RESEARCH ARTICLE T C P L ( M ) T : HE AREER ATHS ESS OR ORE RAVELED A S A IT C H , EQUENCE NALYSIS OF AREER ISTORIES M P , C S 1 OBILITY ATTERNS AND AREER UCCESS Damien Joseph, Wai Fong Boh, Soon Ang Nanyang Business School, Nanyang Technological University, Nanyang Avenue, SINGAPORE 639798 {[email protected]} {[email protected]} {[email protected]} Sandra A. Slaughter College of Management, Georgia Institute of Technology, Atlanta, GA 30303-0520 U.S.A. {[email protected]} This paper examines the objective career histories, mobility patterns, and career success of 500 individuals, drawn from the National Longitudinal Survey of Youth (NLSY79), who had worked in the information technology workforce. Sequence analysis of career histories shows that careers of the IT workforce are more diverse than the traditional view of a dual IT career path (technical versus managerial). This study reveals a new career typology comprising three broad, distinct paths: IT careers; professional labor market (PLM) careers; and secondary labor market (SLM) careers. Of the 500 individuals in the IT workforce, 173 individuals pursued IT careers while the remaining 327 individuals left IT for other high-status non-IT professional jobs in PLM or lower-status, non-IT jobs in SLM careers. Findings of this study contribute to refining the concept of “boundaryless” careers. By tracing the diverse trajectories of career mobility, we enrich our understanding of how individuals construct boundaryless careers that span not only organizational but also occupational boundaries. Career success did not differ in terms of average pay for individuals in IT and PLM careers. By contrast, individuals in SLM careers attained the lowest pay. We conclude this study with implications for future research and for the management of IT professionals’ careers. Keywords: Management of IT human resources; longitudinal, careers, sequence analysis, IT profession boundaryless, mobility 1 Two roads diverged in a wood, and I— Traditionally, researchers have portrayed IT careers as com- I took the one less traveled by, posed of two distinct paths: a technical IT career or a manag- And that has made all the difference erial IT career (e.g. Chesebrough and Davis 1983; Kaiser Robert Frost (1916, The Road Not Taken) 1983; Zabusky and Barley 1996). Individuals in technical IT careers are characterized as holding deep commitment to the technical orientation of the IT profession (Ginzberg and Baroudi 1988) and choosing to remain in technical IT jobs for the duration of their careers (Kaiser 1983). Such technical IT professionals work in multiple organizations over the course 1Mike Gallivan was the accepting senior editor for this paper. Tom Ferratt of their technical IT careers (Kaiser 1983). Technical IT professionals typically measure career success in terms of served as the associate editor. MIS Quarterly Vol. 36 No. 2 pp. 427-452/June 2012 427 Joseph et al./IT Career Histories, Mobility Patterns, and Success accumulated technical accomplishments, expertise, and the organizations within their subscribed careers, (3) differentiate reputation for skill (Ginzberg and Baroudi 1988; Zabusky and the profiles of individuals associated with a given career, and Barley 1996). (4) assess the objective career success associated with the identified contemporary careers. By addressing these issues, Individuals in managerial IT careers, on the other hand, are this paper espouses a more nuanced understanding of contem- characterized as holding to a managerial orientation (Ginzberg porary careers in the IT workforce compared to those and Baroudi 1988), moving from technical IT jobs to manag- presented in current IT literature. erial IT jobs, usually within an organization. Managerial IT professionals typically measure career success in terms of Figure 1 presents the overall heuristic model that guides the promotions attained within an organization (Zabusky and theoretical foundations of this study, the data analysis, and the Barley 1996). The assumption underlying both IT careers discussion of the results. In the following sections, this study (technical IT and managerial IT) is that individuals in these first provides a discussion of the theoretical foundations of careers in terms of the career paths and career mobility. This careers will continue to remain within the IT profession for study then expounds on the antecedents (i.e., individual the duration of their careers (Karahanna and Watson 2006). profiles) and consequences (i.e., objective career success) associated with careers in the IT workforce. Specifically, we In a closer reading of the IT careers literature, Ginzberg and discuss the profiles of individuals (i.e., the overall set of Baroudi (1988) noted that the empirical evidence indicates personal attributes that are argued to characterize individuals that individuals’ careers in the IT workforce are comprised of in career paths). We then focus on understanding objective multiple organizations and occupations. This description of career success, in terms of pay as an external and observable careers as comprising multiple organizations and occupations indicator of accomplishment (Feldman and Ng 2007). is consistent with the concept of a “boundaryless” career (Greenhaus et al. 2008). A boundaryless career is composed of a sequence of jobs that go beyond the boundaries of a single employment setting (DeFillippi and Arthur 1996, p. Characterizing Careers in the 116). The boundaryless career is often contrasted with the IT Workforce traditional career of an “organization man” that unfolds