The Good, the Bad, and the Unknown About Telecommuting: Meta- Analysis of Psychological Mediators and Individual Consequences

Journal of Applied Psychology
Copyright 2007 by the American Psychological Association
2007, Vol. 92, No. 6, 1524 –1541
0021-9010/07/$12.00
DOI: 10.1037/0021-9010.92.6.1524
The Good, the Bad, and the Unknown About Telecommuting: Meta-
Analysis of Psychological Mediators and Individual Consequences
Ravi S. Gajendran and David A. Harrison
Pennsylvania State University
What are the positive and negative consequences of telecommuting? How do these consequences come
about? When are these consequences more or less potent? The authors answer these questions through
construction of a theoretical framework and meta-analysis of 46 studies in natural settings involving
12,883 employees. Telecommuting had small but mainly beneficial effects on proximal outcomes, such
as perceived autonomy and (lower) work–family conflict. Importantly, telecommuting had no generally
detrimental effects on the quality of workplace relationships. Telecommuting also had beneficial effects
on more distal outcomes, such as job satisfaction, performance, turnover intent, and role stress. These
beneficial consequences appeared to be at least partially mediated by perceived autonomy. Also,
high-intensity telecommuting (more than 2.5 days a week) accentuated telecommuting’s beneficial
effects on work–family conflict but harmed relationships with coworkers. Results provide building
blocks for a more complete theoretical and practical treatment of telecommuting.
Keywords: telecommuting, distributed work, virtual work, meta-analysis
Most current work arrangements still bear the imprint of the
Investigations of the impact of such flexible work locations
Industrial Revolution. Employees mainly transact their time, rather
began in earnest over 20 years ago (e.g., Ramsower, 1983). More
than their products, with employing firms. That time is tightly
recent studies use the umbrella term distributed work: arrange-
bound to task and place (Lakoff & Johnson, 1980). In recent
ments that allow employees and their tasks to be shared across
decades, however, an information revolution has compelled firms
settings away from a central place of business or physical organi-
to unbind time and task from place (Harrison, Johns, & Martoc-
zational location (Be´langer & Collins, 1998). The most well-
chio, 2000). Digital technologies have enabled common, even
known form of distributed work, telecommuting, which is also
synchronous activities to be distributed across employees at re-
known as telework or remote work, has become a widespread
mote locations (Herschel & Andrews, 1997). These decentralized
practice. It is growing steadily in the United States and abroad
work arrangements have been publicized as a way for firms to
(Davis & Polonko, 2001). An estimated 45 million American
reduce real estate expenses and comply with government regula-
employees telecommuted in 2006, up from 41 million in 2003
tions (e.g., the Americans With Disabilities Act of 1990, Daven-
(WorldatWork, 2006). Increasingly, U.S. federal and state govern-
port & Pearlson, 1998; U.S. Equal Employment Opportunity Com-
ments have encouraged this work arrangement through the intro-
mission, 2005; Wells, 2001). They are likewise touted as means for
duction of legislation (e.g., Department of Transportation and
employees to adjust their schedule to meet household needs and
Related Agencies Appropriations Act of 2000; U.S. Office of
family demands or to save commuting costs by working from
Personnel Management, 2005). Outside the United States, a survey
home or satellite offices (HR Focus, 2002; Nickson & Siddons,
of 254 senior-level executives revealed that staff in two thirds of
2004). That is, through practices that permit flexibility in the
their global firms were involved in distributed work (AT&T,
“where” of tasks, organizations have adapted work arrangements
2004).
to fit changing environments and labor needs (Igbaria & Guima-
Such a sustained rise in popularity suggests a received wisdom
raes, 1999).
of positive outcomes or clear benefits of telecommuting for firms
and their employees. Indeed, at the firm level, a number of prac-
titioner articles and company statistics have argued and demon-
Ravi S. Gajendran and David A. Harrison, Department of Management
strated that telecommuting reduces real estate costs (e.g., Apgar,
and Organization, Pennsylvania State University.
1998; AT&T, 1997; Dannhauser, 1999). At the individual level as
We are grateful to Jim Detert, Dennis A. Gioia, Raghu Garud, Arijit
Chatterjee, Jayashree Sridhar, and members of the Organizations Research
well, the benefits of telecommuting are almost unswervingly
Group at Pennsylvania State University for their comments on drafts of this
claimed by the business press. Improved work–life balance,
article and to the eBusiness Research Center at the Smeal School of
heightened morale, and increased productivity are widely adver-
Business, Pennsylvania State University, for their support.
tised to practitioners (Apgar, 1998; Brownson, 2004; Tergesen,
Correspondence concerning this article should be addressed to Ravi S.
1998; Young, 1993). Yet such claims are often criticized by
Gajendran, Department of Management and Organization, Pennsylvania
scholars as being based on inconsistent findings, methodologically
State University, 426 B Business Building, University Park, PA 16802, or
weak supporting studies, and unconvincing theory regarding why
to David A. Harrison, Department of Management and Organization,
Pennsylvania State University, 403 Business Building, University Park, PA
and when those benefits should occur (e.g., Be´langer & Collins,
16802. E-mail: rsg183@psu.edu or dharrison@psu.edu
1998). Moreover, negative consequences for individuals, such as
1524

