Journal of Vocational Behavior 67 (2005) 215232

www.elsevier.com/locate/jvb

Convergence between measures

of work-to-family and family-to-work conict:
A meta-analytic examination
Jessica R. Mesmer-Magnus*, Chockalingam Viswesvaran
Department of Psychology, College of Arts and Sciences, Florida International University,
Miami, FL 33199, USA
Received 1 December 2003
Available online 12 August 2004

Abstract
The overlap between measures of work-to-family (WFC) and family-to-work conict
(FWC) was meta-analytically investigated. Researchers have assumed WFC and FWC to be
distinct, however, this assumption requires empirical verication. Across 25 independent samples (total N = 9079) the sample size weighted mean observed correlation was .38 and the reliability corrected correlation was .48. The pattern of external correlates for the two types of
conict was also examined. Both forms of conict had similar (.41) reliability corrected correlations with measures of organizational withdrawal. WFC conict correlated .41 (k = 15,
N = 4714) with job stressors and .17 (k = 13, N = 3312) with non-work stressors whereas
FWC conict correlated .27 with job stressors and .23 with non-work stressors. Correlations
between the two forms of work/family conict and other variables such as organizational commitment, job and life satisfaction, and health were examined. Implications for the discriminant
validity of the two types of conict measures are discussed.
2004 Elsevier Inc. All rights reserved.
Keywords: Work/family conict; Scales; Work stressors; Non-work stressors; Organizational commitment;
Job satisfaction; Life satisfaction; Organizational withdrawal; Health; Discriminant validity

Corresponding author. Fax: 1-305-348-3879.

E-mail address: jmesm001@u.edu (J.R. Mesmer-Magnus).

0001-8791/$ - see front matter 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jvb.2004.05.004

216

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

1. Introduction
Given the increasing prevalence of dual-breadwinner families and single working
parents, the challenges workers face in meeting demands of the work and family domain have become numerous and progressively more complex (Duxbury & Higgins,
1991). It is not surprising that the demands of work and family are not always compatible, leading to feelings of conict between these domains. The potential for negative eects of work/family conict have spurred research in this area.
Researchers have found that workers facing high levels of work/family conict are
at increased physical and mental health risk, have less satisfactory job performance,
poorer parental performance, more incidences of work withdrawal behaviors (e.g.,
tardiness, absenteeism, turnover, and low job involvement), decreased morale, and
lower satisfaction with job, life, marriage, and family (Duxbury & Higgins, 1991;
Frone, Russell, & Cooper, 1992; Hammer, Bauer, & Grandey, 2003; Leiter & Durup,
1996; ODriscoll, Ilgen, & Hildreth, 1992).
Early research has employed measures of work/family conict that did not distinguish between the direction of conict. That is, distinctions were not made between
conict caused by work roles interfering with family from that of family roles interfering with work. Work and family are considered to be distinct spheres or domains
of a persons life. As such, requirements in the work domain that impede performance in the family domain (work-to-family conict; WFC) and family demands
that impede performance in the work domain (family-to-work conict; FWC) are
conceptually distinct (Netemeyer, Boles, & McMurrian, 1996). To better understand
the antecedents and consequences of these two forms of conict, researchers have begun measuring levels of work/family conict using this bi-directional conceptualization (Frone et al., 1992).
But to what extent is this conceptual clarity reected in the measures employed to
assess these two forms of work/family conict (WFC and FWC)? If individuals who
score highly on one type of conict measure (e.g., WFC) also obtain high scores in
measures of conict in the other direction (i.e., FWC), what are the implications for
research and practice? To some extent an overlap is expected (Frone et al., 1992).
Because each individual has xed amounts of physical and psychological resources
(e.g., time, mental energy, etc.), conicts in one direction are likely to be coupled
with expressions of conict in the other direction. Further, to the extent there are
stable individual dierences in perceptions of role conict, a positive overlap between measures of WFC and FWC is expected. The question is whether the overlap
is so substantial that empirical distinction between the two directions of conict becomes problematic. Individual studies have reported correlations between WFC and
FWC that vary widely across samples [e.g., r = .10, p > .05 reported by Gutek,
Searle, and Klepa (1991); compared with r = .59, p < .05 reported by Flye, Agars,
and Kottke (2003)]. Researchers have proceeded as though the two types of conict
are distinct and no one needs to be convinced of their distinctiveness. However, such
an assumption needs to be empirically veried. Thus, our rst objective was to
meta-analytically cumulate the correlations between measures of WFC and measures of FWC.