within a single firm (Whyte 1956). This traditional career is typified A career is defined as ‘‘a sequence or combination of occu- by individuals climbing a career ladder comprising of jobs pational positions held during the course of a lifetime’’ (Super with increasing responsibility and status within an organi- 1957, p. 286). The traditional career model prescribes that zation (Hall 1976; Rosenbaum 1979). careers progress in an orderly fashion and in sequential stages within an occupation and organization (Levinson 1978; Super Following Ginzberg and Baroudi’s (1988) review of IT 1957). Yet, scholars question the applicability of this tradi- careers and reconceptualization of careers as boundaryless tional career model in a work environment where careers are (DeFillippi and Arthur 1996), we propose that traditional por- increasingly shaped by multiple employers (Arthur and Rous- trayals of IT careers as having only two homogenous career seau 1996; Sullivan 1999) and occupational communities paths are unduly restrictive. IT careers have been buffeted by (Ginzberg and Baroudi 1988; Van Maanen and Barley 1984). drastic changes in the demand and supply of IT labor since Consequently, scholars have concluded that the traditional the 1970s (Ang and Slaughter 2000). In the 21st century, career model is an exception rather than the norm (Arthur and rapid technological advancements coupled with market Rousseau 1996; Sullivan 1999), labeling nontraditional pressures arising from globalization have forced organizations careers as “boundaryless careers” and “protean careers” to become leaner and less able to offer job security and life- (Arthur 1994; Hall and Mirvis 1995). long employment (Ang and Slaughter 2000; Slaughter and Ang 1996). In fact, research (e.g., Ituma and Simpson 2009) Research describes boundaryless careers and protean careers has begun to notice that contemporary IT careers are indeed as “overlapping but distinct” concepts (Briscoe and Hall boundaryless in nature—with IT professionals moving across 2006a, p. 6). A protean career is self-directed, proactively organizations, not bound to a single employer (Josefek and managed, and driven by personally meaningful values and Kauffman 2003; O'Mahony and Bechky 2006), or moving goals (Briscoe and Hall 2006a). It is depicted by psycho- across occupations, not bound to the IT profession (Reich and logical mobility, referring to one’s perceived capacity to enact Kaarst-Brown 1999, 2003). job changes (Sullivan and Arthur 2006). Psychological mobility, its antecedents and outcomes are interpreted from The goals of this study, therefore, are to (1) identify and the career actor’s perspective, “who may perceive a boundary- describe the contemporary careers of individuals in the IT less future regardless of structural constraints” (Arthur and workforce, (2) examine their mobility across occupations and Rousseau 1996, p. 6). 428 MIS Quarterly Vol. 36 No. 2/June 2012 Joseph et al./IT Career Histories, Mobility Patterns, and Success Characterizing Careers Individual in the IT Workforce by: Objective Career Profiles - Career Paths Success - Career Mobility Figure 1. The Heuristic Model Guiding This Study In comparison, boundaryless careers are described in terms of characterize the careers in terms of (1) the types of career interorganizational mobility; that is, job changes “across the paths and (2) the types and timing of career mobility within a boundaries of separate employers” (Greenhaus et al. 2008, p. career path. By doing so, we obtain a comprehensive under- 280). It is depicted by both psychological and physical standing of boundaryless careers in the IT workforce. boundary-crossing (Greenhaus et al. 2008). Physical mobility refers to actual job changes across structural or institutional boundaries such as jobs, firms, or occupations (Briscoe and Career Paths Hall 2006b; Sullivan and Arthur 2006). In essence, physical mobility is conceptualized as an objective career change while Career paths are models or prototypes characterizing the psychological mobility is conceptualized as a subjective career sequences of a group of individuals. A career sequence career change (Sullivan and Arthur 2006). depicts the succession of occupational jobs within an indi- vidual’s work history. In their seminal paper, Abbott and IT research has predominantly examined the psychological Hrycak (1990) analyzed the career sequences of German aspects of IT careers (since Ginzberg and Baroudi 1988). musicians in the 17th to 19th centuries. They described four Consequently, research on the physical boundary-crossing major career paths (i.e., organists, court musicians, church behaviors in IT professionals’ careers remains scarce (Ang musicians, and others) as musicians advanced from jobs lower and Slaughter 2000). A recent review of the broader bound- in the hierarchy of the musical establishment to that of the music director or Kapellmeister. Blair-Loy (1999) added to aryless careers literature also calls for research examining the this stream of research by examining the career sequences of physical components of boundaryless careers, its antecedents, women in the finance occupation. She identified two broad and their consequences because psychological mobility may career paths: one characterized by orderly advancement up not be directly correlated with the physical, boundaryless corporate career ladders within firms, and the other charac- notions of career mobility (Greenhaus et al. 2008). terized by “disorderly career shifts between disparate fields and among several different organizations” (p. 