or seniority associated with shifting to different technologies (Ryan, 1999). Therefore, managers can show support for retraining by allocating sufficient funds for retraining, verbally communicating the importance of skill currency, and providing intrinsic rewards such as encouragement or recognition. IT workers have valuable skills that employers need to retain and recognizing that skill obsolescence is part of the complex set of multifactor

relationships involved in IT turnover suggests that employers need to work with IT professionals to allow and encourage them to maintain and enhance their skills. Our findings suggest that maintaining skills, managing the amount of sacrifice, and fitting within the organization are important to the retention of technical professionals. Implications for Research In terms of research, confirming the influence of embeddedness on turnover intention responds to Joseph et al.'s (2007) call for additional research on embeddedness and provides specific evidence for its applicability to IT workers. Our research adds to the literature by incorporating perceived obsolescence in the nomological net with organizational embeddedness when investigating IT workers' turnover intentions. Our findings provide additional insight to researchers investigating perceived obsolescence in IT workers. We found that IT workers' perceptions of the currency of their skills contributed to their perceived job embeddedness. A plausible interpretation is that, in terms of fit, as an IT worker’s skills become obsolete, they may feel as if they are no longer a good match for their company and are not as likely to achieve their professional goals or be satisfied with the possibilities for professional growth and development there. In terms of sacrifice, they have less to sacrifice if they left the company because they may be less likely to receive increased compensation for technical merit, be promoted as a result of their technical skills, or be esteemed for technical prowess by other technical professionals. As employees age, however, the negative correlations perceived obsolescence and job embeddedness weakens. Recently, Joseph et al. (2010) has investigated the relationship between perceived obsolescence, coping strategies (such as direct retraining or indirect coping through emotions), and turnover intention. We recommend that future research also investigate the relationship between perceived obsolescence and the embeddedness dimensions in light of various strategies for coping with perceived obsolescence. Much of the prior IT turnover literature has focused on issues other than embeddedness. However, our model shows that job embeddedness is an essential construct when studying IT turnover because it is a mediator to turnover in IT professionals. The subsequent disaggregated analysis we preformed showed that perceived obsolescence was significantly related to the embeddedness dimensions of organizational fit and organizational sacrifice. Further work on the role of embeddedness in IT turnover is


important because we found that sacrifice mediates the relationship between perceived obsolescence and turnover intention but that the fit and links dimensions involve even more complex interrelationships. Our data also show that the influence of perceived obsolescence in decreasing embeddedness is moderated by age. Given that the higher the age the weaker the effect one can speculate that older workers may be nearing retirement and less concerned about career development and mobility. These results support the contention that this is a model involving relationships worthy of future research. Additional research is also needed to see how embeddedness, perceived obsolescence, and age are related with other variables in the IT literature that are associated with turnover intention. For example, work exhaustion in IT professionals has been found to be significantly related to turnover in the IT workforce (Moore, 2000). Questions arise such as: “Does being embedded in an organization, having strong links with co-workers, or having a good fit with the job and the organization, prevent or decrease the probability of job burnout?” And, “do the perceptions of skill obsolescence and age exacerbate the perception of work overload, and/or job burnout?” With the dispersed nature of many IT organizations, it will also be important to investigate how virtual interaction with coworkers and virtual teams are related to embeddedness, perceived obsolescence, age, and turnover intention. The question of whether the interrelationships between these variables change depending on whether the work context is virtual or co-located should be researched. The relevance of individual personality traits and cognitive style may also have some bearing on the role of embeddedness and perceived obsolescence. The incorporation into embeddedness and perceived obsolescence research of scales such as the Myers-Briggs Type Indicator, the Kirton Adaptation-Innovation Inventory, or others may also be worthwhile and illuminating. Limitations The present study informs researchers and practitioners regarding voluntary turnover in information technology professionals but like any study, it has inherent limitations. One possible limitation of our study is that our model used turnover intention as the dependent variable rather than actual turnover. However, using turnover intention is a common and well-accepted practice within the