---

CONSEQUENCES OF TELECOMMUTING
1525
social isolation, career stagnation, and family conflict, have also
commuters are not really members of the contracting organization
been averred in investigations that sound an opposing drumbeat
(Qvortrup, 1998). It also excludes tasks that are normally per-
(Baruch & Nicholson, 1997).
formed away from a central location of work as part of a conven-
Thus, despite the growing importance and widely spreading
tional work arrangement (e.g., sales trips). Conversely, the defini-
practice of telecommuting, reviews of the last 2 decades of re-
tion acknowledges that telecommuting can be practiced by an
search have concluded that it is unknown whether telecommuting
individual part time (alongside more conventional work arrange-
is good or bad for employees. Existing evidence is indeterminate
ments) or full time (exclusively working from home). In a recent
and often contradictory (McCloskey & Igbaria, 1998). According
review, Bailey and Kurland (2002) suggested that some parts of
to one review, “empirical research to date has been largely unsuc-
the literature have implicitly defined telecommuting as full time
cessful in explaining what happens” (Bailey & Kurland, 2002, p.
(always at home, never “at work”), which does not reflect current
394) after firms and employees adopt distributed work arrange-
practice. In fact, fewer than 10% of telecommuting employees are
ments. A major hurdle in drawing definitive conclusions about
involved in full-time arrangements (Davenport, 2005); part-time
telecommuting’s impact is that studies of this innovation appear in
arrangements prevail (Qvortrup, 1998; Standen, Daniels, & Lam-
dispersed literatures (information systems, logistics, industrial re-
ond, 1999). Our definition also acknowledges that individuals
lations, psychology, operations, real estate, and management; Be´-
differ in their engagement with telecommuting. Doing so for part
langer & Collins, 1998), attracting the interest of scholars within
of the week (1 or 2 days) represents a less intense form of
multiple disciplines. A key contribution of this article, therefore, is
telecommuting compared to spending the major portion of one’s
our systematic attempt at locating these studies and summarizing
work week away from the central location. We propose that this
their findings to address what is still unknown despite decades of
difference represents a continuum of psychological commitment to
investigation. We identify studies that address employee outcomes
the telecommuting arrangement and therefore could have a differ-
of telecommuting, quantitatively summarize available findings,
ential impact on the consequences of telecommuting, an idea we
and provide answers to three fundamental questions for manage-
expand on in a later section.
ment research. First, is telecommuting or distributed work effec-
tive? What are its predictable positive (and negative) consequences
Conceptual Themes in the Telecommuting Literature
for individuals? Second, how do those consequences occur? What
psychological mechanisms carry telecommuting’s effects? Third,
Despite a growing consensus on the structural features and
when do those consequences occur? Under what conditions does
importance of telecommuting as an alternative work arrangement,
telecommuting have its strongest effects? In answering these ques-
there is no single, overarching theory of its consequences. Most
tions, we not only test existing theory, we also forward building
frequently, however, studies have highlighted one or more of the
blocks and a general framework for future theory development.
three conceptual themes we detail below (D. G. Allen, Renn, &
Specifically, we use meta-analytic cumulation to resolve long-
Griffeth, 2003; Shamir & Salomon, 1985). These themes also
standing debates about the individual costs and benefits of tele-
constitute statements about the psychological process or interven-
commuting and identify circumstances under which those effects
ing mechanisms through which telecommuting has its effects (see
are weaker or stronger. We use our findings (and some nonfind-
the first and second research questions highlighted above).
ings) to propose a research agenda for more nuanced and program-
The first conceptual theme deals with psychological control or
matic investigations of distributed work arrangements.
perceived autonomy, which is a key feature of any work arrange-
ment. It comprises employees’ personal assessments of the extent
The Meaning of Telecommuting
to which they can “structure and control how and when they do
their particular job tasks” (Spector, 1986, p. 1006). Scholars as-
We draw on recent reviews and theories to forward a definition
serting this view see telecommuting as a good thing that is thought
of the focal phenomenon that we believe represents an emerging
to enhance perceived autonomy by providing employees with
consensus in the literature. Telecommuting is an alternative work
choice over the location, scheduling (at least for some), and means
arrangement in which employees perform tasks elsewhere that are
of work (e.g., DuBrin, 1991; Standen et al., 1999).
normally done in a primary or central workplace, for at least some
The second conceptual theme concerning telecommuting’s ef-
portion of their work schedule, using electronic media to interact
fects on the work–family interface has been the subject of much
with others inside and outside the organization (Bailey & Kurland,
scholarly debate. Some scholars view telecommuting as a good
2002; Baruch, 2001; Feldman & Gainey, 1997). There is a substi-
thing that leads to greater integration between the work and family
tution of place involved in telecommuting, and a restriction of
roles (e.g., Duxbury, Higgins, & Neufeld, 1998; Raghuram &
interactions occurs because of the physical and psychological
Wiesenfeld, 2004). Others regard telecommuting as a bad thing
distance involved in that substitution. Typically, elsewhere in the
that may intensify conflict by increasing the permeability of work
definition above is home, although telework centers and remote
and family boundaries (e.g., Igbaria & Guimaraes, 1999; Standen
offices are alternative locations (Davis & Polonko, 2001; Hill,
et al., 1999). Empirical evidence resolving this debate has been
Miller, Weiner, & Colihan, 1998). Home was the primary location
inconclusive (Duxbury et al., 1998; Raghuram & Wiesenfeld,
for telecommuting in nearly all the studies included in this meta-
2004).
analysis.
The third conceptual theme deals with concerns about telecom-
This definition is broad enough to include most telecommuting
muting’s potential for relational impoverishment at work. The
forms discussed in academic and practitioner literatures. Yet it
reduction in face-to-face interactions, the lower frequency and
deliberately excludes electronic outsourcing, a type of independent
richness of communication between telecommuters and other or-
contracting in which those who might otherwise be called tele-
ganization members (Daft & Lengel, 1986), and, thus, the dimin-

---

1526
GAJENDRAN AND HARRISON
ished social presence (Short, Williams, & Christie, 1976) telecom-
marized in Figure 1. That framework guides our meta-analysis.
muters have weakens the interpersonal bonds they have with their
Additional conceptual support for our framework, especially for
coworkers or supervisors (Golden, 2006b; Nardi & Whittaker,
such a central role of perceived autonomy, work–family conflict,
2002). These negative consequences are likely to be especially
and relationship quality, comes from models developed by D. G.
severe for individuals who work away from their central work
Allen et al. (2003) and Feldman and Gainey (1997). Our choice of
location for the major portion of their work week.
individual outcomes of telecommuting is also derived from previ-
Taken together, these themes hint at a “telecommuting paradox”
ous treatments of the consequences of telecommuting (e.g., Shamir
of mutually incompatible consequences for employees. If telecom-
& Salomon, 1985). Thus, our framework is a parsimonious col-
muting were to enhance perceived autonomy and lower work–
lection of important proximal (mediating) and distal consequences
family conflict, this would, in turn, enhance job-related attitudes,
of telecommuting, and it also integrates prior theorizing on the
improve performance, and reduce stress. Simultaneously, if tele-
topic.
commuting also were to damage vital work relationships and
hamper career advancement, this would imply that outcomes in the
Telecommuting’s Effects on Psychological Mediators
work and nonwork domains come at the expense of— or are
negatively correlated with— outcomes in the relationship or social
Telecommuters in general are likely to experience increased
domain. The notion that the first two sets of consequences are
feelings of freedom and discretion because they are spatially and
incompatible with the third, at least in conventional work arrange-
psychologically removed from direct, face-to-face supervision
ments, comes from previous meta-analytic research. However,
(DuBrin, 1991). Those employees engaged in part-time telecom-
relationship quality with one’s supervisor and coworkers is posi-
muting arrangements are also likely to experience increased au-
tively associated with attitudinal outcomes, such as increased job
tonomy because of the flexibility they are afforded over the loca-
satisfaction, as well as behavioral and physiological outcomes,
tion of their work (Shamir & Salomon, 1985). An implicit
such as enhanced effectiveness and reduced stress (e.g., Gerstner
assumption in the telecommuting literature has been that flexibility
& Day, 1997; Sousa-Poza & Sousa-Poza, 2000; Viswesvaran,
in work location is likely to increase self-reliance in scheduling
Sanchez, & Fisher, 1999). In the following sections, we examine
particular tasks and to increase control over the means of com-
whether this paradox is sustained by empirical evidence through a
pleting them: Flexibility equals control (Duxbury et al., 1998;
quantitative summary and cumulative tests of telecommuting find-
Raghuram, Garud, Wiesenfeld, & Gupta, 2001). This increased
ings.
flexibility in the timing and execution of tasks enhances employ-
ees’ perceptions of autonomy (Hackman & Oldham, 1976). In
Theoretical Framework for Telecommuting Meta-Analysis
addition, performing one’s tasks at home allows control over
The three conceptual themes summarized above play a central
breaks, clothing, layout, decoration, lighting, ventilation, music,
role as intervening mechanisms in the theoretical framework sum-
and other ambient elements that can contribute to increased feel-
Figure 1.
Theoretical framework for the consequences of telecommuting.