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

217

A high correlation between measures of WFC and measures of FWC does not
necessarily void the separate utility of the two constructs. Nor does the high correlation imply that the two are conceptually similar. For instance, verbal ability and
numeric ability are highly correlated but are conceptually distinct. Dierent patterns
of correlations are hypothesized and empirically supported for verbal and numeric
ability. Thus, high intercorrelation between two constructs but with dierent patterns of correlations with external variables merely suggests the presence of an overarching construct. In this investigation, it is possible (and even likely) that WFC and
FWC will have a high intercorrelation, suggesting an overarching general factor, but
have dierent patterns of correlations with external variables.
Previous research has hinted at the possibility of such a dierential pattern of
correlations for WFC and FWC. For example, one could hypothesize that WFC
will correlate higher with job stressors and job satisfaction than FWC, but FWC
will correlate higher (than WFC) with life satisfaction and non-job stressors. Kossek and Ozeki (1998) found that WFC correlated more with job satisfaction than
FWC. Judge, Boudreau, and Bretz (1994), however, reported that both WFC and
FWC had similar correlations with job satisfaction. A meta-analytic investigation
of the pattern of correlates for both WFC and FWC for several variables will illuminate the potential for discriminant validity for the two types of conict measures. Specically, we investigate whether one type of conict (e.g., FWC)
contributes incremental variance to explaining certain outcome variables over
the other (i.e., WFC). By investigating which outcome variables WFC explains
over and above FWC as well as which outcome variables FWC explains over
and above WFC, this paper contributes to the extant literature in FWC and
WFC.
1.1. Theoretical overview of WFC and FWC constructs
Greenhaus and Beutell (1985) dened work/family conict as a form of inter-role
conict in which the role pressures from the work and family domains are mutually
non-compatible in some respect. That is, participation in the work (family) role is
made more dicult by virtue of participation in the family role (p. 76). The general
demands of each role include the responsibilities, requirements, duties, commitments, and expectations related to performance in a given domain (Netemeyer
et al., 1996). The nite resources required to fulll such role demands are frequently
in a state of imbalance, leading to feelings of conict between domains. Exposure to
stressors in one domain may lead to petulance, fatigue, and/or preoccupation with
those problems, further restricting ones ability to adequately perform role functions
in the other domain (Greenhaus & Beutell, 1985).
1.2. Measures of WFC and FWC
The relationship between WFC and FWC could be moderated by the measure of
conict employed in a given study. Since Greenhaus and Beutells (1985) article
urging the examination of work/family conict from a bi-directional perspective,

218

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

researchers have voiced concern that the lack of psychometrically sound measures of
work/family conict reected poorly upon the validity of their ndings (Laster,
2002). Further, Kossek and Ozeki (1998) speculated that the variation within and
across these measures likely explains the widespread discrepancies in research results.
In fact, of the 25 independent studies included in the present meta-analysis, researchers have reported internal consistency reliabilities ranging from .56 to .95, and have
utilized measures that vary in length from 2 to 22 items. Indeed, wide variability in
item content, scale length, and internal consistency will likely yield varying research
outcomes and discrepant, even contradictory conclusions.
1.3. Incremental variance of WFC and FWC measures
Widely accepted is the notion that measuring both forms of work/family conict
(WFC and FWC) will provide researchers with increased information regarding constructs of interest. However, to what extent does measuring one form of work/family
conict add incremental variance over and above the other? Does incremental variance of WFC or FWC dier depending on the constructs to which they are related?
For example, researchers have suggested that worker levels of FWC are more
impacted by variables within the family domain, while WFC is more inuenced
by work-related variables (e.g., Kossek & Ozeki, 1998). Therefore, when exploring
the eects of work/family conict within these domains, is it necessary to measure
both forms of conict, or will one suce? An examination of partial correlations
of WFC and FWC with frequently assessed external correlates will clarify this
issue.
A review of the literature reveals eight categories of external variables that are
generally examined in studies of WFC and FWC: (a) job-related stressors, (b) level
of support received from work environment, (c) organizational attachment and commitment, (d) behaviors indicative of organizational withdrawal, (e) job/career satisfaction, (f) life satisfaction, (g) physical and mental health, and (h) other non-work
or family-related inuences.
Job-related stressors include workrole overload, workrole ambiguity or conict, job tension, job stress or distress, lack of autonomy in performing work functions, schedule inexibility or unpredictability, etc. Signicant positive relationships
have been reported between the presence of job stressors in the workplace (e.g.,
workrole overload, role ambiguity, non-participation in the work role, and heavy
responsibilities) and reports of WFC (Eagle, Miles, & Icenogle, 1997). This positive
correlation is believed to reect the nite physical and psychological resources of the
worker. Parasuraman, Purohit, and Godshalk (1996) found that increased autonomy
in accomplishing work requirements resulted in decreased work/family conict, since
the worker has the exibility to handle work and family demands without having to
contend with restrictive schedules or micromanagement.
Level of support received from the work environment is reected in: (a) presence
of company-sponsored work/family policies and programs, (b) social support
provided by supervisors and co-workers, (c) work culture conducive to dealing with
conicting family demands, and (d) good quality of interpersonal relations with