literature for several reasons. First, using actual turnover requires access to sensitive personnel records. Next, because of the lag time between intention and action, this type of research takes multiple years to conduct. Third, turnover intention is a variable of interest because it can alert organizations of the need to take actions that would prevent or reduce the occurrence of actual turnover. In addition, meta-analyses have confirmed that turnover intention is a direct antecedent to actual turnover (Griffeth et al., 2000), although the difference between the two constructs is considerably less than perfect. Therefore, future research using actual turnover is also recommended. A second limitation could be attributed to the fact that our respondents were from the federal government rather than the private sector. The results of prior workforce studies comparing private and public sectors have been mixed. Non-profit organizations may differ from for-profit business/corporate units in terms of goals, operations, and performance evaluations. However, there is some evidence that at least some variables predicting turnover intention do not operate differently in the public sector (e.g., Thatcher et al., 2002). While we recommend that care be taken when generalizing our results to the private sector, we emphasize the value of these public sector results. The public sector, especially government agencies, represents an important and growing component of the IT workforce. Third, our sample for this study differs from the IT labor market in terms of age, gender, education and IT experience. Therefore, care must be taken when attempting to generalize this research and additional studies are required. Also, because the present study used cross-sectional data, we cannot confirm the directions of causality implied in the model. Although the directions of the individual relationships proposed in the model have been supported by prior studies, future research using longitudinal data is encouraged. Another limitation is that the present study did not investigate the interrelations between embeddedness, organizational commitment, and turnover intention Organizational commitment is often an important variable included in traditional turnover studies. While organizational commitment has some similarities with aspects of embeddedness in the organizational fit subconstruct, Mitchell et al. (2001) state that it is conceptually distinct because organizational commitment is affect-driven whereas organizational fit represents a cognitive belief. They also argue that while the normative portion of the organizational commitment construct stems from a sense of obligation, embeddedness has items that are not


focused on this issue. Nevertheless, future researchers may want to include organizational commitment as a variable.

Conclusions

Our research contributes to the IT discipline by applying the embeddedness theory to IT workers and extending it to account for contextual influences. Because of the rapidly changing technological environment, IT professionals are continually faced with acquiring updated skill sets to a greater extent than most other professions. Therefore, IT workers are susceptible to professional obsolescence. Our work shows that professional obsolescence, in turn, impacts their job embeddedness within the organization and, in turn, their intent to leave the organization.

References

Ang, S., and Slaughter, S. The missing context of information technology personnel: a review and future directions for research, in Framing the domains of IT management: projecting the future from the past, R. W. Zmud (ed.), Cincinnati, OH: Pinnaflex Educational Resources, 2000.

Armstrong, J.S., and Overton, T.S. (1977), "Estimating Nonresponse Bias in Mail Surveys", Journal of Marketing Research, 14, 396–402. Bagozzi, R.P. Structural equation models in

marketing research basic principles, In Principles of Marketing Research, R. Bagozzi, ed. Oxford: Blackwell, 1994, 317-385.

Banel, A.P, and Sicherman, N. Technological change and the skill acquisition of young workers, Journal of Labor Economics, 16, 4 (1998), 718-755.

Barclay, D., Higgins, C. and Thompson, R. The partial least square approach to causal modeling, personal computer adoption and use as an illustration, Technology Studies 2, 2 (1995), 285-309.

Becker, G.S. Human Capital. Chicago: University Of Chicago Press, 1993.

Bollen, K. Structural Equations With Latent Variables, New York: John Wiley & Sons, 1989.

Bretz, R.D. and Judge, T.A. person-organization fit and the theory of work adjustment - implications for satisfaction, tenure, and career success, Journal of Vocational Behavior, 44, 1 (1994), 32-54.

Business Communications Review. Demand still strong for skilled IT workers. 37, 9, (2007), 6.

Cammann, C., Fichman, M., Jenkins, Klesh, J. Michigan organization assessment questionnaire: Internal work motivation, in Cook, J.D, Hepworth S.J.,Wall T.D., and Warr, P.B. Experience of

Work: A Compendium and Review of 249 Measures and Their Use, New York: Academic Press, 1981, p.126.