---

CONSEQUENCES OF TELECOMMUTING
1527
ings of autonomy (Elsbach, 2003; Standen, 2000). These argu-
relationships between telecommuters and their supervisors and
ments lead us to propose the following:
coworkers (e.g., Duxbury & Neufeld, 1999; Harrison, Johns, &
Martocchio, 2000; Higa, Sheng, Shin, & Figueredo, 2000; Work-
Hypothesis 1: Telecommuting is positively related to per-
man, Kahnweiler, & Bommer, 2003). Face-to-face communication
ceived autonomy.
is considered the medium with the highest social presence and
media richness. Because telecommuting reduces face-to-face com-
As we suggested earlier, debate continues about the work–
munication, both theories make similar predictions about quality
family consequences of telecommuting. On one hand, telecommut-
and frequency of interaction, implying mainly negative impacts on
ing increases the permeability of boundaries in life domains,
interpersonal relationships for telecommuters (see also Hallowell,
making it easier for one domain to intrude on the other, potentially
1999; Nardi & Whittaker, 2002; Nohria & Eccles, 1992; see
leading to work–family conflict (Standen et al., 1999). Boundary
updates and expansions of both theories by Carlson & Zmud,
permeability in the context of telecommuting refers to the degree
1999; Lombard & Ditton, 1997).
to which either family or work encroaches on the other because
Changing one’s workplace from a conventional office to a home
they occupy the same place and, potentially, the same time (Ash-
or an alternate location is likely to alter the frequency, the quality,
forth, Kreiner, & Fugate, 2000; Edwards & Rothbard, 2000;
and, by definition, the modality of interaction one has with other
Nippert-Eng, 1996). Such permeability could also make psycho-
organization members. Telecommuting therefore has the potential
logical disengagement from work more difficult, increasing the
to degrade the quality of the manager–subordinate relationship
likelihood of time-based conflict (Greenhaus & Beutell, 1985). For
(Reinsch, 1999). Managers fear reduced control over their subor-
example, the information and communication technologies sup-
dinates, while employees fear isolation and information impover-
porting telecommuting may encourage employees to continue
ishment (Kurland & Cooper, 2002; McCloskey & Igbaria, 2003).
working at home even after normal work hours (Boswell & Olson-
Reduced face-to-face interaction also makes immediate feedback,
Buchanan, 2004). This may be especially true for individuals with
as well as affective signals, more difficult to send, sustain, and
an integrative boundary management strategy who find it difficult
receive (Hallowell, 1999). Employees who choose to telecommute
to separate activities between home and work (Olson-Buchanan &
may also find their loyalty and commitment being questioned by
Boswell, 2006).
managers (Desrosiers, 2001; McCloskey & Igbaria, 2003).
On the other hand, increased boundary flexibility from telecom-
In addition, managers might have to change their strategies for
muting can help employees regulate and synchronize demands
monitoring employees from behavior-based to output-based con-
between work and family and, potentially, reduce work–family
trols (Kurland & Cooper, 2002), such as management by objec-
conflict (Duxbury et al., 1998; Kirchmeyer, 1995; Raghuram &
tives (Konradt, Hertel, & Schmook, 2003), to adapt to the changes
Wiesenfeld, 2004). Boundary flexibility from telecommuting re-
brought on by telecommuting. Behavior-based controls refer to the
fers to the degree to which the location (home vs. central location)
fairly common practice of supervisors evaluating performance on
and timing of work are under the employees’ control (Ashforth et
the basis of employees’ observable actions. Output-based controls
al., 2000). Accordingly, the prediction forwarded below empha-
refer to supervisor’s evaluation of the telecommuter’s performance
sizes the boundary flexibility or control offered by most forms of
on the basis of the evaluation of output, products, or deliverables
telecommuting (see also L. T. Thomas & Ganster, 1995). This
of work, not on the process or behaviors involved in producing the
flexibility can also mitigate the negative effects of permeability by
output. Managers who are unwilling to or who lack the training to
allowing employees to schedule work optimally to minimize in-
change their management and control styles would likely see
terference from family. Further, employees can introduce greater
deterioration in the depth and vitality of their connection with
segmentation at home through the creation of detached home
telecommuting subordinates (Shin, El Sawy, Sheng, & Higa,
office spaces that discourage interruptions from family members.
2000). Altogether, these mechanisms and lines of evidence and
Although this could mean that telecommuters work longer hours
logic lead to the following prediction:
than when at a central location, they may be able to schedule those
hours to mesh with the schedules of family members, reducing
Hypothesis 3a: Telecommuting is negatively related to
time-based conflict. Through the reduction in commuting hours,
telecommuter–supervisor relationship quality.
telecommuting also increases temporal resources that might be
available to family activities (rather than the transition to and from
Face-to-face interactions with coworkers provide access to in-
work; Greenhaus & Beutell, 1985). Therefore, we expect the
formal networks and create opportunities for serendipitous (job-
following:
relevant) interactions. Working in the physical proximity of co-
workers makes strong, positive, and deep ties easier to develop and
Hypothesis 2: Telecommuting is negatively related to work–
maintain (Monge, Rothman, Eisenberg, Miller, & Kirstie, 1985).
family conflict.
By decreasing the frequency of such “high bandwidth” interac-
tions, telecommuting could diminish the richness of a focal em-
Telecommuting researchers have adopted many theoretical per-
ployee’s connection with his or her peers by making it more
spectives to propose that telecommuting leads to detrimental social
difficult to transmit the symbolic and nonverbal, personalized cues
consequences in the workplace (e.g., D. G. Allen et al., 2003, used
of everyday encounters (Rice, 1992). Spatial distance from others
social identity theory; Feldman & Gainey, 1997, used social iso-
at work likely translates into psychological distance; for telecom-
lation theory). However, scholars have most commonly invoked
muters this might mean becoming “out of sight, out of mind”
media richness theory (Daft & Lengel, 1986) and social presence
(McCloskey & Igbaria, 2003). Telecommuters may also face re-
theory (Short et al., 1976) to predict reduced meaningfulness of
sentment and jealousy from coworkers who do not telecommute