1362). In Responding to these calls, this paper investigates the physical, another study of women in the workforce, Huang et al. (2007) boundaryless career mobility of individuals in the IT work- and Huang and Sverke (2007) examined the career sequences force. As highlighted by Greenhaus et al. (2008, p. 289), it is of a cohort of Swedish women to identify two broad career critical for researchers to “clarify the meaning and measure- paths: one conforming to traditional notions of upward ment of boundarylessness” by identifying different forms of mobility within an organization, and the other marked by exits nontraditional career mobility. Prior research predominantly from and reentry to the workforce for family reasons. conceptualizes and operationalizes boundaryless careers as interorganizational mobility. This paper takes an initial step Adding to the above stream of research, this study analyses of broadening this conceptualization and operationalization by the career sequences of a cohort of individuals in the IT also considering mobility within and across occupational and workforce. As noted earlier, prior research has identified two organizational boundaries. divergent career paths for IT professionals: a technical IT career and a managerial IT career (e.g., Chesebrough and In broadening the conceptualization and operationalization of Davis 1983; Kaiser 1983). However, given the increased boundaryless careers, research advocates analyzing the types heterogeneity in the IT workforce and the volatility of the IT and timing of career mobility revealed in individuals’ work work environment, it is important to examine anew whether histories (Greenhaus et al. 2008). Accordingly, we utilize the contemporary IT professionals still follow traditional career information available from individuals’ work histories to paths or subscribe to other types of career paths. MIS Quarterly Vol. 36 No. 2/June 2012 429 Joseph et al./IT Career Histories, Mobility Patterns, and Success Career Mobility contrast to organizational initiatives to move IT professionals into line functions, Moore et al. (2001) report on initiatives by As noted above, this study broadens the conceptualization and businesses, government, and universities to ease the entry into operationalization of boundaryless careers by examining the IT jobs from non-IT jobs and to reduce the time to proficiency types and timing of career mobility revealed in work histories. in IT. These initiatives were introduced when the IT industry Based on the established career literature, this section dis- faced an acute shortage of IT professionals. cusses the various types of career mobility that comprise a career sequence and their timing within a career path. Given the possibilities for both organizational and occupa- tional mobility in the IT workforce, this study examines three types of career mobility: (1) movements across organizations Types of Career Mobility within occupations, (2) movements across occupations within organizations, and (3) movements across both occupations Individuals may construct careers by moving across organi- and organizations. By examining these types of career zational or occupational boundaries (Greenhaus et al. 2008) mobility, we gain insights into whether IT careers are indeed or both simultaneously. Examining occupational mobility or boundaryless across organizations as well as occupations. organizational mobility alone presents a limited aspect of one’s career. Organizational mobility has been frequently studied, as evidenced by the extant organizational turnover Timing of Career Mobility literature (e.g., Griffeth et al. 2000; Joseph et al. 2007), but this literature rarely examines individuals’ destinations after In addition, we investigate the timing of different types of they leave their organization (Kirschenbaum and Weisberg career moves within individuals’ career paths. For example, 2002). Consequently, there is a lack of understanding of do individuals move early in their careers or late in their whether those who change jobs across organizations remain careers, or do they exhibit a steady rate of mobility throughout within or leave their occupation. their careers? A longitudinal perspective of careers provides insights into the timing of career moves (Huang and Sverke Within an organization, the predominant focus of research has 2007) which could not otherwise be obtained from cross- been on hierarchical mobility up a career ladder (Rosenfeld sectional approaches (Bailyn 2004). 1992). Such studies typically employ organizationally based tournament models (i.e., reaching a certain organizational Careers are increasingly recognized to be nonlinear, without level by a certain age or within a certain period of organi- clearly delineated career stages and predictable timing (Abbott 2003). Instead, individuals’ careers may exhibit zational tenure; Sullivan 1999) to predict career moves to the different rates of career mobility at different points in time. next higher rung on the corporate career ladder (Rosenbaum Indeed, Abbot and Hrycak (1990) urge researchers to examine 1979). Such research, however, ignores the possibility that the pattern of job changes that form a work history because individuals may develop and progress in their careers via individuals continually plan and structure their careers over lateral moves within an organization. In these instances, time. Accordingly, we add to the careers literature by moving to a new occupation within an organization does not examining the timing of occupational and organizational necessarily