Ahuja et al.

/Mitigating Turnover Intentions

RESEARCH NOTE

IT ROAD WARRIORS: BALANCING WORK–FAMILY

CONFLICT, JOB AUTONOMY, AND WORK OVERLOAD
1
TO MITIGATE TURNOVER INTENTIONS

By: Manju K. Ahuja D. Harrison McKnight

Operations and Decision Technologies Accounting and Information Systems Department
Kelley School of Business The Eli Broad College of Business
Indiana University Michigan State University
Bloomington, IN 47405-1701 East Lansing, MI 48824
U.S.A. U.S.A.
[email protected] [email protected]

Katherine M. Chudoba Joey F. George

MIS Department MIS Department
College of Business College of Business
Florida State University Florida State University
Tallahassee, FL 32306-1110 Tallahassee, FL 32306-1110
U.S.A. U.S.A.
[email protected] [email protected]

Charles J. Kacmar
Department of Information Systems, Statistics,
and Management Science Abstract
Culverhouse College of Commerce
University of Alabama This study examines the antecedents of turnover intention
Tuscaloosa, AL 35487-0226 among information technology road warriors. Road warriors
U.S.A. are IT professionals who spend most of their workweek away
[email protected] from home at a client site. Building on Moore’s (2000) work
on turnover intention, this article develops and tests a model
that is context-specific to the road warrior situation. The
model highlights the effects of work–family conflict and job
autonomy, factors especially applicable to the road warrior’s
circumstances. Data were gathered from a company in the
computer and software services industry. This study provides
empirical evidence for the effects of work–family conflict,
1
Ritu Agarwal was the accepting senior editor for this paper. Kathryn M. perceived work overload, fairness of rewards, and job
Bartol, Jo Ellen Moore, and Damien Joseph served as reviewers. autonomy on organizational commitment and work exhaus-

MIS Quarterly Vol. 31 No. 1, pp. 1-17/March 2007 1

Ahuja et al./Mitigating Turnover Intentions

tion for road warriors. The results suggest that work–family including how a vendor’s system fits within the business pro-
conflict is a key source of stress among IT road warriors cesses and systems of the client. The RWs’ employer organi-
because they have to juggle family and job duties as they zations are impacted by turnover not only because they lose
work at distant client sites during the week. These findings a consultant who has in the past represented them but also
suggest that the context of the IT worker matters to turnover because it means the client has to train and socialize a new
intention, and that models that are adaptive to the work consultant, straining the vendor-client relationship.
context will more effectively predict and explain turnover
intention. This research examines the antecedents of turnover intention
for the IT RW, responding to a call by Ang and Slaughter
Keywords: Turnover, turnover intention, IT personnel, road (2000) to study IT professionals within the context in which
warrior, organizational commitment, work–family conflict, they work. While turnover models have been tested in many
work overload, autonomy, fairness, work exhaustion types of organizations, contextual factors may affect how
these models work in a given context. Suggesting that
context matters, Hom and Griffeth (1995) concluded their
turnover meta-analysis by stating that “most correlations
Introduction changed across settings or populations” (p. 37). Workplace
attributes have been shown to be important predictors of per-
Scholars have learned a great deal about the antecedents of ceptions about the job and turnover (Oldham and Rotchford
turnover intention in work organizations in which people are 1983). Thus, it is important to study IT RW turnover in a
physically collocated (Cotton and Tuttle 1986; Mobley et al. context-sensitive manner.
1979) and particularly among technical professionals (Igbaria
and Greenhaus 1992; Moore 2000). However, little is known Our contribution is to adapt and empirically test Moore’s
about the causes of turnover intention among those we call (2000) turnover intention model in the RW context. This is
information technology road warriors (RWs). We define IT done by substituting work–family conflict (WFC) for Moore’s
RWs as IT consultants who spend most of their workweek at role stressors and by adding organizational commitment, both
distant client sites (including overnight), representing their of which we believe apply to the unique challenges of IT
employer (Madden 1995). When not at a client site, they RWs. Like Moore, we examine the effects of autonomy,
travel back to a home office. While management consultants, perceived work overload, work exhaustion, and fairness of
salespeople, and others may also be characterized as RWs, rewards, all relevant in the RW context. We suggest the key
our focus is on those who (1) hold an IT position, (2) are from source of stress among RWs is WFC, which we therefore
a primarily IT-based or IT-driven company, and (3) work at substitute for Moore’s role stressors, role conflict and role
the client site for the sole purpose of IT support. ambiguity. WFC is the role tension that occurs as job
demands interfere with the performance of family duties
The number of IT RWs is significant and growing, although (Netemeyer et al. 2004). Juggling family and job duties is
exact figures are not available. Studies suggest that outside more difficult for RWs since they work at distant client sites
consultants represent 12 to 23 percent of all IT staff in during the week. We also examine organizational commit-
organizations today (King 2003), and sizeable numbers of ment since IT RWs spend most of their time with clients
these are RWs. Procurement of complex vendor software instead of with other company personnel, straining ties to
systems (e.g., ERP systems like SAP R/4) requires customi- their employer. We propose the model as a parsimonious and
zation, installation, and support by either specialists from context-appropriate way to study IT RW turnover intention.
vendor firms or outside consultants. That there are such large
numbers of these workers is partly due to the difficulty and The paper proceeds as follows. We begin by briefly dis-
expense of training in-house employees on a myriad of cussing Moore’s model and how it was adapted to IT RWs.
specialized skills and software tools. The trend toward We then present the details of our methodology. Next, the
outsourcing undoubtedly also increases the number of RWs.2 results of the data analysis are presented and discussed. We
conclude with implications for research and practice.
IT RWs are crucial to their own companies as well as to the
clients they serve (Lovett et al. 1997). Turnover harms the
client because much of the RW’s knowledge is client-specific,
Theory Development
IT researchers have studied the causes of turnover since at
2
We are grateful to the senior editor for this insight. least the 1980s (Baroudi 1985; Bartol 1983; Dittrich et al.

2 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

1985), and several recent turnover models have significantly social interaction rewards they would experience if they
increased our understanding of IT turnover. Jiang and Klein worked at headquarters. Pay and reward equity was important
(2002) produced a model predicting turnover indicators from to the RWs we interviewed, especially as they compared their
the discrepancy between employee wants and how the organi- job and career path to those at headquarters. Adopted from
zation satisfies those wants. Other models used job satisfac- Moore’s full-mediation model, then, we offer several
tion and job utility to predict search and quit intentions hypotheses.
(Thatcher and Stepina 2001), and organizational commitment
as a mediator of job satisfaction and other turnover predictors H1: Perceived work overload will positively influence work
(Thatcher et al. 2002). Speier and Venkatesh (2002) found exhaustion among IT road warriors.
that person-technology fit was negatively related to absen-
teeism and turnover. Igbaria and his associates developed H2: Autonomy will negatively influence work exhaustion
influential models of turnover that included such factors as among IT road warriors.
organizational commitment, career and job satisfaction, and
role stressors (e.g., Igbaria and Greenhaus 1992; Igbaria and H3: Fairness of rewards will negatively influence work
Guimaraes 1999). exhaustion among IT road warriors.

We base our model on the influential recent IS turnover H4: Work exhaustion will positively influence turnover
model proposed by Moore (2000). It suggests a key factor intention among IT road warriors.
shaping IT turnover intention is work exhaustion, modeled
first as fully mediating the effects of autonomy, perceived In adapting the Moore model to RWs, we add two constructs
work overload, fairness of rewards, and two stressors, role that we believe are very salient to RWs: WFC and organi-
ambiguity and role conflict. A second, partially mediated zational commitment (Figure 1). The pre-study interviews
model with direct links from the other predictors to turnover suggested that WFC was a very critical factor in the lives of
intention had a better fit than the full mediation model. We RWs and a primary source of stress, making it worthy of
build on Moore’s model because it is influential, recent, and further investigation. They had to balance work and family
has many constructs applicable to the RW context, especially duties in creative ways because of their travel schedule. To
the potential for work overload and work exhaustion. our knowledge, WFC has been studied in other fields (Boles
et al. 1997; Frone et al. 1992), but not in IT.