---

1528
GAJENDRAN AND HARRISON
(Roberts, 2001). Because coworkers cannot readily see telecom-
Hypothesis 6: Telecommuting is negatively related to turn-
muters’ efforts or their contributions to the unit, they might also
over intent.
perceive telecommuters as parasocial, rather than present and
working, and less instrumental to shared goals (Duxbury &
Telecommuting reduces the stress associated with getting ready
Neufeld, 1999; Lombard & Ditton, 1997).
for and commuting to work; commuting itself is a potential work
role stressor. Further, Pierce and Newstrom (1980) argued that
Hypothesis 3b: Telecommuting is negatively related to
having to arrive at work, especially at a fixed time every day,
telecommuter– coworker relationship quality.
causes distress because lateness has negative reputational conse-
quences at the workplace. Similarly, the flexibility provided by the
Telecommuting’s Effects on Individual Outcomes
most common forms of telecommuting could allow greater partic-
ipation in recreational social or sports activities that could mitigate
Telecommuting provides employees with the choice of working
negative physiological consequences of role stress experienced on
away from a central location. It also leads to reduced costs of
the job (Konradt et al., 2003). Hence, we expect the following:
working, via savings in transportation hassles, time, and money for
employees, and (in many cases) formal business dress is not
Hypothesis 7: Telecommuting is negatively related to em-
required. By providing the opportunity to telecommute, such an
ployee role stress.
arrangement could also symbolize an employer’s willingness to
alter the work environment in response to employees’ needs. This
“Face time,” or visibility, at a central location is thought to be
might also be perceived by employees as reflecting or allowing a
critical for outstanding performance evaluations (O’Mahony &
greater fit between themselves and their job, which is an aspect of
Barley, 1999), and such evaluations are pivotal for career success.
positive work role adjustment (Baltes, Briggs, Huff, Wright, &
Therefore, telecommuters are likely to experience concerns that
Neuman, 1999; Dawis, England, & Lofquist, 1968). Altogether,
working away from a central location could hamper their career
these benefits of telecommuting lead us to assert the following:
prospects. Further, because they are more likely to be out of sight,
out of mind (McCloskey & Igbaria, 2003), they are likely to
Hypothesis 4: Telecommuting is positively related to job
perceive fewer behavioral opportunities to demonstrate high per-
satisfaction.
formance in a face-to-face, highly salient context. They may also
sense that others might view them as less committed and less loyal
Improved productivity is probably the most widely touted ben-
to the organization and as prioritizing personal life over profes-
efit associated with telecommuting (McCloskey & Igbaria, 2003;
sional obligations, which may also contribute to their concerns
Pinsonneault & Boisvert, 2001). A chief explanation for this
about their prospects for advancement (McCloskey & Igbaria,
proposed advantage in performance is that doing tasks remotely
2003). This confluence of factors leads to our next proposition:
also means fewer disruptions while working (Bailey & Kurland,
2002). Telecommuting is also expected to increase productivity
Hypothesis 8: Telecommuting is negatively related to per-
through increased work hours made possible by time saved from
ceived career prospects.
not commuting (Apgar, 1998). Finally, telecommuting provides
individuals the opportunity to tailor or modify the work environ-
ment to better match how and when they do their work most
Telecommuting, Psychological Mediators, and Individual
effectively. Baltes et al. (1999) made a similar argument for
Outcomes
improved performance in the context of flexible work scheduling.
In our earlier sections, we have shown that conceptual argu-
Therefore, we expect the following:
ments form the connection between telecommuting and psycho-
Hypothesis 5: Telecommuting is positively related to job
logical mediators. We have also offered our reasoning above for
performance.
anticipating telecommuting’s effects on individual outcomes. The
remaining connections in our framework are the upshot of well-
Another often-cited advantage of telecommuting, especially in
established evidence linking our proposed psychological mediators
the practitioner literature, has been its presumed capability to help
to those outcomes, and therefore we do not offer them as separate
retain employees who might otherwise have quit (if such a work
hypotheses. In particular, empirical research has consistently dem-
arrangement had not been offered; Kraut, 1987; Pinsonneault &
onstrated that perceived autonomy has beneficial effects on job
Boisvert, 2001). Firms are advised to provide telecommuting to
satisfaction (Loher, Noe, Moeller, & Fitzgerald, 1985) and other
their employees as a form of competitive edge in attracting and
individual outcomes, such as turnover intent, performance, and
keeping the best talent (AT&T, 2004; Vega, 2003). Moreover,
stress, as summarized in a comprehensive meta-analysis (Spector,
Scandura and Lankau (1997) suggested that work arrangements
1986). Other meta-analyses provide robust evidence for the ben-
such as telecommuting increase the value of the psychological
eficial effects of lower work–family conflict on job satisfaction,
contract employees have with their organization. These authors
performance, turnover intent, and stress (T. D. Allen, Herst, Bruck,
proposed that organizations providing employees with the flexi-
& Sutton, 2000; Kossek & Ozeki, 1998). Yet another stream of
bility to work from home are providing a positive signal, visibly
findings from meta-analyses and nationally representative samples
demonstrating their trust and support for employees’ well-being.
suggests that employees’ quality of relationships with supervisors
This signal from organizations should, in turn, generate greater
and coworkers can influence job satisfaction, performance, turn-
psychological commitment and a lowered tendency to quit
over intent, stress, and perceptions of career progress (Gerstner &
(Rhoades & Eisenberger, 2002). Thus, we anticipate the following:
Day, 1997; Kinicki, McKee-Ryan, Schriesheim, & Carson, 2002;