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

219

co-workers. There is some evidence that the support received from the work environment is related more to measures of WFC than to FWC (e.g., Frone et al., 1992).
Organizational commitment is another correlate of both FWC and WFC. Highly
committed workers are also more likely to be involved in their work roles. They are
more likely to report FWC than WFC. Thus, a stronger relationship is expected between FWC and organizational commitment than between WFC and commitment.
Both WFC and FWC are predictive of tardiness, absenteeism, family-related interruptions at work, and intent to leave an organization (Go, Mount, & Jamison,
1990; Hammer et al., 2003; Kossek & Ozeki, 1999). However, FWC seems to be
more to blame for worker absence than WFC because workers are more likely to
miss work to contend with family-related role-demands (a source of FWC), than
for work-related reasons (Kossek & Ozeki, 1999).
High levels of both forms of work/family conict (work-to-family and family-towork) have been associated with decreased levels of job satisfaction (Adams, King, &
King, 1996; Duxbury & Higgins, 1991; Thompson & Blau, 1993; Wiley, 1987), and
similar correlations have been reported for life satisfaction (Judge et al., 1994). Life
satisfaction is increased through involvement in activities within the family domain
and receipt of emotional support from family members (Judge et al., 1994; Wiley,
1987). Therefore, when work interferes with family (WFC), life satisfaction is negatively impacted. FWC is also thought to impact negatively on levels of life satisfaction, as the work role is an arguably large component of a persons life (Wiley, 1987).
Measures of both WFC and FWC have been related to physical and mental
health (Judge et al., 1994). WFC and FWC have been linked with incidence of clinical depression and work distress (Frone et al., 1992), emotional exhaustion (Leiter &
Durup, 1996), and experience of burnout (Kossek & Ozeki, 1999). Finally, we also
examined the relationship of WFC and FWC with non-work or family-related
stressors.
In summary, we expect to nd evidence to support discriminant validity between
measures of WFC and FWC by observing dierential correlations with certain external variables. Specically, we hypothesize that WFC will correlate higher with factors most relevant to the work domain (i.e., job-related stressors, supportive work
environment, organizational attachment, and withdrawal behaviors) while FWC will
correlate higher with factors most related to the home domain (i.e., non-work or
family stressors and health). Finally, based on past research ndings (e.g., Frone
et al., 1992) and given the relative importance of the work and home domains to
worker evaluations of job and life satisfaction, we expect that both WFC and
FWC will have similar levels of correlations with these variables.

2. Method
2.1. Database
Twenty-ve independent studies reported in 20 articles examining WFC and
FWC were included in this meta-analysis. To ensure a comprehensive search, these