We predict that WFC will have stronger effects than the

Adaptation of Moore’s Turnover Model stressors Moore used, role ambiguity and role conflict, for a
to IT Road Warriors theoretical reason as well. Netemeyer et al. (2004) studied
salespeople and used identity theory to argue that WFC has to
The research model (Figure 1) was adapted from Moore. do with both work and family role identities of the employee.
After a literature review, we conducted 12 semi-structured Both roles are highly salient. The more salient the identity,
interviews with IT RWs in a large software company to the stronger its effect upon perceptions. Netemeyer et al. sug-
uncover issues important to RW turnover. These interviews gested “that clashes between salient role identities can have
clarified how Moore’s model applies to IT RWs. Several of more pronounced effects than within-role clashes/conflicts”
Moore’s constructs are important in the RW context. First, (2004, p. 51). Conflict involving work and family can cause
work exhaustion and work overload apply to RWs because of serious distress, but role conflict and role ambiguity represent
the long hours RWs often incur at the client site in order to within-role conflicts, which are likely to be less salient than
accomplish their objectives before coming home on the WFC because they pertain only to one’s work role identity.
weekend. Work overload has a strong influence on work Thus, WFC should be a stronger predictor of job outcome
exhaustion (Moore 2000), and this should hold for RWs perceptions than role conflict and role ambiguity. Effects of
because they can burn out when overburdened. Second, WFC on turnover intention have been found to be signifi-
autonomy is important since most IT RW work is done at cantly greater than the effects of role ambiguity, and the
client sites, with few opportunities for corporate-based effects of WFC on job stress were greater than that of role
superiors to observe work directly. Autonomy provides RWs conflict (Netemeyer et al. 2004). The differential effects of
freedom and flexibility to manage their own workloads such WFC will be magnified for RWs since family and work roles
that they do not unduly increase stress or work exhaustion. are difficult to manage because of frequent time away from
Third, fairness of rewards is important to RWs because they home. Hence, we use WFC in lieu of Moore’s stressor vari-
need to feel that the extra travel and offsite work they do will ables as an antecedent of work exhaustion and organizational
be rewarded, especially given the lack of counterbalancing commitment.

MIS Quarterly Vol. 31 No. 1/March 2007 3

Ahuja et al./Mitigating Turnover Intentions

Perceived Work
Overload +

+ H1
H8
−

H6 Job − H2 Work + H4 Turnover

Autonomy Exhaustion Intention
+
− −
H7 H12
H5 H10 H9
+
Work Family −
− H13
Conflict Organizational
Commitment
H3

− H11
Fairness +
of Rewards

Figure 1. Theoretical Model (Adapted from “One Road to Turnover: An Examination of Work
Exhaustion in Technology Professionals,” J. E. Moore, MIS Quarterly (24:1), 2000.

In addition, we incorporate organizational commitment in the effect on turnover intention, so we excluded it. We included
model because RWs interact more frequently with clients than marital status since it may relate to the effects of WFC.
with members of their own organizations and may, in time, Finally, we included promotability (Igbaria and Greenhaus
identify less with their employer, decreasing organizational 1992) in the model because RWs may be more willing to tip
commitment. Some RW interviewees were frustrated at not the balance toward work (and away from family) if they
feeling connected with the company because of physical expect to be promoted in the near future. The RWs we
distance from both coworkers and supervisors. Their feelings interviewed said they often felt out of the loop regarding
of being alone were not cries for more supervision, as they career paths and job opportunities at the headquarters.
valued their autonomy. Rather, RWs wanted to keep in touch
with what their peers were doing and where the organization
was headed in order to better understand how to move Work–Family Conflict as an Antecedent
forward in the company. Not feeling fully connected with the of Work Exhaustion
company could lead to low commitment, which could affect
turnover intention. In fact, organizational commitment has WFC can be a source of occupational stress. In the “elec-
been found to be an influential and consistent predictor of tronic briefcase” age, workers need not be physically located
turnover intention in both IT and other settings (Cotton and at the employer’s site, but can work from anywhere, anytime,
Tuttle 1986; Igbaria and Greenhaus 1992). and can communicate with colleagues electronically. Green-
haus and Beutell (1985, p. 77) define WFC as a “form of
Control variables were also contextualized to fit the RW inter-role conflict in which the role pressures from the work
situation. Like Moore, we controlled for age and organiza- and family domains are mutually incompatible in some
tional tenure, both traditional turnover controls. Moore con- respect.” We adapt this definition to include family, signi-
trolled for negative affectivity but did not find a significant ficant others, and close friends. Models of WFC suggest that

4 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

conflict arises when demands of participation in one domain Job autonomy is inherently higher in jobs that have a signi-
of life are incompatible with demands of participation in ficant IT component in them because IT affords employees
another, and that this conflict can affect the quality of both more opportunities to respond to their task demands through
work and family life (Greenhaus 1988; Greenhaus and Beutell managing schedules and adapting technologies to fit the
1985; Netemeyer et al. 1996). A European survey revealed specific circumstances of their life (Ahuja and Thatcher,
travel/commuting was the single most stressful aspect of a 2005). We suggest that when work is mediated by IT, as it is
worker’s job because it led to family tension (http:// for RWs, it frees employees from rigid schedules or tight
www.mori.com/polls/2001/mitel.shtml). In virtual settings, control systems. The employee can, therefore, balance com-
a blurring of home and work boundaries has been linked with peting demands more effectively.
stress and exhaustion (Salaff 2002).
H7: Job autonomy will negatively influence work–family
H5: Work–family conflict will positively influence work conflict among IT road warriors.
exhaustion among IT road warriors.

Effects of Autonomy on Perceived

Antecedents of Work–Family Conflict Work Overload

Two antecedents of WFC may be especially salient to RWs: Because of the flexibility it provides, autonomy should also
perceived work overload and autonomy. Perceived work have a negative influence on perceived work overload. RWs
overload—the perception that one has too much to do (Leiter can easily become overburdened with many tasks, but auto-
and Schaufeli 1996; Schaufeli et al. 1995)—is associated with nomy allows them to make adjustments to accommodate other
higher levels of WFC (Frone et al. 1997; Parasuraman et al. aspects of their lives as needed, offsetting negative implica-
1996). IT RWs may be especially susceptible to perceived tions of high workload. Empirical examination of the rela-
work overload because of tensions inherent in their boundary tionship between task dimensions like autonomy and stressors
spanner role (Singh et al. 1994), since they must meet the like perceived work overload have produced ambiguous
needs of both employer and client. Although IT RWs have findings (see Aryee et al. 1999; Mannheim and Schiffrin
easy access to information and communication technologies 1984; Singh et al. 1996); however, in a meta-analysis, Lee
(ICT) to facilitate interaction with family and friends at home and Ashforth (1996) demonstrate a relationship between per-
and are frequent ICT users in their role as boundary spanners, ceived work overload and lack of autonomy. Moore also
ICT use is not likely to be a sufficient substitute for face-to- found that autonomy was correlated with perceived work
face interaction with family members (Hinds and Kiesler overload.
1995). Hence, work overload will have a positive influence on
RW work–family conflict. H8: Autonomy will negatively influence perceived work
overload among IT road warriors.
H6: Perceived work overload will positively influence
work–family conflict among IT road warriors.
Organizational Commitment: Effects
Job autonomy is “the degree to which the job provides sub- and Antecedents
stantial freedom, independence and discretion in scheduling
the work and in determining the procedures to be used in Effects on Turnover Intention
carrying it out” (Hackman and Oldham 1975, p. 162). In its
most general form, job autonomy influences employees’ Organizational commitment is the extent to which one is
perceptions of their authority to initiate, perform, and com- involved in, and identifies with, one’s organization (Mowday
plete tasks (Kaldenberg and Becker 1992; Xie and Johns et al. 1982). When employees feel committed to an organiza-
1995). It allows workers to manage WFC in a way that tion, they are likely to stay with the organization (Cotton and
makes sense for them personally, given their personal Tuttle 1986; Igbaria and Greenhaus 1992; Mobley et al.
constraints. Thomas and Ganster (1995) studied healthcare 1979). Both the physical isolation of RWs and their prox-
professionals and found higher autonomy was associated with imity to clients put tension on their organizational commit-
lower levels of WFC. Likewise, the telecommuting literature ment, which could easily erode. Because working continually
suggests flexibility in the timing of work activities—auto- in isolated locations requires high commitment to the firm,
nomy over when work is done—can reduce WFC (Goldstein RW commitment should be a strong predictor of turnover
2003; Pratt 1999). intention.