---

CONSEQUENCES OF TELECOMMUTING
1529
Sousa-Poza & Sousa-Poza, 2000; Stanton et al., 2002; Viswesva-
their work and family demands, while office-centered or low-
ran et al., 1999).
intensity telecommuters sought freedom from interruptions. Sim-
Now that we have assembled each of the piecewise elements
ilarly, Wiesenfeld et al. (1999) found that high virtual status
and relations in our framework (Figure 1), one more set of prop-
employees (those who work 3 or more days per week away from
ositions is logically deduced from and virtually defined by its
a central work location, usually home) had different communica-
recursive structure. Perceived autonomy, work–family conflict,
tion patterns relative to low virtual status employees (those who
and relationship quality serve as intervening mechanisms or, at the
work 3 or more days a week at a central location). Coveyduck
least, partial conveyors of the effects of telecommuting onto indi-
(1997), DeLay (1995), Mackie-Lewis (1998), Schneider-Borowicz
vidual outcomes. That is, telecommuting indirectly influences job
(2003), Scott and Timmerman (1999), and Taveras (1998) also
satisfaction, performance, and turnover intent by raising percep-
used similar splits of scheduled work time at work and at home as
tions of control over the location, timing, and means of completing
an indicator of behavioral immersion in telecommuting.
one’s work. Indeed, perceived control has long been theorized as
In our framework, telecommuting intensity moderates the ef-
an antidote to stress (Karasek, 1979; Ganster & Schaubroeck,
fects of telecommuting on the three psychological mediators—
1991). Similarly, by creating opportunities for greater synchroni-
perceived autonomy, work–family conflict, and relationship qual-
zation between demands from the work and family domains,
ity. Moving further into the first conceptual theme in this literature,
telecommuting could indirectly influence individual outcomes,
we propose that high-intensity telecommuters are likely to receive
such as heightened satisfaction and reduced role stress, by lower-
and perceive a greater sense of autonomy relative to those who
ing work–family conflict. However, earlier in our discussion of the
telecommute less frequently. That is, high-intensity telecommuters
telecommuting paradox, we contended that telecommuting is ex-
are likely to experience enhanced discretion over the means of
pected to have negative effects on relationship quality. By poten-
completing their tasks relative to those in low-intensity arrange-
tially harming the quality of an employee’s ties to supervisors and
ments because of lower levels of scrutiny from their supervisors.
coworkers, telecommuting could indirectly have harmful effects
Such perceptions of discretion could also be a result of increased
on individual outcomes of telecommuting. More formally, we
self-reliance from completing one’s tasks by accessing appropriate
therefore offer the following:
virtual technologies (e.g., knowledge management systems, intra-
net databases) instead of relying on relational sources for task
Hypothesis 9: Telecommuting’s beneficial effects on individ-
support, such as supervisors and coworkers (Raghuram et al.,
ual outcomes are mediated by (a) perceived autonomy and (b)
2001), which are relatively less accessible. In addition, high-
work–family conflict, and its detrimental effects are mediated
intensity telecommuters have more opportunities to exercise con-
by (c) relationship quality with the supervisor and (d) rela-
trol over scheduling work for peak productivity times, exercising
tionship quality with coworkers.
greater control over their availability to other organizational mem-
bers, and, therefore, better managing interruptions to their work.
The Moderating Role of Telecommuting Intensity
Finally, at lower intensities, an employee’s identity as a telecom-
muter is likely to be less salient (Bailey & Kurland, 2002), which
By treating telecommuting as a single, undifferentiated pro-
further limits feelings of autonomy with such an arrangement.
gram, the previous hypotheses tend to overlook potentially impor-
Thus, we expect the following:
tant structural distinctions among work arrangements. The chief
structural distinction made by previous investigators deals with
Hypothesis 10: Telecommuting intensity moderates the pos-
what we refer to as telecommuting intensity: the extent or amount
itive impact of telecommuting on perceived autonomy by
of scheduled time that employees spend doing tasks away from a
accentuating its positive effects.
central work location. This idea has been referred under various
guises, as virtual status by Wiesenfeld, Raghuram, and Garud
Earlier, we proposed that a second conceptual theme in the
(1999, p. 782), as virtuality by Scott and Timmerman (1999, p.
telecommuting literature is that it reduces work–family conflict. In
242), and as home-centered versus office-centered telework by
this section, we suggest that the intensity of telecommuting accen-
Konradt et al. (2003, p. 62), among other terms (Hill, Ferris, &
tuates that reduction. The greater boundary flexibility experienced
Martinson, 2003).
by high-intensity telecommuters relative to those in low-intensity
An emerging perspective on telecommuting intensity in the
arrangements translates to a higher degree of potential synchroni-
literature is that when telecommuters spend the majority, versus a
zation between the work and family domains (Ashforth et al.,
minority, of their scheduled time away from a central location, it
2000). High-intensity telecommuters also experience greater sav-
crosses a psychological threshold (in a sense, creating two “tax-
ings of time and energy from reduced commutes and less time
ons,” or classes, of employees in telecommuting arrangements;
spent in transitions to and from work relative to low-intensity
Meehl, 1992). High-intensity telecommuters spend the majority
telecommuters. Reducing conflict between domains, this could
(or all) of their workdays away from a central location. Low-
mean more time and energy available to attend to family respon-
intensity telecommuters spend the majority of their workdays at a
sibilities.
central (conventional) location, working remotely for only 1 or 2
Further, high-intensity telecommuting represents a greater psy-
days a week. Konradt et al. (2003) found that telecommuters who
chological commitment to the unconventional work arrangement.
spent more than 50% of their time away from the office (home
Such a level of commitment possibly reflects a strong desire for
centered) had different motivations for telecommuting relative to
greater integration between the domains of work and family (Kon-
those who spent less than 50% of their time away (office centered).
radt et al., 2003). This greater commitment means that high-
Home-centered or high-intensity telecommuters sought to balance
intensity telecommuters are likely to proactively evolve norms and