220

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

studies were located using the following strategies: (1) conducting a computerized
search of the PsycInfo (1887 to present) and ABI Inform (1971 to present) databases
using appropriate keywords and phrases (i.e., work AND family, work-to-family
conict OR workfamily conict, AND family-to work conict OR familywork conict, worknon-work conict, workfamily interference, job interference AND o-job
interference, and WFC AND FWC), (2) conducting a manual search of references
cited in studies included in this meta-analysis and those cited in recently published
reviews (e.g., Frone, 2003; Kossek & Ozeki, 1998, 1999), and (3) obtaining related
studies from the recent conference presentations (i.e., Society of Industrial Organizational Psychology, April 2003). Our objective in examining studies from the recent
conference was to incorporate research results that had not yet been accepted into
journals.
Studies were included only if a measure of both WFC and FWC was administered, a correlation between these two constructs stated, and internal reliability estimates for each measure reported. Studies were omitted if they included only
qualitative data, administered only a global measure of work/family conict, measured one type of work/family conict (WFC or FWC), or dealt with constructs that
diered from those examined in this research. When authors reported separate correlations for dierent subgroups (e.g., males and females), samples, administrations
(as in a longitudinal study), or measures of the same construct, those correlations
were examined separately. While no specic cut-o dates were observed when selecting studies for inclusion, the rst study we found to employ both measures of WFC
and FWC, and report a correlation between them did not appear until 1987. The
studies included in this meta-analysis, along with sample characteristics and the
WFC and FWC measures used, are presented in Table 1, as well as listed in the references prexed with an asterisk.
2.2. Coding procedure
The rst author undertook an independent eort to code the 25 studies that
met the criteria for inclusion in the analysis. A random subset of the articles
was coded by the second author in an eort to determine coder reliability. Inter-coder agreement was very high (100%), likely due to the objective nature of
the information coded. Data coded included study sample size, sample characteristics, type of WFC and FWC measure utilized, reported correlation between
WFC and FWC, reported reliability coecients for WFC and FWC measures,
correlations between WFC/FWC and relevant external correlates (job stressors,
non-work stressors, supportive work environment, organizational attachment, organizational withdrawal behaviors, job satisfaction, life satisfaction, and health),
and reliability coecients of these measures. Specic coding for the quality of
studies was not conducted. However, all studies included in this meta-analysis underwent some form of peer-review process prior to selection, and corrections were
made for unreliability in the measures that were included in these studies. As
such, the potential for results of poor quality studies to inuence our ndings
is minimal.

Table 1
Studies included in meta-analysis
WFC/FWC

Sample characteristics

WFC measure

FWC measure

Adams et al. (1996)

.30

Kopelman et al. (1983, 4 items)

Burley (1989, 4 items)

163

Boyar et al. (2003)

Brewster et al. (2003)

.42
.53

Full-time workers living

together
University employees
Full-time workers,
mostly male and married

Own (3 items)
Carlson, Kacmar, and Williams
(2000; 9 items)

698
142

Eagle et al. (1997)

.42

University employees,
mostly married

Frone et al. (1992, 2 items);

Kopelman et al.
(1983, 4 items); Burke et al. (1979)

Flye et al. (2003)

.59

Full-time workers

Netemeyer et al. (1996, 5 items)

Frone et al. (1992)

.33

Own (2 items)

Frone and Yardley (1996)

.44

Full-time workers,
married
Full-time workers,
mostly married

Own (3 items)
Carlson, Kacmar,
and Williams
(2000; 9 items)
Frone et al. (1992, 2
items); Burley (1989,
4 items); Burke et al.
(1979)
Netemeyer et al.
(1996, 5 items)
Own (2 items)

Frone et al. (1997)

.45

Full-time workers,
mostly married

Grzywacz and Bass (2003)

Gutek et al. (1991)

.50
.20

Gutek et al. (1991)

.10

Hammer et al. (2003)

.33

General population
Psychologists with
at-home family members
Sr. Managers, mostly
married
Full-time workers,
married, all male

Judge et al. (1994)

Laster (2002)

.27
.54

Male executives
Full-time workers,
mostly married

Frone et al. (1992, 2 items) and

Kopelman et al.
(1983, 4 items)
Frone et al. (1992, 2 items) and
Kopelman et al.
(1983, 4 items)
Own (4 items)
Kopelman et al. (1983, 4 items)
Kopelman et al. (1983, 4 items)
Kopelman et al. (1983, 4 items) and
Go et al. (1990, 8 items)
Kopelman et al. (1983, 4 items)
Carlson, Kacmar, and Williams
(2000; 9 items)