Campbell, D.T., and Fiske, D.W. Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, (1959), 81-105.

Chien, C., and Chen, L. Data mining to improve personnel selection and enhance human capital: a case study in high-technology industry. Expert Systems with Applications 34, 1 (2008), 280-290.

Chin, W.W. The partial least squares approach to structural equation modeling. In G. A. Marcoulides (ed.), Modern Methods for Research, Mahwah, NJ: Lawrence Erlbaum, 1998, pp. 295-336.

Chin, W.W., Marcolin, B.L., and Newsted, P.R. A partial least squares latent variable modeling approach for measuring interaction effects: Results from a Monte Carlo simulation study and an electronic mail emotion/adoption study. Information Systems Research, 14, 2 (2003), 189-217.

Cohen, J. Statistical power analysis for the behavioral sciences, Hillsdale, N.J.: L. Erlbaum Associates, 1988.

Cohen, A. An examination of the relationships between work commitment and nonwork domains. Human Relations, 48, (1995), 239-263.

Cohen, J.D. (2006) The aging nursing workforce: how to retain experienced nurses, Journal of Healthcare Management, 51, 4 (2006), 233-245.

Cole, T. Update skills to cut staff turnover. Computer Weekly, 11/16/2004, 32.

Collett, S. Hot IT skills for 2011, Computerworld, 44, 17 (2011), 34-37.

Coombs, C.R. Improving retention strategies for IT professionals working in the public sector. Information & Management, 46, 4, (2009), 233-240.

Cotton, J.L and Tuttle, J.M. employee turnover: a meta-analysis and review with implications for research. Academy of Management Review, 11, (1986), 55-70.

Crossley, C., Bennett, R.J., Jex, S.M., and Burnfield, J.L. Development of a global measure of job embeddedness and integration into a traditional model of voluntary turnover. Journal of Applied Psychology, 92, 4 (2007), 1031-1042.

Dess, G.G., and Shaw, J.D. Voluntary turnover, social capital, and organizational performance. Academy of Management Review 26, (2001), 496-524.

Diamantopoulos, A., and Winklhofer, H.M. Index construction with formative indicators: an alternative to scale development. Journal of Marketing Research, 38, 2 (2001), 269-277.


Falk, R.F., and Miller, N.B. A Primer for Soft Modeling. Akron, OH: The University of Akron Press, 1992.

Felps, W., Mitchell, T.R., Hekman, D.R., Lee, T.W., Holtom, B.C. and Harman, W.S. Turnover contagion: how coworkers' job embeddedness and job search behaviors influence quitting, Academy of Management Journal, 52, 3 (2009), 545-561.

Ferratt, T.W., Agarwal, R., Brown, C.V., Moore, J. IT human resource management configurations and it turnover: theoretical synthesis and empirical analysis. Information Systems Research, 16, 3 (2005), 237-255.

Figueroa, E.B., and Woods, R.A. Industry output and employment projections to 2016, Monthly Labor Review (November 2007), 53-85.

Fornell, C., and Larcker, D.F. Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18, 1 (1981), 39-50.

Gallivan, M. J. “Examining IT Professionals’ Adaptation to Technological Change: The

Influence of Gender and Personal Attributes,” The DATA BASE for Advances in Information Systems, 35, 3 (2004), 28-49.

Griffeth, R.W., Hom, P.W., and Gaertner, S. A meta-analysis of antecedents and correlates of employee turnover: update, moderator tests, and research implications for the next millennium. Journal of Management 26, 3, (2000), 463-488.

Halbesleben, J.R.B. and Wheeler, A.R. The relative roles of engagement and embeddedness in predicting job performance and intention to leave. Work & Stress 22, 3 (2008), 242-256.

Hall, T. Jagielska, D., and Baddoo , N. Motivating developer performance to improve project outcomes in a high maturity organization, Software Quality Journal, 15, 4 (2007), 365-383.

Harman, H. H. (1976).Modern Factor Analysis, 3ed. Chicago: University Press. Harman, W.S., Lee, T.W., Mitchell, T.R., Felps, W.,

and Owens, B.P. The psychology of voluntary employee turnover. Current Directions in Psychological Science, 16, 1 (2007), 51-54.