MIS Quarterly Vol. 31 No. 1/March 2007 5

Ahuja et al./Mitigating Turnover Intentions

H9: Organizational commitment will negatively influence important to IT RWs since it provides them the freedom to
turnover intention among IT road warriors. perform their work independently, reducing frustration from
actions like playing telephone tag to get approval for work
Work Exhaustion activities from remote supervisors who may not understand
the circumstances at a particular site. Research on IT workers
RWs frequently face long hours, excessive travel, and stress who telecommute found positive relationships between
associated with project deadlines (Goff 2001), making them autonomy and organizational commitment (Bailyn 1994;
susceptible to work exhaustion. The literature shows that the Belanger 1999; Hill et al. 1998).
consequences of work exhaustion include reduced organiza-
tional commitment (Lee and Ashforth 1996; Leiter and H12: Job autonomy will positively influence organi-
Maslach 1988; Thomas and Williams 1995). RWs will likely zational commitment among IT road warriors.
decrease their organizational commitment as their work
exhaustion increases, because they will lose faith that the Work–Family Conflict
company can take care of them by providing an acceptable
work life. Regular IT workers experiencing work exhaustion Ability to work from home on days in which family respon-
can socially interact with others at the firm as they com- sibilities require attention can be an important consideration
miserate over the work conditions in a positive manner in evaluating other job opportunities (Pratt 1999), suggesting
(“we’re all in this together”). Such social interaction may a likely relationship between WFC and organizational
mitigate the effects of work exhaustion on organizational commitment. RWs, who are susceptible to WFC issues, may
commitment. With RWs, on the other hand, physical distance decrease their commitment as WFC increases because high
severely restricts face-to-face social interaction, which means WFC indicates the company is placing undue burdens that
that the effects of work exhaustion on organizational cannot be reconciled with family duties.
commitment will not be mitigated.
H13: Work–family conflict will negatively influence
H10: Work exhaustion will negatively influence organi- organizational commitment among IT road warriors.
zational commitment among IT road warriors.

Fairness of Rewards Mediating Role of Work Exhaustion

and Organizational Commitment
Social exchange theory, based on the role of relationships
between employees (Cropanzano, Rupp, and Byrne 2003; The research model depicts work exhaustion and organiza-
Cropanazano et al. 2001), suggests employees are inclined to tional commitment as the proximal antecedents of turnover
form social exchange relationships with others so long as they intention, mediating the effects of all other factors. Because
perceive they are fairly and reciprocally receiving benefits of of their conceptual diversity, using both work exhaustion and
value to them as a result of the social exchange. In turn, organizational commitment as mediators increases the likeli-
social exchange relationships affect organizational commit- hood that the other model antecedents will be fully mediated.
ment (Cropanzano et al. 2001). RWs isolated at clients’ sites For example, if autonomy is not mediated by the effects of
may believe it is more difficult to both detect fairness viola- work exhaustion alone, it may be mediated by the effects of
tions and to correct any that arise, so they will be more organizational commitment. In contrast, Moore tested a
sensitive about fairness issues. For regular IT employees, single mediator, work exhaustion, and found that it only
social interaction rewards might mitigate the effects of reward partially mediated the effects of other variables. The two
fairness on commitment. However, the lack of social inter- mediators represent key, powerful, and complementary rea-
action for RWs means that fairness of rewards will have a sons one might decide to exit a company. If one does not feel
significant influence on organizational commitment. committed to the organization, then one will have less posi-
tive beliefs about the company, its management, and anything
H11: Fairness of rewards will positively influence organi- else one associates with the company, such as its responsive-
zational commitment among IT road warriors.
ness to one’s career desires and needs. This tie-in to global
beliefs and feelings about the company is one reason why
Autonomy organizational commitment has been a powerful turnover
intention antecedent in prior research. Work exhaustion, on
Eby et al. (1999) found that autonomy was positively related the other hand, is about the work itself and reflects salient
to organizational commitment. Autonomy may be especially frustration about job outcomes (Moore 2000), making it a key

6 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

Table 1. Demographics of Sample

Total Responses: 171
Gender Female 54%
Male 46%
Age 22–30 27%
31–40 28%
41–50 33%
> 50 13%
Marital Status Single 41%
Married 59%
Responsible for dependents on a regular basis? Yes 36%
No 64%
Tenure at company < 2 years 34%
2–5 years 48%
> 5 years 18%

indicator of dissatisfaction levels that likely lead to turnover Background data were gathered during six hour-long tele-
intention. The two variables are complementary both because phone interviews with human resource representatives who
one is about the job while the other is about the organization, supported RWs. Next, three researchers visited two client
and because organizational commitment inheres positive sites and conducted semi-structured interviews with twelve
affect while work exhaustion inheres negative affect. They RWs, each lasting 45 to 60 minutes. RWs described positive
should, therefore, fully mediate the effects of the other and negative characteristics of their jobs. They reported on
factors. such topics gleaned from the telephone interviews as project
teams, employee–manager relationships, maintaining a sense
H14: Work exhaustion and organizational commitment of community, and compensation. They completed a series of
will fully mediate the effects of job autonomy, ranking exercises around the theme, “What really frustrates
perceived work overload, work–family conflict, and me at [company name] is….” Interviews were tape-recorded.
fairness of rewards on turnover intention among IT
road warriors. Company management sent an e-mail to the 700 RWs, asking
them to complete a web-based questionnaire. The question-
naire was server-hosted at a researcher’s university. A second
e-mail was sent 10 days after the first, reminding RWs to
Method complete the questionnaire. Of the 700 RWs contacted, 171
completed the instrument for a 24.4 percent response rate.
We studied employees at a company in the computer and soft- Demographics of the sample are shown in Table 1. The
ware services industry, with over 3,000 employees, most respondents ranged from 22 to 50 years in age, with the
residing at company headquarters located in a large Mid- 22–30, 31–40, and 41–50 age-groups represented almost
western U.S. city. Of these, about 700 employees are RWs evenly. Despite this wide range, 82 percent had 5 years or
because they work at client sites to install and maintain the less of company tenure.
information management systems developed at company
headquarters. The company permits RWs to live wherever
they wish and then travel to client sites, typically on a Measures
Monday morning through late Thursday schedule, with
Fridays spent at home completing paperwork. When possible, The questionnaire items, mainly from existing scales, are
RWs are assigned to clients located within the same region of listed in the Appendix, along with their sources. The auto-
the country as their home to minimize travel time. RW nomy measure was adapted from the organization behavior
engagements range from several days to more than a year, and social psychology literature (Beehr 1976). This approach
with most lasting 3 to 6 months. differs from Moore’s (2000) measure in that her items reflect

MIS Quarterly Vol. 31 No. 1/March 2007 7

Ahuja et al./Mitigating Turnover Intentions

input in decision-making, while ours reflect input regarding obtained the following results: Null model (P2 = 1088.21, df
how, when, and what work is done. = 406); measurement model with a single factor (P2 =
4281.84, df = 377, p = .00, RMSEA = .246, NFI = .75, NNFI
WFC was measured using items from Adams et al. (1996). = .76, CFI = .78, AIC = 4398, and standardized RMR = .15);
Turnover intention was measured using a four-item scale multi-trait model: (P2 = 678.77, df = 356, p = .00, RMSEA =
adapted from Moore (2000). Our work exhaustion scale is .073, NFI = .94, NNFI = .97, CFI = .97, AIC = 837, and
taken from Moore, whose items originated from eight items standardized RMR = .058); multi-trait model with a method
developed by Maslach (1982) and Maslach and Jackson factor—a measurement model with items assigned to both
(1984). Of the five items Moore selected, other researchers their respective latent factor (i.e., seven factors) as well as to
have reported reliability problems with the item “Working the method factor (P2 = 524.47, df = 327, p = .00, RMSEA =
with people all day is really a strain for me” (e.g., Boles et al. .059, NFI = .95, NNFI = .97, CFI = .98, AIC = 740, and
2000; Kickul and Posig 2001), so it was removed. standardized RMR = .050).