---

1530
GAJENDRAN AND HARRISON
routines to manage the increased boundary permeability between
ployed while telecommuting, gender (cf. McCloskey, Igbaria, &
the work and family domains, which could offset the gains due to
Parasuraman, 1998), and experience with telecommuting (cf. Ra-
greater flexibility (Ashforth et al., 2000). For example, over time,
ghuram et al., 2001). Information about the first four of these
high-intensity telecommuters might be more likely to create ded-
moderators in the studies included in this meta-analysis was,
icated home office spaces or make other, relatively permanent
unfortunately, not regularly included in the sample or work ar-
investments in supporting the work arrangement, such as meshing
rangement description, and authors who were directly contacted
child care arrangements with home-working hours. Low-intensity
about their studies could not provide it. Hence, available data
telecommuters might be less effective at managing boundary per-
allowed us only to do an examination of the latter two potential
meability because their work arrangements might be perceived by
moderators: gender and experience with telecommuting.
themselves and family members as too ad hoc or infrequent to
First, gender has been hypothesized to alter the effect of tele-
negotiate major adjustments in boundary management. Taken as a
commuting on autonomy and work–family conflict (Dixon &
whole, these arguments point to the following hypothesis:
Webster, 1998; McCloskey et al., 1998; see also Harrison &
Martocchio, 1998, for arguments about the ubiquity of gender as a
Hypothesis 11: Telecommuting intensity moderates the neg-
moderator of similar work– home transition processes). Second, a
ative impact of telecommuting on work–family conflict by
process-based view of telecommuting suggests that, over time, as
accentuating its negative (beneficial) effects.
individuals gain experience with telecommuting, they begin to
modify the technology and processes of working from a distance to
Our third conceptual theme, relational impoverishment, sug-
have lesser costs and greater benefits (see Walther, 1992, and
gests that higher telecommuting intensity exacerbates the negative
Hinds & Bailey, 2003, for a related discussion).
impacts of telecommuting on quality of interpersonal interactions
with others in one’s workplace. Spending the majority of one’s
work week away from office implies fewer opportunities for rich
Method
face-to-face interactions with supervisors and coworkers, present-
Collecting Effect Sizes
ing fewer opportunities for spontaneous informal interactions
(Mackie-Lewis, 1998). The interactions that do occur are mediated
As the consequences of telecommuting arrangements have in-
mainly by leaner media (Daft & Lengel, 1986), contributing to
terested scholars across many disciplines, we searched electronic
lower social presence (Short et al., 1976) relative to those practic-
databases in management, psychology, information systems, man-
ing low-intensity telecommuting. This leads to sharply diminished
agement science, logistics, engineering, sociology, and education.
perceptions of intimacy and immediacy, which are crucial to
Those databases included ABI/Inform, ProQuest, PsycINFO,
effective interpersonal communication (Lombard & Ditton, 1997).
Elsevier Science Direct, Academic Ideal, EBSCO, ERIC, Socio-
Altogether, this drives a much greater deterioration in the relation-
logical Abstracts, JSTOR, Web of Science, and Dissertation Ab-
ship quality between high-intensity telecommuters and their su-
stracts International. Original studies were found in journal arti-
pervisors or coworkers when compared to that of low-intensity
cles, dissertations, and book chapters. Keywords for our search
telecommuters. Indeed, Kurland and Cooper (2002) found that
included telecommuting, remote work, telework, distributed work,
some employees reduced their telecommuting intensity when they
mobile work, work at home, and flex-place. We also searched the
sensed their relationship with their manager was at risk or that they
Internet and http://scholar.google.com with the same terms to
were targets of coworker gossip and resentment. Therefore, we
locate working papers and unpublished conference papers.
propose the following, countervailing (to those relating to per-
To minimize the potential file drawer problem (Rosenthal,
ceived autonomy and work–family conflict) influences of telecom-
1984), we posted to several listservs and personally contacted
muting intensity. Together, Hypotheses 10, 11, and 12 suggest that
authors who had completed dissertations or had published in this
the potential tension of a telecommuting paradox will be especially
research domain, asking for unpublished or soon-to-be published
sharp for those who spend more of their work time in such an
evidence. In addition, we searched through reference lists of ex-
arrangement.
isting articles. Through this process, we identified approximately
212 works that might have contained effect sizes to contribute to
Hypothesis 12: Telecommuting intensity moderates the neg-
our meta-analysis. However, most of those works lacked data.
ative impact of telecommuting on relationship quality, accen-
They were often instructional rather than empirical (e.g., forward-
tuating its negative effects on interpersonal ties with (a)
ing “how to” advice on setting up remote work arrangements).
supervisors and (b) coworkers.
Inclusion and Exclusion Criteria
Other Moderators
Papers we eventually retained for this meta-analysis fitted sev-
Our hypotheses for telecommuting intensity follow from our
eral inclusion criteria. One criterion was data based and straight-
discussion of the theoretical framework. However, we acknowl-
forward. The paper needed to include an effect size for telecom-
edge that additional moderators might temper or strengthen tele-
muting (e.g., r, t, F, chi-square) or report enough data to compute
commuting’s effects on the psychological mediators. On the basis
one (e.g., means and variances across two groups). A second,
of a search of the telecommuting literature, we identified and
related criterion was that the paper needed to address one of the
attempted to code the following moderators: voluntariness of the
broadly defined, individual-level consequences given in our hy-
telecommuting arrangement, task interdependence, job type of the
potheses. The third criterion was conceptual and more of a meta-
telecommuter (cf. Feldman & Gainey, 1997), technology em-
analytic judgment call (Wanous, Sullivan, & Malinak, 1989). We