reflect a promotion or a loss of a tournament; mobility over the span of individuals’ careers. rather, it provides new opportunities for individual devel- opment (Greenhaus et al. 2008). Kaiser (1983) noted that some IT professionals held relatively higher organizational Antecedents and Consequences attachment and readily moved into functional jobs from IT jobs. She concluded that mobility to functional jobs might be of Career Paths motivated by restricted career opportunities within IT. In this section, we examine the profiles of individuals as Reich and Kaarst-Brown (1999) provided an alternative antecedents characterizing individuals in career paths. We explanation for IT professionals’ mobility to functional jobs. then focus on objective career success as a consequence of They uncovered organizational initiatives that moved IT subscribing to a particular career path. professionals out of IT jobs into functional jobs when job opportunities became available within an organization. Such moves to non-IT jobs are facilitated by the pervasiveness of Individual Profiles IT in an organization and by the firm-specific skills gained by IT professionals when developing information systems for Individual profiles comprise a set of attributes that are best non-IT functional areas (Reich and Kaarst-Brown 2003). In understood by the analysis of overall patterns of attributes 430 MIS Quarterly Vol. 36 No. 2/June 2012 Joseph et al./IT Career Histories, Mobility Patterns, and Success rather than by analysis of specific attributes (Greenwood and capital is transferable across domains (e.g., across jobs, Hinings 1993). In essence, individual profiles present a organizations, industries, and professions) while specific holistic perspective of attributes that describe individuals in human capital is constrained to a particular domain. Invest- different career paths. Prior research has shown that the ments in general human capital facilitate career mobility as individual attributes of gender and human capital investments the accumulated human capital may be productively utilized strongly influence individuals’ career decisions and oppor- in another occupation or organization (Fugate et al. 2004). tunities in the workforce (Feldman and Ng 2007; Griffeth et On the other hand, investments in specific human capital al. 2000; Josefek and Kauffman 2003; Joseph et al. 2007). decrease one’s career mobility as changing an occupation Therefore, we focus on these sets of individual attributes in and/or organization would entail a forfeit of the specific describing individuals in different career paths. human capital that has been accumulated in a different occu- pation and/or organization (Feldman and Ng 2007; Joseph et There is an extensive body of research examining the careers al. 2011). of females. The research suggests that females, compared to males, are more likely to pursue a boundaryless careers for its self-directedness and for its attainment of personally meaning- Objective Career Success ful goals (Mainiero and Sullivan 2005). The research on career paths of females also tends to support the notion that The career paths that individuals follow have implications for females are more likely than males to leave the workforce their objective career success. Prior research provides evi- either temporarily or permanently for family reasons (Blair- dence that career paths conforming to traditional notions of Loy 1999; Huang et al. 2007; Huang and Sverke 2007). upward mobility within an organization are associated with Moreover, the labor market tends to be segmented by gender higher levels of pay (Blair-Loy 1999; Huang et al. 2007; with an overrepresentation of females in secondary labor Huang and Sverke 2007). In this study, we examine one key market jobs such as clerical and production-oriented jobs. dimension of an individual’s objective career success: pay. Secondary labor market jobs are characterized by their Individuals’ career decisions often have a significant impact relatively lower status, limited opportunities for promotion, on their pay, because the returns to individuals’ accumulated autonomy, or the ability to independently bargain for human capital may differ depending on their career decisions favorable pay (Kalleberg and Sorensen 1979; Piore 1975). By (Mithas and Krishnan 2008; Parent 2000), and individuals comparison, males tend to congregate in the technical and may be compensated differently because of the nature of their managerial jobs, which are characterized by their relatively jobs (Combs and Skill 2003; Gerhart and Milkovich 1990). higher status, opportunities for promotion, autonomy, and favorable pay (Rosenfeld 1992; Tomaskovic-Devey and Human capital theory suggests that obtaining a fit between Skaggs 2002). one’s human capital investment and the job requirements generates rewards in terms of higher pay (Ang et al. 2002; Within male dominated occupations, females report experi- Slaughter et al. 2007). For example, if individuals possess encing barriers to career advancement (Goodman et al. 2003; general human capital, moving frequently across domains Stroh et al. 1996) and to mobility within the external labor may not entail a decrease in pay (Mithas and Krishnan 2008). market in search of alternative jobs (Keith and McWilliams On the other hand, if individuals have accumulated 1999). As the IT workforce tends to be male dominated occupation-specific human capital, such as data modeling and (Information Technology Association of America 2005), programming skills specific to the IT profession, leaving the several