Henseler, J., Ringle, C.M. and Sinkovics, R.R. The use of partial least squares path modeling in international marketing. Advances in International Marketing, 20, (2009), 277–319.

Holtom, B.C., and Inderrieden, E. J. Integrating the unfolding model and job embeddedness model to better understand voluntary turnover. Journal of Managerial Issues, 18, 4 (2006), 435-452.

Holtom, B. C., Mitchell, T. R., and Lee, T. W. Increasing human and social capital by applying job embeddedness theory. Organizational Dynamics 35, 4 (2006), 316-331.

Hom, P.W., Tsui, A.S., Lee, T.W., Ping Ping F., Wu, J.B., Zhang, A.Y., and Li, L. Explaining Employment Relationships With Social Exchange and Job Embeddedness. Journal of Applied Psychology, 94, 2 (2009), 277-297.

Igbaria, M., and Guimaraes, T. Exploring differences in employee turnover intentions and its determinants among telecommuters and non-telecommuters. Journal of Management Information Systems, 16, 1 (1999), 147-164.

Jackson, S.E. and Turner, J.A. Correlates of burnout among public service lawyers. Journal of Occupational Behavior, 8, 4 (1987), 339-350.

Jarvis, C B., Mackenzie, S.B., Podsakoff, P.M., Mick, D.G., and Bearden, W.O. Critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research, 30, 2, (2003), 199-218.

Jöreskog, K.G. and Sörbom, D. LISREL 8: Structural equation modeling with the SIMPLIS command language. Chicago, IL: Scientific Software International (1996).

Joseph, D., and Ang, S. The threat-rigidity model of professional obsolescence and its impact on occupational mobility behaviors of IT professionals. International Conference on Information Systems Proceedings, (2001), http://aisel.aisnet.org/icis2001/73.

Joseph, D., Koh, C.S.K., and Hao Foo, A.C. Sustainable IT-specific human capital: coping with the threat of professional obsolescence. International Conference on Information Systems Proceedings, (2010), http://aisel.aisnet.org/icis2010_submissions/46.

Joseph, D., Ng, K., Koh, C., and Ang, S. Turnover of information technology professionals: a narrative review, meta-analytic structural equation modeling, and model development, MIS Quarterly, 31, 3 (2007), 547-577.

Kaplan, D.M. and Lerouge, C. Managing on the edge of change: Human resource management of information technology employees, Human Resource Management, 46, 3 (2006), 325-330.

Kenny, D.A., Kashy, D.A. and Bolger, N. The Handbook of Social Psychology, Fourth Edition, in Gilbert, D.T., Fiske, S.T., Lindzey, G. and Allport, G.W., Oxford: University Press, 1985.

Kleinbaum, D.G., Kupper, L., and Muller, K.E. Applied Regression Analysis and Other Multivariable Methods, 2nd ed., Boston: PWS-Kent, 1988.

Kopelman, R.E. 'Psychological stages of careers in engineers: An expectancy theory taxonomy. Journal of Vocational Behavior, 10, (1977). 270-286.


Krijnen, W.P., Dijkstra, T.K., and Gill, R.D. Conditions for factor (in) determinacy in Factor Analysis. Psychometrika, 63, 4 (1998), 359–367.

Lee, K. and Mirchandani, D, Dynamics of the importance of IS/IT skills The Journal of Computer Information Systems, 50, 4 (2010), 67-78.

Lee, T.W., Mitchell, T.R., Sablynski, C.J., Burton, J.P., and Holtom, B.C. The effects of job embeddedness on organizational citizenship, job performance, volitional absences, and voluntary turnover. Academy of Management Journal, 47, 5 (2004), 711-722.

Lee, T.W. and Mowday, R.T. Voluntarily leaving an organization: an empirical investigation of Steers and Mowday's model of turnover, Academy of Management Journal, 30, (1987), 721-743.

MacCallum, R.C., and Browne, M.W. The use of causal indicators in covariance structure models: Some practical issues. Psychological Bulletin 114, 3 (1993), 533-541.