---

CONSEQUENCES OF TELECOMMUTING
1531
excluded studies based on (a) employees doing “supplemental
given above (see Baltes et al., 1999, for a similar dichotomization
work” (tasks done after scheduled hours in addition to work done
in the context of alternative scheduling arrangements). Of the 46
at the conventional office), (b) definitions of remote locations that
studies in the meta-analysis, 37 provided information that allowed
equated them with the more common idea of branch offices, (c)
us to code for telecommuting intensity. Of these, 19 studies (51%)
investigation of home-based craft work, and (d) participants who
involved high-intensity telecommuting, and 18 studies (49%) in-
were self-employed or independent contractors. Our fourth and
volved low-intensity telecommuting.
final criterion was that our retained studies examined the phenom-
Other moderators.
We also checked for the effects of two
enon as it was being experienced by employees in organizations
other moderators that were not explicitly mentioned in our hypoth-
(i.e., in the field).
eses: gender and personal experience with telecommuting. We
Data from 46 studies met all of our inclusion criteria and
used the percentage of women in each meta-analyzed sample as a
contributed to a test of at least one of the hypotheses above
study-level index for gender. When possible, we calculated expe-
(highlighted by an asterisk in our reference list). The 46 studies
rience with telecommuting as the time that had elapsed between
included 27 published works and 19 unpublished dissertations. We
introduction of telecommuting in the organization at hand and the
checked whether there were any systematic differences in effect
date of the data collection within a study. Because many studies
sizes based on the rigor of peer review. We assigned the highest
only reported a range (e.g., “more than 12 months”) for this
level of such rigor to journal articles, the lowest to unpublished
measure, we classified studies into two groups. One group in-
dissertations, and an intermediate level to book chapters and con-
cluded studies in which the average telecommuter had a year or
ference papers. We conducted a study-level categorical moderator
less experience with the arrangement; another included studies in
test (Hedges & Olkin, 1985) to test for differences in cumulative
which the average telecommuter had more than a year’s experi-
effect sizes between the three categories of rigor of peer review.
ence.
We did not find any systematic differences (Q
.96, ns).
Dependent constructs.
We followed widely accepted defini-
b
tions for most of our dependent constructs and their construct-label
Final Sample and Coding
synonyms. The constructs constituting the psychological media-
tors in our theoretical framework included perceived autonomy
Each of the 46 studies in our final sample used survey or
(perceived control, job discretion), work–family conflict (work–
interview techniques to measure the proximal and distal outcomes
family conflict and work–family balance, the latter of which was
of telecommuting. The average response rate for those forms of
reversed coded), quality of relationship with supervisor (satisfac-
data elicitation was 51%. The typical telecommuter was a manager
tion with supervision, trust in supervisor, degree of interaction or
(composing an average of 31% of any original sample) or a
communication with one’s supervisor, extent of support from
professional from either the information technology or the sales
one’s boss, and supervisor–subordinate leader–member exchange
and marketing function of a firm (averaging 57% of the functional
quality), and quality of relationship with coworkers (satisfaction
background of a typical sample). The mean age of telecommuters
with coworkers, trust in coworkers, degree of interaction or com-
was 39 years. The average proportion of women in any original
munication with coworkers, support from coworkers, perceived
sample was 49%.
cohesiveness with the work group, and identification with or
Independent construct.
The independent construct—telecom-
commitment to the focal person’s team or work unit). Individual
muting—was typically treated as a dichotomous or categorical
outcomes included job satisfaction (work satisfaction, individual
indicator. That is, in original studies, consequences for individuals
morale), turnover intent (withdrawal cognitions, likelihood of
working under a conventional arrangement were usually compared
changing job, intent to stay; the latter of these was reverse coded),
to consequences for those working under a telecommuting ar-
role stress (tension, role ambiguity, interrole conflict), and per-
rangement. Other studies measured multiple levels of telecommut-
ceived career prospects (satisfaction with promotion, career op-
ing, from zero telecommuting to full-time telecommuting (e.g.,
portunities, chances for promotion). Job performance included
0 –5 days). A study-level categorical moderator test (Hedges &
measures of assignment completion, assessments of productivity,
Olkin, 1985) for dissimilarity in cumulative effect sizes between
and overall evaluations or ratings. We coded separately for self-
the two types of indicators of telecommuting did not reveal any
rated performance and supervisor or objective ratings of perfor-
systematic differences (Q
.04, ns). Therefore, we treated effect
mance.
b
sizes derived from the two types of indicators of telecommuting as
equivalent. Almost all papers used telecommuting to describe the
Meta-Analytic Techniques and Statistical Adjustments
arrangement, so identification of the independent variable was
straightforward.
We used methods described by Hedges and Olkin (1985) to
Hypothesized moderator.
For many of the accumulated stud-
calculate the meta-analytic effect sizes for testing each of our
ies, the average number of days per week spent telecommuting was
hypotheses. We first transformed reported statistics such as means
also reported. We used that average to code our main moderator
and standard deviations, chi-square values, t tests, and F tests into
variable: telecommuting intensity. If the telecommuters in a given
correlations. Prior to calculating the estimated population effect
sample spent the majority of their workdays working remotely—
size, its variance, and confidence intervals (CIs), we used the
2.5 or more days per week—the sample was coded as one involv-
sample-adjusted meta-analytic deviancy statistic, developed by
ing high-intensity telecommuters. Studies reporting an average of
Huffcutt and Arthur (1995), to identify outliers in effect sizes for
fewer than 2.5 days per week working remotely were coded as a
relationships of telecommuting to each of its proposed proximal
sample of low-intensity telecommuters. Conceptual, substantive,
(mediating) or distal outcomes. This resulted in the elimination of
and empirical reasons for using these two classifications were
one study entirely from the meta-analysis because the effect sizes