studies have documented the negative career experi- IT profession will result in a depreciation of their stock of ences of females in the IT workforce. These negative career human capital, which will be reflected as lower pay. In addi- experiences include stereotyping of and hostility toward tion, if individuals have accumulated firm-specific human females (Ahuja 2002), restricted career advancement oppor- capital such as knowledge of an organization’s policies, they tunities (Leventman 2007), and a lack of informal networks may find higher rewards by staying within the organization. to obtain information about alternative employment oppor- tunities (Hartmann et al. 1986; Margolis and Fisher 2002). The nature of the job held may also influence one’s level of pay due to the roles performed, responsibilities held, or power Another key factor influencing individuals’ careers is human wielded (Combs and Skill 2003). For example, Gerhart and capital. Human capital refers to individuals’ productive com- Milkovich (1990) reason that managers are compensated more petencies that result from accumulating knowledge, skills, and than others lower in the organizational hierarchy because experience (Becker 1975). These competencies vary in their managers are perceived to have a greater impact on organi- level of specificity from general to specific. General human zational performance compared to non-managers. Similarly, MIS Quarterly Vol. 36 No. 2/June 2012 431 Joseph et al./IT Career Histories, Mobility Patterns, and Success a study of managerial pay (Carpenter and Wade 2002) shows ques that collect detailed data on a broad range of topics that managers are likely to receive higher pay by virtue of including work history, job duration, and occupation. their responsibilities, experience, and position. As such, organizations tend to pay more for managers compared to For this study, we used the most recent NLSY79 data set those in technical or staff positions (Cappelli and Cascio released (as of 2008) by the BLS containing respondents’ 1991). Accordingly, we compare the pay of individuals in work histories from 1979 to 2006. As the respondents have different career paths as a means of evaluating their objective yet to report retirement and their careers have yet to be com- career success. pleted, our data are right censored. Nevertheless, our data captures a significant portion of respondents’ careers from their start between the ages of 18 and 21 years to the end of this study (as of 2006), between the ages of 42 and 49 years. Method Moreover, this data set is ideal for the analysis of careers as it contains detailed career related information on jobs held and This study adopts an inductive and quantitative approach to job destinations following a career move, thus allowing us to examine the career paths of individuals in the IT workforce. construct career sequences from the beginning of respondents’ Identifying career sequences requires analyses not just at careers. single points in time but across individuals’ work histories (Abbott 2003). Accordingly, we employ an inductive ap- proach to understand the complexities of individuals’ career Sample changes across time and employ quantitative analyses to ad- dress the generalizability of conclusions to the IT workforce. We selected a sample of respondents from the NLSY79 data The analytical approach here differs from the approaches set using the following criteria: (1) respondents be at least 18 employed in prior studies on IT careers (Chesebrough and years and have attained a high school diploma or its equi- Davis 1983; Kaiser 1983; Tanniru 1983) in that a formal, valent; (2) respondents have worked in a full-time IT job for replicable analysis of career sequences is conducted on a at least one continuous year anytime after entering the work- relatively large sample, using optimal matching (a sequence force between 1979 and 2006; and (3) respondents provide at analysis technique) and cluster analyses. least five contiguous years’ worth of workforce data. The U.S. Youth Employment Provisions of the Fair Labor Stan- We also make use of systematic coding strategies to describe dards Act (FLSA) restricts the types of employment for individuals’ career sequences. This study is also atypical in persons under the age of 18. Accordingly, we applied the “18 that we do not test empirical models identifying variables that years and older” criterion to include respondents who were explain or predict the career mobility or the career paths to able to work in any job. We also considered respondents to which individuals will subscribe. Instead, we make use of a begin their careers after attaining a high school diploma (or its series of statistical tests to verify that the conclusions reached equivalent) to avoid capturing transient vacation or part-time are indeed statistically significant. This inductive and quanti- IT jobs. tative approach, we believe, is the most appropriate to examine the complexities and nuances of career sequences in Consistent with the BLS definition of a full-time, permanent a relatively large sample of individuals in the IT workforce. job, incumbents in IT jobs were included in our sample if they had worked “for more than 35 hours a week” for “one con- tinuous year in a job” (see Polivka 1996). This criterion Data ensured the selection of respondents who have regarded IT jobs as a full-time permanent job and not just as a temporary We analyze actual work history data of a sample of respon- job.2 It has been argued that individuals in temporary jobs are dents drawn from