Mallol, C.M., Holtom, B.C., and Lee, T.W. Job embeddedness in a culturally diverse environment. Journal of Business & Psychology, 22, 1 (2007), 35-44.

March, J. G., and Simon, H. A. Organizations. New York: Wiley, 1958.

Mathieson, K., Peacock, E., and Chin, W.W. Extending the technology acceptance model: The influence of perceived user resources, The Data Base for Advances in Information Systems, 32, 3 (2001), 86-112.

Mint, Indian IT firms redefine career path for engineers. (2010, February 15).Retrieved July 17, 2011, from ABI/INFORM Global. (Document ID: 1989846951).

Mitchell, T.R., Holtom, B.C., Lee, T.W., Sablynski, C.J., and Erez, M. Why people stay:

using organizational embeddedness to predict voluntary turnover, Academy of Management Journal, 44, 6 (2001), 1102-1122.

Moore, J.E. One road to turnover: an examination of work exhaustion in technology professionals. MIS Quarterly, 24, 1 (2000), 141-167.

Ng, T.W.H. and Feldman, D.C. Re-examining the relationship between age and voluntary turnover, Journal of Vocational Behavior, 74, 3 (2009), 283-294.

Ng, T.W.H. and Feldman, D.C The relationships of age with job attitudes: a meta-analysis, Personnel Psychology, 63, 3 (2010), 677-718.

Nunnally, J.C., and Bernstein, I.H. Psychometric Theory, 3rd ed., New York: McGraw-Hill, 1994.

O'Connell, M. and Kung, M.C. The cost of employee turnover. Industrial Management 49, 1 (2007), 14-19.

Perelman, D. 10 ways to jump an IT pay grade eWeek, 25, 19 (2008), 41-42.

Petter, S., Straub, D. and Rai, A. Specifying Formative Constructs in Information Systems Research. MIS Quarterly 31, 4 (2007), 623-656.

Podsakoff, P., MacKenzie, S. Lee, J. Common method biases in behavioral research: a critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 5 (2003), 879–903.

Reich, B.H., and Kaarst-Brown, M.L. “Seeding the Line": Understanding the transition from IT to non-IT careers, MIS Quarterly, 23, 3 (1999), 337-364.

Reich, B.H., and Kaarst-Brown, M.L Creating social and intellectual capital through it career transitions, Journal of Strategic Information Systems, 12, 2 (2003), 91-109.

Roberts, N. and Thatcher, J. B. Conceptualizing and Testing Formative Constructs: Tutorial and Annotated Example. The DATA BASE for Advances in Information Systems 40, 3 (2009), 9-39.

Rutner, P.S., Hardgrave, B.C, and McKnight, D.H. Emotional dissonance and the information technology professional, MIS Quarterly, 32, 3 (2008), 635-652.

Ryan, S.D. A model of the motivation for IT retraining, Information Resources Management Journal 12, 4 (1999), 24-32.

Ryan, S.D., and Harrison, D.A. Motivation to retrain: an empirical study. Proceedings of the Seventh Americas Conference on Information Systems, (2001), 1947-1949.

Standridge, J. and Autrey, R. Rapid skill obsolescence in an IT company: A case study of Acxiom corporation. Journal of Organizational Excellence, 20, 3 (2001), 3-9.

Sekiguchi, T., Burton, J.P., and Sablynski, C.J. The role of job embeddedness on employee performance: the interactive effects with leader–member exchange and organization-based self-esteem. Personnel Psychology, 61, (2008), 761-792.

Settoon, R.P., Bennett, N., and Liden, R.C. (1996). Social exchange in organizations: Perceived organizational support, leader-member exchange, and employee reciprocity. Journal of Applied Psychology, 81, 3 (1996), 219–227.

Shaw, J., Delery, J.E., Jenkins, D.G., and Gupta, N. An organization-level analysis of voluntary and involuntary turnover. Academy of Management Journal, 41, 5, (1998) 511-525.

Slaughter, S. and Ang, S. Employment outsourcing in information systems. Communications of the ACM, 39, 7 (1996), 47-54.