As expected, model 4, with the addition of the method factor,

improved fit (Williams et al. 1989). P2 difference tests among
Results the models were performed, with the result that model 4 fit
the data better than did model 3 () P2 = 154.30, df = 29). The
Response Bias and Common Method improvement in fit was nominal, however, with both sets of
fit statistics falling in the ranges recommended by Hu and
Variance Testing Bentler (1999). The loadings for the items from model 4 were
then used to compute the AVE (Barclay et al. 1995; Chin
Response bias was assessed on gender, age, education, marital 1998) for each latent construct, including the method factor.
status, number of dependents, tenure with the company, Results indicated that the method factor accounted for 7
tenure in current position, tenure in current project, and per- percent of the variance in the model. Since the AVE's for all
ceived promotability, using the Armstrong and Overton other latent constructs met or exceeded the minimum cutoff
(1977) procedure. The sample was divided into three parts, of 0.50 (Chin 1998), and since the percent variance explained
with early, middle, and late respondents categorized by the by the method factor was less than the critical method factor
date the questionnaire was received. An analysis of variance effect value of 25 percent as recommended by Williams et al.
contrasting the early third of respondents with the late third of (1989), we concluded that common method variance was not
respondents indicated a nonsignificant difference for all a significant contributor to study results.
variables but gender (F(9,89) = 5.04, p = .027). During the
early period, 24 females and 30 males responded while during Partial least squares (PLS) was selected for data analysis,3
the late period 30 females and 15 males responded. using a two-step analytic approach (Anderson and Gerbing
1988). First, the measurement model is evaluated to assess
We used the procedure recommended by Widaman (1985), the validity and reliability of the measures, and once it is
combined with Williams et al. (1989), to test for the effects of accepted, the structural model is evaluated to assess the
common method variance. Following this approach, four strength of the hypothesized links among the variables.
models are estimated: (1) a null measurement model, (2) a
model with all items pointed to a single method factor, (3) a
multifactor trait measurement model with items pointed to the
Measurement Model Evaluation
proposed latent constructs, and (4) a trait measurement model
like model 3 but with an additional method factor. If a
Means, standard deviations, reliability measures (ICR),
method effect of some magnitude exists, model 4 will fit the
average variance extracted (AVE), and correlations for the
data significantly better than model 3. Next, this procedure
variables are shown in Table 2. The lowest ICR in Table 2 is
allows the researcher to determine the amount of variance in
0.87, well above the accepted level of 0.70 (Fornell and
the model contributed by the method factor by computing the Larcker 1981). For convergent validity, PLS requires that
average variance extracted (AVE) for the latent constructs AVE figures be 0.50 or above as an indicator that the items
vis-à-vis the method factor (Chin 1998). The variance within a variable converge (Chin 1998). As Table 2 shows,
explained by the method factor should be less than 25 percent each construct exceeds this requirement.
of total (Williams et al. 1989).
3
PLS was chosen over LISREL because of the complexity of the model to be
Four models were estimated in LISREL 8.71 using a tested (i.e., the number of constructs and links; see Barclay et al. 1995; Chin
covariance matrix of the observed variables as input. We 1998).

8 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

Table 2. Descriptives, Correlations, and Measurement Model Statistics

Std.
Variable Mean Dev. ICR 1 2 3 4 5 6 7 8 9 10 11
1. Work–Family Conflict 5.04 1.37 .95 .79
2. Job Autonomy 5.05 1.12 .87 -.20** .63
3. Work Overload 4.10 1.46 .94 .48** -.31** .79
4. Work Exhaustion 3.59 1.37 .94 .52** -.34** .65** .80
5. Organizational 5.29 1.21 .93 -.32** .55** -.41** -.49** .76
Commitment
6. Fairness of Rewards 4.36 1.30 .95 -.32** .58** -.33** -.40** .60** .80
7. Turnover Intention 2.55 1.35 .95 .22** -.32** .25** .42** -.67** -.38** .82
8. Tenure at Organization 1.80 1.74 1.00 .13 .25** .17* .22** .03 .01 -.04 1.00
9. Age n/a n/a 1.00 .17* -.06 .06 -.05 .09 .02 -.20** .08 1.00
10. Promotability 4.69 1.86 1.00 -.15 .40** -.17* -.27** .51** .49** -.46** -.17* -.15* 1.00
11. Marital Status n/a n/a 1.00 .05 -.01 -.02 -.03 .02 -.01 -.05 .07 .31** -.07 1.00
(1 = Single; 2 = Married)
Notes: 1. Average Variance Extracted (AVE) is on the diagonal.
2. ICR refers to internal composite reliability.
3. Significance of correlations: **p < .01; *p < .05

Discriminant validity was assessed using the Fornell and tional commitment, and work exhaustion. Overall, we found
Larcker (1981) test. First, each latent variable correlation support for the proposed model; specifically, 11 of the 14
should be less than the square root of the AVE on the same hypotheses were supported. As predicted, both work exhaus-
row and column. We apply a more stringent test—that the tion and organizational commitment (H4, H9) were significant
correlation be less than the AVE itself, as Table 2 shows. For turnover intention predictors, although the latter was more
example, the correlation between WFC and job autonomy predictive. From the original Moore hypotheses, perceived
(-0.20) would be compared to the bold diagonal items above work overload had a strong, positive influence on work
it (0.79) and to its right (0.63) to verify that it does not exceed exhaustion (H1), and autonomy negatively influenced work
either diagonal element. Comparing each correlation against exhaustion (H2). Fairness of rewards did not significantly
its corresponding diagonal figure, each correlation is less than impact work exhaustion (H3). Importantly, WFC was
the numbers on the diagonal, indicating that the discriminant positively associated with work exhaustion (H5) and
validity test is met (Chin 1998). The highest correlation was perceived work overload affected WFC (H6), as predicted.
between turnover intention and organizational commitment Surprisingly, autonomy had no effect on WFC (H7), even
(0.67). These constructs passed the stringent version of the though it influenced perceived work overload (H8). Work
Fornell and Larcker test in that their respective AVEs were exhaustion (H10), fairness of rewards (H11), and autonomy
0.82 and 0.76. Given that the measures demonstrated ade- (H12) significantly predicted organizational commitment, but
quate construct validity, the measurement model was WFC did not (H13).
accepted, and we proceeded to test the structural model.
To confirm full mediation by organizational commitment and
work exhaustion (H14), we added direct paths to turnover
Structural Model Results intention from job autonomy, WFC, perceived work overload,
and fairness of rewards and then reran the PLS structural
Figure 2 reports the results of the structural model test. We model. None of the added links were significant in predicting
used a PLS bootstrapping technique with 100 resamples to turnover intention. They had only a minor effect, increasing
assess the significance of model linkages. Control variables the variance in turnover intention explained from 52 percent
were entered as predictors of turnover intention, organiza- to 54 percent. Thus, H14 was supported.

MIS Quarterly Vol. 31 No. 1/March 2007 9

Ahuja et al./Mitigating Turnover Intentions

Perceived Work
Overload
R2 = .10
.43***

.46***
-.31***

Work Turnover
Job -.13* .14*
Exhaustion Intention
Autonomy R2 = .54 R2 = .52
.22***
-.06 -.21***

.26***
-.48***
Work Family
Conflict -.09
Organizational
R2 = .24 Commitment
R2 = .54

-.07
Control Work Organizational Turnover
Fairness .22**
Variable Exhaustion Commitment Intention
of Rewards
Age -.14** .14** -.18**
Tenure at Org. .14*** .06 ns -.08 ns
Marital Status -.02 ns -.01 ns .01 ns
*** p < .001 Promotability -.06 ns .28*** -.22***
** p < .01
* p < .05 Note: Control variables above were entered simultaneously with
model variables

Figure 2. Structural Model

The variables in the Figure 2 model explained 10 percent of to the Figure 2 model but found no significant effect on any
the variance in perceived work overload with one predictor, of the three dependent variables.
24 percent of the variance in WFC with two predictors, and
over 50 percent of the variance in work exhaustion, organiza-
tional commitment, and turnover intention with the predictors
shown. Discussion
Among the control variables, the most pronounced effects The model had reasonable explanatory power, accounting for
were those of promotability on organizational commitment ($ slightly over half of the variance in turnover intention,
= 0.28***) and turnover intention ($ = -0.22***). Age had implying that work exhaustion and organizational commit-
modest but significant effects on organizational commitment, ment are key RW turnover factors. These antecedents of turn-
work exhaustion, and turnover intention, with the older over intention also had slightly over half of their variance
workers being less exhausted, more committed, and less likely explained. The results imply that WFC is an important pre-
to turnover. By contrast, longer company tenure was related dictor of work exhaustion for IT RWs and, through work
to more work exhaustion, indicating the toll the RW role pays exhaustion, affects organizational commitment. Autonomy is
over time. Tenure did not affect commitment or turnover also salient for RWs, positively affecting organizational
intention. Marital status had no effect. To verify that we did commitment and negatively affecting work exhaustion. This
not ignore relevant variables, we next added the control is not a comprehensive model, however, as many other factors
variables of gender and number of dependents (one at a time) such as role conflict and role ambiguity could be used.