393

313
631

221

Frone et al. (1992, 2

252
items) and Burley
(1989, 4 items)
Frone et al. (1992, 2
372
items) and Burley
(1989, 4 items)
Own (4 items)
1,986
Burley (1989,
423
4 items)
Burley (1989,
209
4 items)
Burley (1989, 4 items)
359
and Go et al. (1990,
8 items)
Burley (1989, 4 items)
1062
Carlson, Kacmar, and
221
Williams (2000; 9 items)
(continued on next page)

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

Study

222

WFC/FWC

Sample characteristics

WFC measure

FWC measure

Leiter and Durup (1996)

.26

Kopelman et al. (1983, 4 items)

Burley (1989, 4 items)

151

Netemeyer et al. (1996)

.48

Netemeyer et al. (1996, 22 items)

Netemeyer et al.
(1996, 21 items)

182

Netemeyer et al. (1996)

.33

Netemeyer et al. (1996, 22 items)

.42

ODriscoll et al. (1992)

Parasuraman et al. (1996)

.37
.16

Thompson and Blau (1993)

.40

Wiley (1987)

.40

Williams and Alliger (1994)

.36

Netemeyer et al.
(1996, 21 items)
Netemeyer et al.
(1996, 21 items)
Own (7 items)
Kopelman et al.
(1983, 4 items)
Burke et al.
(1979, 7 items)
Burke et al.
(1979, 5 items)
Burley (1989, 4 items)

162

Netemeyer et al. (1996)

Female hospital
workers, mostly full-time
School teachers, mostly
women and mostly
married
Small business owners,
mostly married
Real-estate salespeople,
mostly married
Full-time workers
Business owners
Full-time workers,
mostly married
Employed MBA
students
Working parents

Netemeyer et al. (1996, 22 items)

Own (7 items)
Kopelman et al. (1983, 6 items)
Burke et al. (1979, 7 items)
Burke et al. (1979, 5 items)
Kopelman et al. (1983, 4 items)

186
120
111
234
191
159

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

Table 1 (continued)
Study

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

223

2.3. Analysis
The meta-analytic methods outlined by Hunter and Schmidt (1990) were employed. When a study employed more than two scales for either WFC or
FWC, our rst choice was to compute the composite correlation across the dierent measures. The composite correlation provides a better estimate of the relationship than either the component or the average (of the components) correlation
(Hunter & Schmidt, 1990; Nunnally, 1978; Schmidt, 1971). Linear composites
with unit weights were formed as unit weights have been found to be comparable
to regression-based weights or weights based on factor loadings on the common
factor (Harman, 1976). To compute composite correlations, one would need the
intercorrelations across the measures. Fortunately, these correlations were available in the studies included here, and as such the composite correlations were
computed and included in the meta-analysis. However, when examining the moderating eects of the measure used, we coded the corresponding correlations separately.
Since all studies reported reliabilities for their measures, individual correction of
the correlations was possible. No recourse was made to artifact distributions. Corrections were made for unreliability in the two measures and coecient as were used.
No corrections for range restrictions or other artifacts were possible. The sample size
weighted mean observed correlations as well as the reliability-corrected mean correlations were computed. The associated standard deviations were examined for potential moderating inuences as well as used in computing the 90% credibility interval
around the reliability-corrected (true) correlation. Interested readers can also compute a condence interval around the sample-weighted mean observed correlations
using the information provided.
2.3.1. External correlates
WFC and FWC were examined with relation to eight commonly measured external correlates. Variables reported by researchers were grouped into the following
categories: (1) job-related stressors (includes autonomy (reverse coded), schedule inexibility, workrole overload, workrole conict, job stress, job distress, work
role ambiguity, and job tension), (2) supportiveness of work environment (includes
perceived workfamily culture, workfamily policies, supervisor support, and coworker support), (3) organizational attachment (includes organizational attachment, organizational commitment, job involvement, and time commitment to
work), (4) organizational withdrawal (includes intent to leave, intent to search
for a new job, absenteeism, tardiness, and turnover), (5) job/career satisfaction,
(6) life satisfaction, (7) physical and mental health (includes burnout, physical
symptoms of stress, depression, dysphoric mood, and health (reverse coded)),
and (8) non-work or family-related stressors (includes parental demands, family/
friend support (reverse coded), family personal conict, emotional or instrumental
assistance (reverse coded), family-role conict, family-role ambiguity, family overload, family involvement, family satisfaction, time-commitment to family, and perceived family demand).