the National Longitudinal Survey of Youth unlikely to undergo the same cognitive processes with regard 1979 (NLSY79; see Bureau of Labor Statistics 2008). The to their careers as individuals in full-time. permanent jobs Bureau of Labor Statistics (BLS) began the NLSY79 program (Hulin et al. 1985). in 1979 by surveying a nationally representative sample of 12,686 respondents who were between 14 and 21 years old (as of January 1, 1979) and residing in the United States of America. The NLSY79 continues to track these respondents 2We conducted sensitivity analyses using a criterion of two years in an IT job periodically over time. In each wave of data collection, in selecting respondents for inclusion in our sample. The results from this respondents are surveyed using structured interview techni- analysis are consistent with our original results. 432 MIS Quarterly Vol. 36 No. 2/June 2012 Joseph et al./IT Career Histories, Mobility Patterns, and Success We chose to exclude respondents who provided less than five onto the remaining non-IT job categories (e.g., professional, years of workforce data because we required a sufficiently technical, managerial, clerical, sales, etc.) and their codes (in long career sequence for each respondent to compare the brackets) as follows: non-IT manager (“G”), non-IT profes- similarity of one career sequence adequately with another.3 sional (“P”), technical administration and support (“J”), tech- After applying these criteria, our final sample comprised 500 nician (“T”), sales (“S”), clerical (“C”); craft, production and respondents (or 3.94 percent) out of 12,686 respondents in the service (“O”), military (“Y”), in school (“X”), unemployed data set. (“U”), and out of the workforce (“Z”). We arranged respondents’ careers into sequences of coded Coding jobs from the point at which the individual entered the workforce. The average length of career sequences for the To obtain a career sequence for each respondent in our study individuals in our sample is 22.2 years (SD = 5.8 years), with data set, we followed Abbott and Hrycak’s (1990) approach the maximum and minimum lengths at 28.0 years and 5.0 by expressing each respondent’s work history as a sequence years, respectively.5 For example, consider the following of jobs. We used one calendar year as the basic time interval career sequence for a respondent i: P P P P P P P P I I I I I I in the career sequence. Respondents who switched jobs with- M M. The length of respondent i’s career is 16 years. On in a given calendar year were coded for their dominant job in entering the workforce, respondent i spent eight years in a that year. The dominant job refers to the job that the respon- non-IT professional job, then spent six years in a technical IT dent held for the most amount of time (and for at least 13 job, after which she moved into a managerial IT job for two weeks) in a calendar year (Bureau of Labor Statistics 2008). years. The work history data in the NLSY79 contains detailed information about jobs held by respondents in each year. Jobs Analysis are coded, described, and categorized according to the Occu- pation Classification System (OCS; U.S. Census Bureau We employed Abbott and Hrycak’s (1990) sequence analytic 1971, 2000) which distinguishes between 12 major job cate- approach to identify distinct career paths in our data. This gories (e.g., professional, technical, managerial, clerical, approach comprises two separate analyses: an optimal sales, etc.) and identifies 417 separate subcategories (e.g., matching analysis of career sequences to establish the inter- accountants, bank officers, chemical engineers, etc.). The sequence distances between all pairs of career sequences in OCS was refined in 2000 and 2002 to reflect the proliferation the study data set and a cluster analysis that uses the inter- of jobs and job titles in the U.S. labor market. IT jobs coded sequence distances to group career sequences into meaningful by OCS include “Computer and Information Systems clusters (i.e., career paths). Managers,” “Computer Systems Analysts,” and “Database Administrators.” A full list of IT jobs and their respective OCS codes are presented in Table 1. Optimal Matching Analysis Using these OCS codes, we noted each respondent’s job in Optimal matching analysis (Sankoff and Kruskal 1983) is a each year from 1979 to 2006 using the appropriate coding statistical method that yields a measure of resemblance scheme in force at the time of data collection. For ease of between pairs of career sequences. The technique calculates analyses, we recoded jobs held by respondents into one of 13 similarity scores between pairs of career sequences and broad job categories presented in Table 2. All non- measures the extent to which two career sequences differ by managerial IT jobs codes (in Table 1) were recoded as counting the number of substitutions, insertions, and deletions technical IT jobs (i.e., “I”) while Computer and Information needed to transform one sequence into the other. By allo- Systems Manager jobs were recoded as IT managerial jobs cating costs to substitutions, insertions and deletions, pairs of (i.e., “M”).4 All non-IT job codes in our sample were mapped 3We conducted sensitivity analyses by increasing the contiguous workforce (Managers and Administrators, not classified elsewhere) should be coded as data criterion to “at least 10 contiguous years’ worth of workforce data.” The IT management as well. results from this analysis are consistent with our original results. 