10 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

The unique nature of the IT RW provides insight into several sample was representative within the organization. However,
of our results, such as the link from autonomy to perceived it is possible, although we have no evidence for or against it,
work overload. In her study of IT workers, Moore (2000) that those who filled out the questionnaires were those who
found that these two variables correlated at r = -0.20. In this were less work exhausted than average. Another limitation is
study, they correlated at -0.31. Thus, autonomy becomes that the questionnaire was administered at one point in time,
even more important to work overload perceptions for RWs. making it impossible to prove that the model links are causal.
For example, while we show that work exhaustion leads to
While autonomy correlated with WFC in the direction organizational commitment, it is possible that the converse is
hypothesized, it did not significantly influence WFC, contrary true, as Kalliath et al. (1998) found. Also, we adapted pre-
to findings with virtual workers who telecommute (Goldstein vious scales rather than adopting them as-is. This implies that
2003; Pratt 1999). In contrast to telecommuters, IT RWs do our results cannot be compared directly with the results of
not have the option of taking a mid-day break to watch a previous research.
child’s soccer game or attend to family needs. Hence, IT
RWs may not experience reduced WFC resulting from the job
autonomy they experience because they live away from home
Implications for Research
for much of the week. This helps explain why autonomy
affected both work overload and work exhaustion while not
Future work should test this model with other IT RWs in
affecting WFC.
order to extend its generalizability beyond the subject organi-
zation. The extent to which the model’s constructs apply to
The indirect effect of WFC on organizational commitment
implies that WFC creates sufficient energy drain to cause other IT workers should then be tested empirically. We
exhaustion but does not directly make one feel less committed believe that whenever significant work pressures exist that
to the organization. Low fairness of rewards, on the other cause WFC, the variables in this model will capture a large
hand, may make RWs feel bitter toward the organization and share of the variance. The model should also be extended
may reduce commitment directly, but does not cause them to longitudinally to include actual turnover. We note that meta-
feel work exhausted. This finding was corroborated by analyses report that turnover intentions correlate with actual
several interviewed RWs , who mentioned that they did not turnover only from .31 to .36, which means that turnover
feel that superiors at headquarters were necessarily aware of intention should not be used as a surrogate for turnover
their work performance, indicating some distrust in their (Dalton et al. 1999; Hom and Griffeth 1995). Turnover inten-
performance evaluations. tion does not always result in turnover because a number of
other factors enter the equation. Missing factors include
Before moving on to discussing the implications of our geographic or locality preferences, the availability of outside
findings, it is important to acknowledge several limitations. job opportunities (Igbaria and Greenhaus 1992), the expected
First, all data were collected from one organization, with one utility of those opportunities (Mobley et al. 1979), intention
type of IT professional—the RW; hence, the results should to search for another job (Arnold and Feldman 1982), and
not be generalized to other types of IT workers or organi- shock events that produce job-related deliberations (Lee and
zations. Since the study began with a series of interviews that Mitchell 1994). Future research should explore other pre-
enabled us to contextualize the hypotheses to the work dictors that might help determine the point at which an RW
context, it is not surprising that the model applies well to this moves from a state of casual consideration of how long one
group of workers. While the constructs that predicted turn- might stay at the company to a state of active, outside-
over intention antecedents—autonomy, work overload, WFC, company job seeking.
and fairness of rewards—appear to be generic enough that
they may apply to IT RWs in other organizations, future This study suggests that more research should be done to
researchers should empirically test this. Another limitation is ferret out the motivational differences between RWs and tele-
the moderate 24.4 percent response rate on the questionnaire. commuters or other IT workers. For example, while prior
Company management indicated they considered this a rela- research of telecommuters has found a strong negative rela-
tively high rate and were pleasantly surprised. Concurrent tionship between autonomy and WFC (Goldstein 2003; Pratt
with administration of the instrument, RWs were encouraged 1999), this relationship was not found to be significant for
to maximize billable hours, and the questionnaire took about RWs. This and other differences between telecommuters and
30 minutes to complete. Management also confirmed that the road warriors need to be explored further. Thus researchers
demographic characteristics of our sample (see Table 1) must take care to understand the contextual characteristics of
closely matched the demographics of all RWs, suggesting our the work environment to create better models.

MIS Quarterly Vol. 31 No. 1/March 2007 11

Ahuja et al./Mitigating Turnover Intentions

This study extends past research by showing that the prime The result that work exhaustion was a less important turnover
factors of turnover intention (organizational commitment and intention predictor than organizational commitment may be
work exhaustion) have their own small set of causal factors due to the Monday-to-Thursday road schedule practiced by
(work overload, autonomy, fairness of rewards, and WFC), at the subject firm, which may have decreased the salience of
least among IT RWs. This extends the nomological network work exhaustion. If so, it suggests that setting favorable
of turnover model constructs so that researchers can move travel policies and practices is important to decreasing work
closer to the root causes of turnover. If it is important to ask, exhaustion.
“What factors lead to turnover?,” then it is also important to
ask, “What determines the predictors of turnover intention?” This study points to the value of fostering interpersonal
While this study has used several factors, more should be networks between RWs and key experts at headquarters for
identified that relate to the employee’s work, work–family companies to encourage and support autonomy for IT RWs.
situation (such as work-life balance and lifestyle accommo- Our respondents reported that they not only felt emotionally
dation), and stress. For example, additional antecedents and isolated but also felt disconnected from the knowledge of
moderators of WFC should be researched. It is possible that work practices and processes. Interpersonal networks in the
social support of an IT RW might mitigate stress felt due to form of knowledge repositories and employee directories,
WFC. Individual differences such as extraversion and which include areas of expertise and job responsibility, may
neuroticism may affect the positive or negative interpretation help IT RWs find answers to questions on their own or know
of work events, as might the organizational climate (Hart and who to call. In addition, when IT RWs return to company
Cooper 2001). Also, task complexity might moderate the headquarters for training or other reasons, organizations can
effects of job autonomy on WFC (Mack and McGee 2001). encourage face-to-face meetings between RWs and the
headquarters experts who support them. Personal relation-
This model did not examine the effects of model variables on ships between IT RWs and experts may mean that phone calls
absenteeism, client satisfaction, or job performance, each of and e-mails with requests for help are answered promptly.
which is a possible extension. For instance, it may be that the
levels of stress—WFC, work overload, and work exhaus- Researchers may identify additional factors that help support
tion—that lead to turnover intention also decrease job per- an autonomous work environment. For example, it is possible
formance, making them even more critical to ameliorate. that training in project management, problem-solving, and
Researchers should explore whether this is the case or interpersonal skills will help give IT RWs greater confidence
whether some inverted-U curve exists between types of stress in interacting with clients and managing their work
and job performance, as has been proposed (Muse et al. environment.
2003).
In practice, many companies face the issue of the effects of
Commitment and work exhaustion may also have an inter- WFC and autonomy on work exhaustion and find that it has
active effect on turnover intention, which is beyond the scope affected morale and commitment in their organizations (e.g.,
of this study. Further research is needed to tease out and Rothbard et al. 2001). The pressures of work, especially for
more fully explain these relationships. The differential effects those working in areas related to information technologies,
of role conflict, role ambiguity, and WFC should be examined have intensified in recent decades. This, when combined with
to understand in what context one is more salient than a need to travel and be away from family and home, can
another. Researchers may also want to study how specific as- accelerate the burnout rate. The IT organization should try to
pects of the RW job contribute to work exhaustion and WFC. accommodate the complete life needs of each employee
(Agarwal and Ferratt 1999). We suggest that training in
managing life-styles involving virtual work be a part of the
Implications for Practice support available to IT RWs. Employees should also be
provided with options for counseling on clarifying their life-
The findings of this study suggest that managers of RWs priorities. Based on these priorities, they can work with a
should focus on providing autonomy to their workers and company mentor to create a custom-career for themselves.
providing them enough flexibility to reduce the WFC they
feel as a result of the structure of their work situation. Our research is also important to practice because it points to
Managers should also be sure those who are promotable are factors that managers can influence. For example, instead of
told they are promotable, as this may compensate for the trying to change work exhaustion perceptions, managers can
WFC stresses they experience as RWs, improve organiza- more easily influence work overload by reducing what is on
tional commitment, and lower the risk of turnover. an employee’s plate. In the current study, work overload is