224

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

2.3.2. Incremental variance of WFC and FWC

In an eort to determine whether measurement of WFC or FWC adds incremental variance over and above the other in explaining relationships to other variables,
we computed partial correlations between each form of conict with each of the eight
frequently explored external correlates. The variance reduction ratio (Chen & Spector, 1991) was also computed to assess how much the variance explained in the outcome variable by one form of conict decreased due to partialing out the other forms
of conict.

3. Results
The results of the meta-analysis of the correlations between measures of WFC and
FWC are summarized in Table 2. The meta-analytic results of the correlations between WFC and external correlates are presented in Table 3; the corresponding correlations for FWC are provided in Table 4. In each of these three tables, the number
of independent samples included in the meta-analysis (k), the total sample size across
the samples (N), the sample size weighted mean observed correlation (r), the sample
size weighted standard deviation associated with the observed mean (SDr), the reliability corrected correlation (q), the standard deviation associated with the q and the
90% credibility interval around q, percent variance due to sampling error (%SEV),
and the percent variance due to all statistical artifacts (%ARTV) are reported. In addition, a le drawer analysis (File Drawer k) was conducted based on reliability corrected mean correlations (see Hunter & Schmidt, 1990, p. 513). This value represents
the number of missing studies averaging null results that would be needed to reduce
the observed reliability corrected correlation to a specied level (we used a level of
.10 in computing this value).

Table 2
Correlation between work-to-family and family-to-work conict
Measure

SDr

Overall
25 9079 .38 .1205
Burke et al. (1979)
4 818 .40 .0345
Frone et al. (1992)
3 1024 .39 .0721
Kopelman et al. (1983)/ 11 3389 .26 .0845
Burley (1989)
Netemeyer et al. (1996)
4 843 .48 .0980

SDq

90% CI File
%SEV %ARTV
Drawer k

.48
.50
.54
.33

.1277
0
0
.0778

.27.69
.50.50
.54.54
.20.46

95
16
13
26

13.84
100
40.91
39.70

15.84
100
100
42.64

.54 .0782 .41.67

18

29.48

36.99

Note. k = number of correlations meta-analyzed; N = total sample size across the correlations meta-analyzed; r = sample size weighted mean observed correlation; SDr = sample size weighted standard deviation of the correlations; q = sample size weighted mean observed correlation corrected for unreliability in
both measures; SDq = standard deviation of q; 90% CI = 90 percent credibility interval computed as
q = 1.645 (SDq); File Drawer k = the number of studies averaging null results to reduce the reliabilitycorrected correlation to .10; %SEV = percent variance due to sampling error; %ARTV = percent variance
due to all statistical artifacts.

Meta-analysis

Job stressors
Non-work stressors
Supportive work environment
Organizational attachment
Organizational withdrawal behaviors
Job satisfaction
Life satisfaction
Health

15
13
5
10
6
8
7
9

4714
3312
2596
2229
1561
2208
2049
4701

SDr
.34
.14
.14
.06
.18
.12
.26
.23

.1462
.0589
.0017
.1488
.0700
.1033
.1103
.1077

q
.41
.17
.16
.06
.20
.14
.31
.26

SDq

90% CI

File Drawer k

%SEV

%ARTV

.1642
0
.0235
.1637
.0343
.0985
.1068
.1177

.14 to .68
.17 to .17
.13 to .20
.21 to .33
.14 to .26
.02 to .30
.13 to .49
.07 to .46

47
9
3
0
6
3
15
15

11.74
100
100
20.24
73.70
33.27
24.56
14.92

11.84
100
82.57
19.67
78.48
34.01
26.75
14.49

Note. k = number of correlations meta-analyzed; N = total sample size across the correlations meta-analyzed; r = sample size weighted mean observed
correlation; SDr = sample size weighted standard deviation of the correlations; q = sample size weighted mean observed correlation corrected for unreliability
in both measures; SDq = standard deviation of q; 90% CI = 90 percent credibility interval computed as q = 1.645 (SDq); File Drawer k = the number of
studies averaging null results to reduce the reliability-corrected correlation to .10; %SEV = percent variance due to sampling error; %ARTV = percent
variance due to all statistical artifacts.