---

1532
GAJENDRAN AND HARRISON
coded from it were identified as outliers across all the outcomes it
(95% CI
.16 to .22). The Q statistic,
2(10)
15.69, was
involved. Our contact with the authors also revealed that the
nonsignificant, indicating that the variance in this sample of effect
original data were not available for confirming the size of the
sizes was not greater than what would be expected as a result of
effects as unmistaken. Further, for each cumulation involving a
sampling error.
particular mediator or outcome, we did not include a study in that
Our second hypothesis asserted lower work–family conflict as a
particular meta-analytic estimate if its effect size was identified as
proximal consequence (and mediator) of telecommuting. As pre-
a statistical outlier.
dicted, the relationship between telecommuting and work–family
We disattenuated each correlation for the unreliability of the
conflict was negative (ˆ
.13; k
19; n
9,852). It was also
variables involved (as reported in each original study), using
a modest correlation, but the CI for the uncorrected relationship
procedures described by Hedges and Olkin (1985). For studies that
did not include zero (95% CI
.13 to
.10). A significant Q
did not report reliabilities, we imputed an average reliability from
statistic suggested the presence of moderators,
2(18)
51.67,
the other investigations that involved the same construct, an
p
.01.
emerging norm in meta-analyses (e.g., Bhaskar-Shrinivas, Harri-
A subset of original studies that contributed the effect sizes for
son, Shaffer, & Luk, 2005). We calculated meta-analytic estimates
this relationship also examined the work–family consequences of
only when we had at least k
4 independent effect sizes.
telecommuting at a greater level of granularity, investigating bi-
One other adjustment dealt with nonindependence of results—
directional forms of conflict between the work and family do-
conceptual replication within a study—which can bias estimates of
mains. We conducted a post hoc analysis of the effects of work
the true variation in effect sizes across studies and jeopardize tests
interference with family conflict and family interference with work
of moderators (Hunter & Schmidt, 2004). That is, when there were
conflict. Consistent with its overall effect on work–family conflict,
multiple measures of the same dependent construct within a single
we found telecommuting to have a negative effect on work inter-
study and, hence, multiple effect sizes for that construct, we
ference with family conflict (ˆ
.16; k
7; n
1,248; 95%
carried forward the median of the multiple effect sizes. For exam-
CI
.19 to
.08) as well as on family interference with work
ple, if one study reported correlations of telecommuting with (a)
conflict (ˆ
.15; k
6; n
794; 95% CI
.21 to
.07).
satisfaction with supervision, (b) trust in one’s supervisor, and (c)
There was no appreciable difference in the size of these bidirec-
communication frequency with one’s supervisor (all of which we
tional effects.
coded as ways to operationalize relationship quality with one’s
Because of restricted breadth of communication and social
supervisor), we used the median of correlations (a– c) in our
isolation, we predicted that another proximal consequence (psy-
meta-analytic summary tables and tests.
chological mediator) of telecommuting would be damage to inter-
personal relationships at the workplace. Hypothesis 3a dealt with
Results
the proposed negative effect of telecommuting on an employee’s
relationship with his or her supervisor. Contrary to our expecta-
Telecommuting’s Effects on Psychological Mediators:
tions, we found a positive effect of telecommuting on the
Hypotheses 1–3
employee–supervisor relationship. The average corrected correla-
In our first hypothesis, we proposed that telecommuting would
tion was ˆ
.12 (k
14; n
2,888). The meta-analytic 95% CI
be associated with higher perceived autonomy for employees. Our
ranged between .08 and .15. The Q statistic, 2(13)
22.05, was
meta-analytic findings in Table 1 support the idea. Telecommuting
nonsignificant, suggesting that the effect sizes were homogenous
was positively related to this psychological mediator. The average
across the original studies included in this analysis. Hypothesis 3b
correlation corrected for unreliability was ˆ
.22 (k
11; n
predicted that telecommuting would be negatively related to co-
3,040). The CI for the uncorrected correlation did not include zero
worker relationship quality. We did not find support for this
Table 1
Meta-Analytic Relationships of Telecommuting With Proposed Psychological Mediators and Distal Outcomes: Hypotheses 1– 8
Mean
Effect size
Mean 95%
Q statistic
Estimated
Average
Variable
k
n
r
d
CI
2
reliability
Psychological mediator
Perceived autonomy
11
3,040
.19
.39
.16, .22
15.69
.22
.77
Work-family conflict
19
9,852
.11
.23
.13,
.10
51.67**
.13
.84
Relationship quality—supervisor
14
2,888
.12
.23
.08, .15
22.05
.12
.86
Relationship quality—coworker
14
3,269
.00
.00
.03, .03
39.74**
.00
.83
Individual outcome
Job satisfaction
28
7,764
.09
.18
.07, .11
22.92
.10
.83
Performance: self-rated
9
7,419
.01
.02
.01, .03
29.65**
.01
.76
Performance: supervisor-rated or objective measure
4
484
.18
.36
.09, .26
9.79*
.19
.88
Turnover intent
9
7,580
.08
.17
.11,
.06
17.06*
.10
.83
Role stress
11
2,406
.11
.23
.15,
.07
29.08**
.13
.84
Perceived career prospects
8
1,038
.00
.01
.06, .07
32.00**
.00
.83
Note.
CI
confidence interval.
* p
.05.
** p
.01.

---

CONSEQUENCES OF TELECOMMUTING
1533
proposition. Instead, we found the relationship to be very close to
1,038; 95% CI
.06 to .07). The Q statistic was large and
zero (95% CI
.03 to .03; ˆ
.003; k
14; n
3,269). The
significant, 2(7)
32.00, p
.01.
Q statistic was significant, implying the operation of moderators,
2(13)
39.74, p
.01.
Mediating Effects: Hypotheses 9a–9d
Telecommuting’s Effects on Individual Outcomes:
We examined the mediating effects of perceived autonomy,
Hypotheses 4 – 8
work–family conflict, and relationship quality on the individual
outcomes using a series of partial correlations based on the meta-
Hypothesis 4 asserted greater job satisfaction as a (distal) out-
analytic estimates given above about X 3 M and X 3 Y (where X
come of telecommuting. As predicted, the telecommuting–
is telecommuting, M is the proposed psychological mediator, and
satisfaction relationship was positive (ˆ
.10; k
28; n
7,764).
Y is the proposed individual outcome). We used prior meta-
The CI for the uncorrected relationship did not include zero (95%
analytic estimates or correlations from large, nationally represen-
CI
.07 to .11), and the Q statistic was nonsignificant, 2(27)
tative samples to estimate mediator– outcome connections (M 3
22.92.
Y). The perceived autonomy–individual outcome correlations were
Hypothesis 5 stated that telecommuting leads to enhanced job
sourced from Spector (1986); correlations between work–family
performance. Results did not match theoretical and practical ex-
conflict and individual outcomes were taken from T. D. Allen et al.
pectations for self-rated performance, as there was no demonstra-
(2000), and the correlations between relationship quality and in-
ble connection between this variable and telecommuting (ˆ
.01;
dividual outcomes came from Kinicki et al. (2002) and Stanton et
k
9; n
7,419). Its 95% CI for the uncorrected correlation
al. (2002).
included zero (95% CI
.01 to .03). However, telecommuting’s
We computed a partial correlation (corrected for unreliability; ˆ)
relationship with supervisor ratings or archival records of perfor-
for each telecommuting–individual outcome pair after controlling
mance was positive (ˆ
.19; k
4; n
484), and its 95% CI did
for the effect of their respective psychological mediators, given in
not contain zero (95% CI
.09 to .26). The Q statistic was
Hypotheses 9a–9d and shown in Figure 1. Table 2 provides the
significant for both types of indicators of performance—self-rated,
original and partial correlation for each of the hypothesized
2(8)
29.65, p
.01; ratings or records,
2(3)
9.79, p
telecommuting–mediator– outcome mechanisms. We computed
.05—indicating the likely operation of moderators.
the sample size for the significance test of the partial correlations
We predicted in Hypothesis 6 that telecommuting would be