12 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

the major cause of both work exhaustion and WFC. Instead Family Conflict with Job and Life Satisfaction,” Journal of
of trying to address organizational commitment head-on, a Applied Psychology (81:4), 1996, pp. 411-420.
manager can control the amount of job autonomy granted and Ahuja, M., and Thatcher J. “Moving Beyond Intentions and Toward
the fairness of rewards. A manager can also take specific Theory of Trying: Linking the Work Environment and Gender
steps (outlined above) to reduce WFC among RWs. Hence, to Post-Adoption Information Technology Use,” MIS Quarterly
these secondary antecedents are important because manage- (29: 3), 2005, pp. 427-459.
ment can more easily address them than broader and more Anderson, J. C., and Gerbing, D. W. “Structural Equation Modeling
abstract concerns like commitment. in Practice: A Review and Recommended Two-Step Approach,”
Psychological Bulletin (103:3), 1988, pp. 411-423.
Ang, S., and Slaughter, S. “The Missing Context of Information
Technology Personnel: A Review and Future Directions for
Conclusion Research,” in Framing the Domains of IT Management, R. Zmud
(ed.), Pinnaflex Education Resources, Inc., Cincinnati, OH.,
In this study, we investigated IT road warriors and the factors 2000, pp. 305-327.
related to turnover intention. IT road warriors are IT pro- Armstrong, J. S., and Overton, T. S. “Estimating Nonresponse Bias
fessionals who are consultants located at client sites and, as in Mail Surveys,” Journal of Marketing Research (16), 1977, pp.
396-402.
such, spend most of their work time on the road. We adapted
Arnold, H. J., and Feldman, D. C. “A Multivariate Analysis of the
Moore’s turnover model to apply to these RWs. The results
Determinants of Job Turnover,” Journal of Applied Psychology
show that the traditional turnover factors, organizational com-
(67), 1982, pp. 350-360.
mitment and work exhaustion, were related to turnover inten-
Aryee, S., Luk, V., Leung, A., and Lo, S. “Role Stressors, Interrole
tion. Of greater interest, however, we clarified the effects of
Conflict, and Well-Being: The Moderating Influence of Spousal
four important antecedents of RW organizational commitment
Support and Coping Behaviors among Employed Parents in Hong
and work exhaustion: WFC, work overload, fairness of
Kong,” Journal of Vocational Behavior (54), 1999, pp. 259-278.
rewards, and autonomy. We also found that autonomy might Bailyn, L. “Toward a Perfect Work Place? The Experience of
be experienced differently among virtual workers, depending Home-Based Systems Developers,” in Information Technology
on the characteristics of the work environment. Although and the Corporation of the 1990s, T. Malone and M. S. Scott
previous research with telecommuters found that autonomy Morton (eds.), Oxford University Press, New York, 1994, pp.
could mitigate WFC, our study showed that while autonomy 410-428.
is important to IT RWs, it did not have a significant effect on Barclay, D., Higgins, C., and Thompson, R. “The Partial Least
WFC. Overall, the study shows that turnover models become Squares (PLS) Approach To Causal Modeling,” Technology
more predictive when researchers contextualize to the cir- Studies (2:2), 1995, pp. 285-309.
cumstances of the particular type of IT employee under Baroudi, J. J. “The Impact of Role Variables on IS Personnel Work
examination. Attitudes and Intentions,” MIS Quarterly (9:4), 1985, pp. 341-56.
Bartol, K. M. “Turnover Among DP Personnel: A Causal
Analysis,” Communications of the ACM (26:10), 1983, pp.
Acknowledgments 807-811.
Beehr, T. “Perceived Situational Moderators of the Relationship
We are very grateful to the managers and employees of our research between Subjective Role Ambiguity and Role Strain,” Journal of
site for their help and cooperation with this project. We would also Applied Psychology (61), 1976, pp. 35-40.
like to thank Senior Editor Ritu Agarwal, our AE and reviewers, and Belanger, F. “Workers’ Propensity to Telecommute: An Empirical
Michael Dickey, Jason Thatcher, and Molly Wasko for their Study,” Information & Management (35:3), 1999, pp. 139-153.
guidance and suggestions. An earlier version was published in the Boles, J., Dean, D., Ricks, J., Short, J., and Wang, G. “The
electronic Proceedings of the 35th Annual Hawaii International Dimensionality of the Maslach Burnout Inventory Across Small
Conference on System Sciences. Business Owners and Educators,” Journal of Vocational
Behavior (56), 2000, pp. 12-34.
Boles, J., Johnston, S. M., and Hair, J. Jr. “Role Stress, Work–
References Family Conflict and Emotional Exhaustion: Inter-Relationships
and Effects on Some Work-Related Consequences,” The Journal
Agarwal, R., and Ferratt, T. W. Coping with Labor Scarcity in of Personal Selling & Sales Management (17:1), Winter 1997,
Information Technology: Strategies and Practices for Effective pp. 17-28.
Recruitment and Retention, Pinnaflex Education Resources, Inc., Chin, W. W. “The Partial Least Squares Approach to Structural
Cincinnati, OH, 1999. Equation Modeling,” in Modern Methods for Business Research,
Adams, G. A., King L. A., and King D. W. “Relationships of Job G. A. Marcoulides (ed.), Lawrence Erlbaum, Mahwah, NJ, 1998,
and Family Involvement, Family Social Support, and Work– pp. 295-336.