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

Table 3
Correlations between work-to-family conict and outcomes

225

226

Meta-analysis

Job stressors
Non-work stressors
Supportive work environment
Organizational attachment
Organizational withdrawal behaviors
Job satisfaction
Life satisfaction
Health

15
13
5
10
6
8
7
9

4714
3312
2596
2229
1561
2208
2049
4709

SDr
.21
.18
.02
.01
.18
.14
.20
.23

.1202
.1056
.0479
.1305
.0654
.0855
.1087
.1115

q
.27
.23
.02
.00
.20
.18
.25
.27

SDq

90% CI

File Drawer k

%SEV

%ARTV

.1407
.1073
.0255
.1439
.0242
.0732
.0167
.1395

.03 to .51
.05 to .40
.07 to .02
.23 to .24
.16 to .24
.06 to .30
.08 to .43
.04 to .50

26
17
0
0
6
7
11
15

20.16
33.16
84.06
26.49
84.69
47.84
26.73
13.83

18.94
33.68
81.94
26.44
86.32
49.89
29.34
11.35

Note. k = number of correlations meta-analyzed; N = total sample size across the correlations meta-analyzed; r = sample size weighted mean observed
correlation; SDr = sample size weighted standard deviation of the correlations; q = sample size weighted mean observed correlation corrected for unreliability
in both measures; SDq = standard deviation of q; 90% CI = 90 percent credibility interval computed as q=1.645 (SDq); File Drawer k = the number of
studies averaging null results to reduce the reliability-corrected correlation to .10; %SEV = percent variance due to sampling error; and %ARTV = percent
variance due to all statistical artifacts.

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

Table 4
Correlations between family-to-work conict and outcomes

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

227

Across 25 independent samples, the sample size weighted mean observed correlation between measures of WFC and FWC was .38. When corrected for unreliability
in the two conict measures, the correlation was .48. The 90% credibility intervals
ranged from .27 to .69 suggesting that the association between WFC and FWC varied across the samples included.
One obvious concern is whether some of the measures have better discriminant
validity in assessing these two types of conicts. Four measures were used in more
than three samples. The Kopelman, Greenhaus, and Connolly (1983) WFC measure
and Burleys (1989) FWC measure (which was created specically to complement the
Kopelman et al. measure) were used in concert in 11 samples and were the most
widely used measures across the studies included in the meta-analytic database.
The WFCFWC correlation was .33 across these 11 samples. In fact, the Kopelman
et al./Burley measures displayed the least overlap across the four measures investigated here. The Frone et al. (1992) and the Netemeyer et al. (1996) measures, though
more recent conceptualizations than the Kopelman et al. (1983) measure, showed
greater overlap between the two forms of conict. These ndings are more intriguing
since the later measures (Frone et al., 1992; Netemeyer et al., 1996) were presumably
developed to better distinguish between the two types of conict. Eect size estimates
for the relationship between WFC and FWC ranged from .26 to. 48.
Tables 3 and 4 summarize the correlations between WFC and FWC with external
correlates. Job satisfaction correlated .14 with WFC and .18 (k = 8, N = 2208)
with FWC. Contrary to our expectation that FWC will demonstrate a stronger relationship with withdrawal behaviors than WFC, we found identical correlations between FWC and WFC with withdrawal behaviors. As expected, the strength of
the relationship between WFC and supportive work environment was much stronger
(.16) than it was between FWC and supportive work environment (.02). Both WFC
and FWC had a weak relationship with organizational commitment.
Table 5 summarizes partial correlations between WFC and FWC with six of the
eight frequently measured external correlates. Two of the eight external correlates
(supportive work environment and organizational attachment) were eliminated from
the analyses due to low initial correlations with WFC ( .14 and .06, respectively)
and FWC ( .02 and .01, respectively). As expected, removing the eects of WFC

Table 5
Partial correlations of work-to-family conict and family-to-work conict with external correlates
External correlates
Job stressors
Non-work stressors
Organizational withdrawal
Job satisfaction
Life satisfaction
Health

FWC
.21
.18
.18
.14
.20
.23

r13.2
.09
.14
.12
.10
.11
.16

VRR
94
40
78
46
68
53

WFC
.34
.14
.18
.12
.26
.23

r23.1
.29
.08
.12
.07
.20
.14

VRR
29
68
78
63
39
53

Note. r1 = FWC, r2 = WFC, and r3 = applicable external correlate. r13.2 indicates the correlation between
FWC and external correlate with WFC partialed out. r23.1 indicates the correlation between WFC and
external correlates with FWC partialed out. VRR indicates the percent reduction in variance explained.