5The average length of career sequences for males and females are not 4We made a reasonable assumption that for all individuals identified as IT significantly different (t = -1.53, df = 452.52, p = 0.172). The average length managers in the OCS 2000/2002 codes, all of their prior contiguous jobs with of career sequences for males in our sample is 21.8 years (SD = 6.0 years), Code 220 (Office Manager, not classified elsewhere) and Code 245 and the average length of career sequences for females is 22.6 (SD = 5.5). MIS Quarterly Vol. 36 No. 2/June 2012 433 Joseph et al./IT Career Histories, Mobility Patterns, and Success Table 1. Occupation Classification Codes and their Corresponding Jobs in the IT Profession Jobs 1970 Code 2000 Code 2002 Code Computer and Information Systems Manager 11 110 Computer Systems Analyst/Computer Scientist 004 100 1000 Computer Programmer 003 101 1010 Computer Specialist 005 Computer Software Engineer 102 1020 Computer Support Specialist 104 1040 Database Administrator 106 1060 Network and Computer Systems Administrator 110 1100 Network Systems and Data Communications Analyst 111 1110 Computer Hardware Engineer 140 1400 Source: Bureau of Labor Statistics “NLSY79 Attachment 3: 1970, 1980 & 2000/2002 Census Industrial & Occupation Classification Codes and 1977 Department Of Defense Enlisted Occupational Codes,” 2006. Table 2. Job Codes and Definitions Jobs Code Definition IT I Technical IT job (i.e., systems analyst; programmer; software and hardware engineer; IT support specialist; database, network and systems administrator; network and data commu- nication analyst; computer specialist) M Managerial IT job (i.e., computer and information systems manager) Non-IT C Clerical job (e.g., clerk, secretary, office and administrative support worker) G Non-IT managerial job (e.g., general manager, financial manager, sales manager) J Technical administration and support job (e.g., computer/peripheral equipment operator, data entry clerk; computer, office equipment or automated teller repairer) O Craft, production and service job (e.g., protective service, food and beverage worker, entertainer, janitor, service worker, laborer) P Non-IT professional job (e.g., accountant, human resource executive, financial analyst, legal professional, engineer, psychologist, scientist, healthcare professional, teacher) S Sales job (e.g., retail salesperson, insurance sales agent, advertising sales agent) T Technician (e.g., engineering technician, healthcare technician, chemical technician) Out of U Unemployed (i.e., is actively searching for employment during a job gap) Civilian X Enrolled in School Workforce Y Military service Z Out of workforce (i.e., is not actively searching for work during a job gap) sequences can be compared, and sequences that “cost” the sequences (Wilson et al. 1999). The similarity scores are least to be transformed into another are considered the most rescaled into distance scores by dividing the similarity scores similar (Abbott and Forrest 1986). Once all sequences are by 100 and subtracting from 1. The resulting distance matrix analyzed, the outcome is a matrix detailing how similar each (dimensions: 500 rows by 500 columns) is then analyzed sequence is to all others in a given data set. using cluster analysis to empirically group the various career sequences into clusters (Hair et al. 1995). Based on a sensi- We use the optimal matching algorithm in ClustalG (Wilson tivity analysis of various insertion, deletion and substitution 2002) to calculate the similarity scores for all career cost matrices and on the recommendations of Abbott and his 434 MIS Quarterly Vol. 36 No. 2/June 2012 Joseph et al./IT Career Histories, Mobility Patterns, and Success colleagues (Abbott and Forrest 1986; Abbott and Hrycak Second, we assessed the robustness of the results of the 1990) and of Sabherwal and Robey (1993), we set the hierarchical agglomeration by reanalyzing the distance matrix insertion, deletion and substitution costs to the same value with a nonhierarchical cluster analysis. Specifically, we across all pairs of job codes. The rationale for doing this is employed a K-means clustering algorithm in SPSS and that the cluster analysis algorithm is not sensitive to variations specified that three clusters be formed. The level of agree- in distance matrices derived from different cost matrices.6 ment in the resulting cluster memberships of the hierarchical and nonhierarchical clustering algorithms was very high (Cohen’s kappa = 0.93, p < 0.001), suggesting near perfect agreement (Landis and Koch 1977) across the two clustering Cluster Analysis algorithms. We then conducted two sets of cluster analyses on the distance matrix derived from optimal matching analysis to Results identify distinct career paths in our data. First, a hierarchical agglomerative clustering technique was used to analyze the distance matrix derived from the optimal matching analysis In this section, we report our results according to the heuristic and to group career sequences into “natural groups” (Alden- model that guided our study (Figure 1). The following sec- tions present the career paths identified from the analyses of derfer and Blashfield 1984, p. 16) of career paths in the IT career sequences, and the types and timings of career mobility workforce. Specifically, we employed the two-step cluster from the analyses of movements across organizational and analysis procedure in SPSS to analyze the distance matrix by occupational boundaries. We then report our findings on the specifying the 500 columns to represent the clustering antecedents and consequences of career paths. variables and to