MIS Quarterly Vol. 31 No. 1/March 2007 13

Ahuja et al./Mitigating Turnover Intentions

Cotton, J. L., and Tuttle, J. M. “Employee Turnover: A Meta- Large Organization,” Organization Science (6:4), 1995, pp.
Analysis and Review with Implications for Research,” Academy 373-393.
of Management Review (11:1), 1986, pp. 55-70. Hom, P. W., and Griffeth, R. W. Employee Turnover, International
Cropanzano, R., Rupp, D., and Byrne, Z. “The Relationship of Thompson Publishing, Cincinnati, OH, 1995.
Emotional Exhaustion to Work Attitudes, Job Performance, and Hu, L., and Bentler, P. “Cutoff Criteria for Fit Indexes in Co-
Organizational Citizenship Behaviors,” Journal of Applied variance Structure Analysis: Conventional Criteria Versus New
Psychology (88:1), 2003, pp. 160-169. Alternatives,” Structural Equation Modeling (6), 1999, pp. 1-55.
Cropanzano, R., Rupp, D., Mohler, C., and Schminke, M. “Three Igbaria, M., and Greenhaus, J. H. “Determinants of MIS Em-
Roads to Organizational Justice” in Research in Personnel and ployees’ Turnover Intentions: A Structural Equation Model,”
Human Resources Management, J. Ferris (ed.), JAI Press, Communications of the ACM (35:2), 1992, pp. 35-49.
Greenwich, CT, 2001, pp. 1-113. Igbaria, M., Greenhaus, J. H., and Parasuraman, S. “Career Orien-
Dalton, D. R., Johnson, J. L and Daily, C. M. “On the Use of ‘Intent tations of MIS Employees: An Empirical Analysis,” MIS
to…’ Variables in Organizational Research: An Empirical and Quarterly (15:2), 1991, pp. 151-168.
Cautionary Assessment,” Human Relations (52), 1999, pp. Igbaria, M., and Guimaraes, T. “Exploring Differences in Employee
1337-1350. Turnover Intentions and its Determinants Among Telecommuters
Dittrich, J. E., Couger, J. D., and Zawacki, R. A. “Perceptions of and Non-Telecommuters,” Journal of Management Information
Equity, Job Satisfaction, and Intention to Quit among Data Systems (16:1), 1999, pp. 147-164.
Processors,” Information & Management (9:2), 1985, pp. 67-75. Jiang, J. J., and Klein, G. “A Discrepancy Model of Information
Eby, L. T., Freeman, D. M., Rush, M. C., and Lance, C. E. Systems Personnel Turnover,” Journal of Management Infor-
“Motivational Bases of Affective Organizational Commitment: mation Systems (19:2), 2002, pp. 249-272.
A Partial Test of an Integrative Theoretical Model,” Journal of Kalliath, T., O’Driscoll, M., and Gillespie, D. “The Relationship
Occupational and Organizational Psychology (72), 1999, pp.
Between Burnout and Organizational Commitment in Two
463-483.
Samples of Health Professionals,” Work and Stress (12:2), 1998,
Fornell, C., and Larcker, D. “Evaluating Structural Equation
pp. 179-185.
Models with Unobservable Variables and Measurement Error,”
Kaldenberg, D. O., and Becker, B. W. “Workload and Psycho-
Journal of Marketing Research (18:3), 1981, pp. 39-50.
logical Strain: A Test of the French, Rodgers, and Cobb Hypo-
Frone, M., Russell, M., and Cooper, M. “Antecedents and Out-
thesis,” Journal of Organizational Behavior (13), 1992, pp.
comes of Work–Family Conflict: Testing a Model of the Work–
617-624.
Family Interface,” Journal of Applied Psychology (77), 1992, pp.
Kickul, J., and Posig, M. “Supervisory Emotional Support and
65-78.
Burnout: An Explanation of Reverse Buffering Effects,” Journal
Frone, M., Yardley, J., and Markel, K. “Developing and Testing an
of Managerial Issues (8:3), 2001, pp. 328-344.
Integrative Model of the Work–Family Interface,” Journal of
Vocational Behavior (50), 1997, pp. 145-167. King, J. “On-the-Fly IT,” Computerworld (37:36), September 8,
Goff, L. J. “Balancing Acts,” Computerworld, May 7, 2001, p. 49. 2003, p. 35.
Goldstein, N. IT at Work: Information Technologies and Remote Lee, R. T., and Ashforth, B. E. “A Meta-Analytic Examination of
Working in the United States, SRI International, Arlington, the Correlates of the Three Dimensions of Burnout,” Journal of
VA, 2003. Applied Psychology (81:2), 1996, pp. 123-133.
Greenhaus, J. H. “The Intersection of Work and Family Roles: Lee, T. W., and Mitchell, T. R. “An Alternative Approach: The
Individual, Interpersonal, and Organizational Issues,” Journal of unfolding Model of Voluntary Employee Turnover,” Academy of
Social Behavior and Personality (3), 1988, pp. 23-44. Management Review (19), 1994, pp. 51-89.
Greenhaus, J. H., and Beutell, N. J. “Sources of Conflict Between Leiter, M. P., and Maslach, C. “The Impact of Interpersonal
Work and Family Roles,” Academy of Management Review (10), Environment on Burnout and Organizational Commitment,”
1985, pp. 76-88. Journal of Occupational Behavior (8), 1988, pp. 297-308.
Hackman, J., and Oldham, G. “Development of the Job Diagnostic Leiter, M. P., and Schaufeli, W. B. “Consistency of the Burnout
Survey,” Journal of Applied Psychology (60), 1975, pp. 159-170. Construct Across Occupations,” Anxiety, Stress, and Coping (9),
Hart, P. M., and Cooper, C. L. “Occupational Stress: Toward a 1996, pp. 229-243.
More Integrated Framework,” in Handbook of Industrial, Work Lovett, S., Harrison, D., and Virick, M. “Managing the Boundary
and Organizational Psychology, N. Anderson, D. S. Ones, H. K. Spanner—Customer Turnover Connection,” Human Resource
Sinanil, and C. Viswesvaran, (eds.), Sage Publications, London, Management Review (7:4), 1997, pp. 405-424.
2001, pp. 93-114. Mack, D. A., and McGee, J. E. “Occupational Stress and the Small
Hill, E. J., Miller, B. C., Weiner, S. P., and Colihan, J. “Influences Business Owner: The Role of Task Complexity and Social
of the Virtual Office on Aspects of Work and Work/Life Support,” working paper, University of Texas at Arlington, 2001
Balance,” Personnel Psychology (51:3), 1998, pp. 667-683. (http://www.usasbe.org/knowledge/proceedings/2001/045.pdf).
Hinds, P., and Kiesler, S. “Communication Across Boundaries: Madden, J., “Road Warriors,” American Demographics, October
Work, Structure, and Use of Communication Technologies in a 1995, pp. 4-9.

14 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

Mannheim, B., and Schiffrin, M. “Family Structure, Job Charac- for Customer Service Representatives,” Journal of Marketing
teristics, Rewards and Strains as Related to Work-Role Centrality Research (31), 1994, pp. 558-669.
of Employed and Self-Employed Professional Women with Singh, J., Verbeke, W., and Rhoads, G. “Do Organizational
Children,” Journal of Occupational Behavior (5:2), 1984, pp. Practices Matter in Role Stress Processes? A Study of Direct
83-107. and Moderating Effects for Marketing-Oriented Boundary
Maslach, C. Burnout: The Cost of Caring, Prentice-Hall, Engle- Spanners,” Journal of Marketing (60:3), 1996, pp. 69-86.
wood Cliffs, NJ, 1982. Speier, C., and Venkatesh, V., “The Hidden Minefields in the
Maslach, C., and Jackson, S. E. “Burnout in Organizational Adoption of Sales Force Automation Technologies,” Journal of
Settings,” in Applied Social Psychology Annual: Applications in Marketing (66:3), 2002, pp. 98-111.
Organizational Settings (5), S. Oskamp (ed.), Sage Publications, Thatcher, J. B., and Stepina, L. P. “Information Technology Worker
Beverly Hills, CA, 1984, pp. 133-153. Turnover: An Integrative Model and Empirical Test,” in Pro-
Mobley, W. H., Griffeth, R. W., Hand, H. H., and Meglino, B. M. ceedings of the 22nd International Conference on Information
“Review and Conceptual Analysis of the Employee Turnover Systems, W. J. Orlikowski, S. Ang, P. Weill, H. C. Krcmar, and
Process,” Psychological Bulletin (86), 1979, pp. 493-522. J. I. DeGross (eds.), Brisbane, Australia, December 10-13, 2001,
Moore, J. E. “One Road to Turnover: An Examination of Work pp. 285-288.
Exhaustion in Technology Professionals,” MIS Quarterly (24:1), Thatcher, J. B., Stepina, L. P., and Boyle, R. J. “Turnover of Infor-
2000, pp. 141-168. mation Technology Workers: Examining Empirically the
Mowday, R. T., Porter, L. W., and Steers, R. M. Employee–Organi- Influence of Attitudes, Job Characteristics, and External Mar-
zation Linkages: The Psychology of Commitment, Absenteeism, kets,” Journal of Management Information Systems (19:3), 2002,
and Turnover, Academic Press, New York, 1982. pp. 231-261.
Muse, L. A., Harris, S. G., and Field, H. S. “Has the Inverted-U Thomas, K. M., and Williams, K. L. “The Role of Burnout on
Theory of Stress and Job Performance Had a Fair Test?,” Human Organizational Attachment and Career Mobility,” in Proceedings
Performance (16:4), 2003, pp. 349-364. of the Work, Stress and Health ’95: Creating Healthier Work-
Netemeyer, R. G., Boles, J. S., and McMurrian, R. “Development places, Washington, DC, September 1995, pp. 265-273.
and Validation of Work–Family Conflict and Family-Work Thomas, L. T., and Ganster, D. C. “Impact of Family-Supportive
Conflict Scales,” Journal of Applied Psychology (81:4), 1996, pp. Work Variables On Work–Family Conflict And Strain: A Con-
400-410. trol Perspective,” Journal of Applied Psychology (80), 1995, pp.
Netemeyer, R. G., Brashear-Alejandro, T., and Boles, J. S. “A 6-15.
Cross-National Model of Job-Related Outcomes of Work Role Tsui, A., Pearce, J. L., Porter, L. W., and Tripoli, A. M. “Alterna-
and Family Role Variables: A Retail Sales Context,” Journal of tive Approaches to the Employee-Organization Relationship:
the Academy of Marketing Science (32:1), 2004, pp. 49-60. Does Investment in Employees Pay Off?,” Academy of
Oldham G. R., and Rotchford, N. L. “Relationship Between Office Management Journal (40:5), 1997, pp. 1089-1121.
Characteristics and Employee Reactions: A Study of the Widaman, K. “Hierarchical Nested Covariance Structure Models
Physical Environment,” Administrative Science Quarterly (28), for Multitrait-Multimethod Data, Applied Psychological Mea-
1983, pp. 542-556. surement (9), 1985, pp. 1-26.
Parasuraman, S., Purohit, Y., Godshalk, V., and Beutell, N. “Work Williams, L., Cote, J., and Buckley, M. “Lack of Method Variance
and Family Variables, Entrepreneurial Career Success, and in Self-Reported Affect and Perceptions of Work: Reality or
Psychological Well-Being,” Journal of Vocational Behavior (48), Artifact?,” Journal of Applied Psychology (74), 1989, pp.
1996, pp. 275-300. 462-468.
Pratt, J. Costs/Benefits of Teleworking to Manage Work/Life Xie, J. L., and Johns, G. “Job Scope and Stress: Can Job Scope Be
Responsibilities, International Telework Association and Council Too High?,” Academy of Management Journal (38:5), 1995, pp.
(ITAC), 1999 (available from [email protected]). 1288-1309.
Rothbard, N. P., Dumas, T. P., and Phillips, K. W. “Role Conflict,
Work–Family Policies, and Employee Preferences for Segmen-
tation,” in Proceedings of the International Association for Con- About the Authors
flict Management 14th Annual Conference, Cergy (Paris), France,
June 24-27, 2001. Manju Ahuja is an assistant professor of MIS at the Kelley School
Salaff, J. W. “Where Home Is the Office: The New Form of of Business, Indiana University. Her publications have appeared in
Flexible Work,” working paper, Department of Sociology, Centre journals such as MIS Quarterly, Management Science, Organization
for Urban and Community Studies, University of Toronto, 2002. Science, Communications of the ACM, Journal of Management,
Schaufeli, W. B., Leiter, M. P., and Kalimo, R. “The General Burn- European Journal of Information Systems, Small Group Research,
out Inventory: A Self-Report Questionnaire to Assess Burnout Decision Support Systems, The Database for Advances in Infor-
at the Workplace,” in Proceedings of the Work, Stress and Health mation Systems, and The Journal of Computer-Mediated Communi-
’95: Creating Healthier Workplaces, Washington, DC, Sep- cations. She is an associate editor at Information Systems Research
tember 1995, pp. 10-23. and also serves on the review boards of Decision Sciences and Jour-
Singh, J., Goolsby, J. R., and Rhoads, G. “Behavioral and Psycho- nal of the Association for Information Systems. She has served as a
logical Consequences of Boundary Spanning: Burnout guest coeditor for Information, Technology, and People (2005-2006)