Staw, B.M., and Hong, H. Sunk costs in the NBA: why draft order affects playing time and survival in professional basketball. Administrative Science Quarterly, 40, 3, (1995), 474-494.

Standridge, J., and Autrey, R. Rapid skill obsolescence in an IT company: A case study of Acxiom corporation. Journal of Organizational Excellence, 20, 3(2001), 3-9.

Steiner. D.D. and Farr. J.L. Career goals, organizational reward systems and technical updating in engineers. Journal of Occupational Psychology 59, (1986), 13-24.

Thatcher, J.B., Liu, Y., Stepina, L.P., Goodman, J.M., and Treadway, D.C. IT worker turnover: an empirical examination of intrinsic motivation. The Data Base for Advances in Information Systems, 37, 2/3 (2006), 133-146.

Thatcher, J.B., Stepina, L.P. and Boyle, R.J. Turnover of information technology workers: examining empirically the influence of attitudes, job characteristics, and external markets. Journal of Management Information Systems 19, 3, (2002/2003), 231-260.

Tsai, H., Compeau, D., Haggerty, N. Of races to run and battles to be won: technical skill updating, stress, and coping of IT professionals. Human Resource Management, 46, 3 (2007), 395-409.

Wold, H. O. Partial least squares. In: S. Kotz and N. L. Johnson (Eds), Encyclopedia of Statistical Sciences (Vol. 6, pp. 581–591). (1985), New York, NY: Wiley.

About the Authors Xiaoni Zhang is an Associate Professor of Business Informatics. She received her Ph.D. from University of North Texas in 2001. Her teaching specialties are in SAP business intelligence, systems analysis and design, ecommerce, research methods and statistics. Her research interests include technology adoption in healthcare, digital government, business intelligence, information security, and IT service quality. Her publications appear in Communications of the ACM, International Conference on Information Systems, IEEE Transactions on Engineering Management, European Journal of Information Systems, Information & Management, and other journals. Sherry D. Ryan is an Associate Professor of Information Technology and Decision Sciences at the University of North Texas. She received her Ph.D. in IS from the University of Texas at Arlington and an

MBA from the University of Southern California. Prior to returning to academia she worked for IBM, teaching courses and speaking at national conferences. Her research interests include IT human resource issues, knowledge management, and social media. Her work has appeared in journals including Journal of Management Information Systems, Decision Support Systems, and Informing Science: the International Journal of an Emerging Transdiscipline. Victor R. Prybutok is a Regents Professor of Decision Sciences in the Information Technology and Decision Sciences Department and Associate Dean of the Toulouse Graduate School at the University of North Texas. He received, from Drexel University, his B.S. with High Honors in 1974, a M.S. in Bio-Mathematics in 1976, a M.S. in Environmental Health in 1980, and a Ph.D. in Environmental Analysis and Applied Statistics in 1984. Dr. Prybutok is an American Society for Quality certified quality engineer, certified quality auditor, certified manager of quality / organizational excellence, and an accredited professional statistician (PSTAT®) by the American Statistical Association. Dr. Prybutok has authored over 125 journal articles, several book chapters, and more than 130 conference presentations in information systems measurement, quality control, risk assessment, and applied statistics. Leon A. Kappelman, Ph.D. is Professor of Information Systems and Director Emeritus of the IS Research Center in the Information Technology & Decision Sciences Department of the College of Business at the University of North Texas, where he is also a Fellow of the Texas Center for Digital Knowledge. Dr. Kappelman assists many public and private organizations improve their technology management, strategic planning, governance, software development, project management, and IT workforce activities. He has presented and written articles on these and other IT management topics, and testified before the US Congress on technology legislation and IT management practices. Professor Kappelman has published several books, over 100 journal articles, and has lectured and conducted seminars and workshops around the world on many management, business, and technology topics. Dr. Kappelman is founding chair of the Society for Information Management’s (SIM) Enterprise Architecture Working Group and brought nearly $2.5 million in research contracts to the university.


Perceived obsolescence, organizational embeddedness, and turnover of it workers

Documents