determine the number of viable clusters in the data. The two-step clustering algorithm is well suited to analyze large data sets in which there are no predetermined Career Paths numbers of clusters. The sequence analysis of work histories reveals that there are Accordingly, we did not limit the two-step clustering algo- three types of career paths in the IT workforce. Table 3 rithm to produce a specific number of clusters but allowed it presents exemplars of career sequences for each career path. to determine, automatically, up to a maximum of 15 clusters Table 4 provides the descriptive statistics showing the extent (the default in SPSS). The clustering algorithm derives the to which each occupation is represented in each career path. candidate number of clusters by comparing model AIC fit statistics (Akaike’s Information Criterion; Akaike 1974) The first cluster, which we label “Information Technology across different clustering solutions. The cluster solution that (IT) Career,” comprises 173 individuals (34.6 percent of the best fits the data is indicated by the greatest ratio of change in sample). From Table 3, we see that individuals in this career the AIC fit statistic between a cluster solution with n clusters path, on average, enter the IT profession at a relatively early and a two-cluster solution (Bozdogan 1987; Norušis 2008). stage of their careers, after some non-IT work experience (3.9 The AIC fit index indicated that a three cluster solution best years, SD = 4.6 years). These individuals tend to remain fits the data. within the IT profession for the remaining duration of their careers or move into a managerial IT position at a later point in time. 6Although substituting one occupation with another in a career sequence The remaining two clusters include individuals who work in should intuitively incur costs, Monte Carlo studies (Abbott and Hrycak 1990) IT jobs for a significantly smaller proportion of their work find that drastically different substitution costs for occupations all result in histories. The second cluster, labeled “Professional Labor “overwhelming similarities” in cluster memberships (p. 164). As such, Market (PLM) Career,” is composed of 147 individuals (29.4 Abbott and Hrycak conclude that the cluster analysis “seems to behave percent of the sample) who typically enter IT jobs at a rela- robustly with variation in substitution costs” and that “differences in minor analytic decisions are unlikely to drastically change results” (p. 164). Our tively early stage of their careers after some non-IT work sensitivity analyses using different substitution costs results in an intra-class experience (5.0 years, SD = 4.8 years). Most of these indi- correlation of 1.0 (p < 0.001), indicating perfect consistency in the cluster viduals (95.2 percent) exit IT jobs after about two years of IT memberships across different substitution scenarios. For example, we con- work experience (2.3 years, SD = 2.9 years). Individuals in sidered scenarios where a change from an IT job to a non-IT job would cost this career path spend most of their careers in non-IT profes- more than transiting from a non-IT job to another non-IT job. As this and sional (27.4 percent) or non-IT managerial (16.9 percent) other scenarios produced identical cluster memberships, we used the same relative costs as Abbott and Hrycak. jobs. MIS Quarterly Vol. 36 No. 2/June 2012 435 Joseph et al./IT Career Histories, Mobility Patterns, and Success Table 3. Exemplars of Career Sequences of Individuals† Clusters ID Sequence 1 I I I I X X X X I I I I I I I I I I I 2 P I I I I I I I I I I I I I I Information Technology (IT) 3 P P P P I I I I I O O I I I I I Career 4 S I I T P C T I I I I P P P I I I (N=173; 34.6%) 5 C O I I I I I I I I M M M M M M M M 6 O O O O C G G I I I M M M M M M M M M M M 7 C G P I C C C C C C C C C C C C G G P P G P 8 T P P P P I T P Z Z Z Z P P P Professional Labor Market 9 O X C O G P C P G C C C G G G P P I (PLM) Career 10 X I I I S S P S P G G G G G G G S S (N=147; 29.4%) 11 P P P P O P I S S S G G G G G G G G 12 P P P P P P P P I G G G G G G G G G G 13 S O O I I I I I I I I I I I O O O 14 C C O C O O G G O G U U X I I I I I Secondary Labor Market (SLM) 15 C C O C C C J J O J J J J J J J J J J I I I Career 16 C C C C C C C C C G C C C I I I C C (N=180; 36.0%) 17 C U Z S C G C C I O O O G C C C 18 O O O Y Y Y Y Y Y Y Y Y Y Y Y X X X I †See Table 2 for detailed meanings of codes. I: IT; M: IT Manager; C: Clerical; G: Non-IT Manager; J: Technical Admin & Support; O: Craft, Production, and Service; P: Non-IT Professional; S: Sales; T: Technician; U: Unemployed; X: In School; Y: Military Service; Z: Out of Workforce. Table 4. Mean Proportion of Work History for Each Job Within Each Career Sequence SLM Jobs F IT Career PLM Career Career IT Jobs Technical IT I 51.0% 12.8% 15.0% 284.88*** Managerial IT M 7.6% 0.1% 0.0% 46.63*** Non-IT Jobs Non-IT Manager G 2.7% 16.9% 4.1% 82.64*** Non-IT Professional P 6.3% 27.4% 3.6% 142.70*** Clerical C 4.2% 10.0% 20.2% 42.84*** Technical Administration & Support J 2.3% 0.8% 8.4% 21.69*** Craft, Production and Service O 5.1% 6.0% 20.6% 57.74*** Sales S 2.0% 4.5% 3.7% 4.90*** Technicians T 1.2% 3.9% 4.9% 8.06*** Out of Civilian Workforce Unemployed U 0.5% 0.5% 1.1% 3.93*** In School X 14.8% 14.2% 6.7% 27.80*** Military Service Y 1.4% 0.6% 8.8% 22.26*** Out of Workforce Z 1.0% 2.2% 2.9% 3.55* Length of Sequence (years) Mean 21.61 21.82 23.07 3.26* SD 5.93 5.53 5.86 MANOVA: Hotelling’s T = 3.458, F = 64.365***; * p < 0.05, ** p < 0.01, *** p < 0.001 (26, 968) 436 MIS Quarterly Vol. 36 No. 2/June 2012