MIS Quarterly Vol. 31 No. 1/March 2007 15

Ahuja et al./Mitigating Turnover Intentions

and The Database for Advances in Informatoin Systems (2004). Dr. D. Harrison McKnight received his Ph.D. in Management
Ahuja is actively involved in research on issues related to virtual Information Systems from the University of Minnesota. He is an
communities, virtual teams, and management of human resources in assistant professor in the College of Business at Michigan State
IT professions. Recently, she was awarded a National Science University. His research interests include trust building within e-
Foundation grant to examine Information Technology workforce commerce and organizational settings and the retention and moti-
issues. Her research has been cited by publication such as Wall vation of information systems professionals. His work has appeared
Street Journal, Strategy+Business, and Computerworld. in such journals as Information Systems Research, Journal of
Strategic Information Systems, International Journal of Electronic
Katherine M. Chudoba is an assistant professor of MIS at Florida Commerce, Electronic Markets, and the Academy of Management
State University. Dr. Chudoba’s research interests focus on the Review.
nature of work in distributed environments, and how ICT are used
and integrated into work practices. She earned her Ph.D. at the Joey F. George is a professor of Information Systems and the
University of Arizona, and her bachelor’s degree and MBA at the Thomas L. Williams Jr. Eminent Scholar in Information Systems in
College of William and Mary. Prior to joining academe, she worked the Management Information Systems Department in the College of
for 8 years as an analyst and manager in the information technology Business at Florida State University. He received his Ph.D. from the
industry. Her research has been sponsored by companies such as
University of California, Irvine, in 1986, and an A.B. from Stanford
Intel and Xerox, and published in Organization Science, Information
University in 1979. His research interests focus on the use of infor-
Systems Journal, Database, Information & Management, Infor-
mation systems in the workplace, including the detection of decep-
mation Technology & People, and Information Research.
tive computer-mediated communication, computer-based moni-
toring, and group support systems. Dr. George has published over
Charles J. Kacmar is an associate professor in the MIS Department
40 articles in such journals as Information Systems Research, MIS
at the University of Alabama. He received his Ph.D. in computer
science from Texas A&M University. His research interests include Quarterly, Communications of the ACM, Journal of MIS, Com-
behavioral and organizational information systems, human-computer munication Research, and Small Group Research. Dr. George is co-
interaction, collaborative systems, and hypertext/hypermedia. He author of several textbooks, including the bestselling Modern
has published in Academy of Management Journal, Information Systems Analysis and Design, fourth edition (Prentice Hall, 2004).
Systems Research, Journal of Strategic Information Systems, Com- He is currently the editor-in-chief for Communications of the AIS
munications of the ACM, ACM Transactions on Information Sys- and recently served as a senior editor for MIS Quarterly. Dr. George
tems, Hypermedia, and Behaviour, and Information Technology. He served as the Conference Co-Chair for the 2001 International
is a member of the Association for Information Systems and the Conference on Information Systems (ICIS) in New Orleans and as
ACM. Co-Chair of the Doctoral Consortium for ICIS 2003 in Seattle, WA.

Appendix
Measurement Scales
Job Autonomy (Beehr 1976)
Indicate the extent to which these statements reflect your feelings about your current job. (1-7, with ends and midpoint anchors)
1. I control the content of my job.
2. I have a lot of freedom to decide how I perform assigned tasks.
2. I set my own schedule for completing assigned tasks.
4. I have the authority to initiate projects at my job.

Work–Family Conflict (Adams et al. 1996)

If you are not married and/or do not have children, you can choose to respond to these questions in terms of your life outside of work in general
(for example, replace “family” with “friends” and think of your other commitments, such as gymnasiums, book clubs, or any other hobbies)
(1-7, with ends and midpoint anchors).
1. The demands of my work interfere with my home and family life.
2. The amount of time my job takes up makes it difficult to fulfill family responsibilities.
3. Things I want to do at home do not get done because of the demands my job puts on me.
4. My job produces strain that makes it difficult to fulfill family duties.
5. Due to work-related duties, I have to make changes to my plans for family activities.

16 MIS Quarterly Vol. 31 No. 1/March 2007

 Ahuja et al./Mitigating Turnover Intentions

Organizational Commitment (Tsui et al. 1997)

Think about your organization. Indicate the extent to which you agree or disagree with these statements (1-7, with ends and midpoint anchors)
1. I am willing to put in effort beyond the norm for the success of the organization.4
2. For me, this is the best of all possible organizations for which to work.
3. I am extremely glad to have chosen this organization to work for over other organizations.
4. This organization inspires the very best in the way of job performance.
5. I show by my actions that I really care about the fate of this organization.

Work Exhaustion (Moore 2000; used the first four of Moore’s items; scaling the same as Moore)
(0 = never; 1 = A few times a year or less, almost never; 2 = Once a month or less, rarely; 3 = A few times a month, sometimes; 4 = Once a
week, rather often; 5 = A few times a week, nearly all the time; 6 = Daily)
1. I feel emotionally drained from my work.
2. I feel used up at the end of the work day.
3. I feel fatigued when I get up in the morning and have to face another day on the job.
4. I feel burned out from my work.

Perceived Work Overload (Moore 2000; items same; scaling adapted for items 1 and 2)
(1 = Daily; 2 = Almost every day; 3 = About once a week; 4 = Two or three times a month; 5 = About once a month; 6 = A few times a year;
7 = Once a year or less) (Reverse scored)
1. I feel that the number of requests, problems, or complaints I deal with is more than expected.
2. I feel that the amount of work I do interferes with how well it is done.
3. I feel busy or rushed.
4. I feel pressured.

Fairness of Rewards (new items) (1-7, with ends and midpoint anchors)
1. My organization has processes that assure that all team members will be treated fairly and equitably.
2. I work in an environment in which good procedures make things fair and impartial.
3. In my workplace, sound practices exist that help ensure fair and unbiased treatment of all team members.
4. Fairness to employees is built into how issues are handled in my work environment.

Turnover Intention (Moore 2000; items same; scaling altered for items 3 and 4)
(1 = very unlikely; 5 = very likely)
1. How likely is it that you will be working at the same company this time next year? (R)
2. How likely is it that you will take steps during the next year to secure a job at a different company?

(7-point scale; 1 = very unlikely; 4 = Neutral; 7 = very likely)

3. I will be with this company five years from now. (R)
4. I will probably look for a job at a different company in the next year.

4
Item dropped after initial inspection of outer model loadings.

MIS Quarterly Vol. 31 No. 1/March 2007 17