228

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

from the correlation of FWC with job stressors results in a large reduction of variance explained (94%), as WFC is thought to be more related to variables in the work
domain than is FWC. Similarly, removing the eects of FWC from the correlation of
WFC with non-work stressors also results in a large (i.e., 68%) reduction in variance
explained. However, contrary to expectations, removing the eects of FWC from the
correlation between WFC and job satisfaction results in a larger reduction in explained variance (63%) than did removing WFC from the correlation between
FWC and job satisfaction (46%). Given the conceptual link between job satisfaction
and the work domain variables, we would have expected to lose more variance by
removing WFC from the correlation between FWC and job satisfaction. Similarly,
given the conceptual link between life satisfaction and non-work variables, we would
have expected to nd a larger reduction in variance in its correlation with WFC
when FWC was removed. However, contrary to our expectations, removing FWC
from the correlation between WFC and life satisfaction resulted in less loss of variance (only 39%) than when WFC was removed from the correlation between FWC
and life satisfaction (68%). The relationship between WFC and FWC and organizational withdrawal and health factors was equivalent. Therefore, removal of one form
of conict from correlations with these variables and the other form of conict
resulted in the loss of an equal, large portion of explained variance (78 and 53%,
respectively).

4. Discussion
It is generally assumed that WFC and FWC are distinct forms of work/family
conict, as they originate from arguably separate life domains. The possibility these
variables are part of a reciprocal relationship is frequently voiced in the literature.
The results summarized here suggest that this is not the casedespite some overlap,
the two measures have sucient unique variance to warrant independent examination. Further, the dierential patterns of correlations of these two types of conict
with external correlates, along with the partial correlations of WFC/FWC with these
variables, were relatively consistent with our hypotheses, thus providing preliminary
support for discriminant validity between WFC and FWC. While it is true that the
credibility intervals for some external correlates overlap when comparing between
WFC and FWC, the small number of studies that were available for conducting
these moderator analyses voids the potential for strict dependence on credibility intervals when interpreting results. Therefore, the conclusion that these two types of
conict possess discriminant validity appears to be credible.
More interesting was the nding that the overlap between WFC and FWC measures was higher for the Frone et al. (1992) and Netemeyer et al. (1996) measures
than it was when WFC was measured using the Kopelman et al. (1983) scale and
FWC was measured using the corresponding Burley (1989) scale. The later day measures were presumably developed specically with the intent to dierentiate between
the two types of conict, and as such one would expect the correlation to be lower.
Of course, since we do not know the true relationship between the two types of

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

229

conict, one cannot conclude that one measure is a better measure (or more construct valid) based on these results. In this area, as in most areas of investigations
in psychological sciences, the construct validity of a measure has to be inferred from
an extended nomological network of related variables. All that can be inferred from
this nding is that if a researcher is interested in choosing a measure of FWC and
WFC that will minimize the overlap between the two, the WFC measure by Kopelman et al. (1983) used in conjunction with the FWC measure by Burley (1989) appear to be promising. (These measures have been published together and in their
entirety in Gutek et al., 1991.)
Future research should employ separate WFC and FWC measures in the same
study so that more focused factor analyses can be conducted to rene the constructs
of WFC and FWC as well as their sub-dimensions. Such renement will facilitate
theoretical advancements in the spillover and compartmentalization literatures. Researchers may also be wise to consider the role of individual dierences in the interpretation of WFC or FWC. Specically, widely used measures solicit only a global
assessment of felt WFC and FWC (e.g., using phraseology like how frequently
does. . . or in general. . .), rather than a workers evaluations or interpretations
of isolated events. By clarifying the dierences between workers in interpretations
of work/family conict, researchers may be better equipped to identify and explain
its antecedents, consequences, and correlates. In addition, given the implicit assumption of bi-directional inuences of work/family conict, longitudinal research is
needed to map out how the two types of conict develop over time. Unfortunately,
across the 25 studies included in this meta-analysis only one employed a longitudinal
design. Further, their results suggested a likely uctuation in levels of WFC and
FWC over time. Certainly this uctuation would have important implications for research and practice. More important than employing a longitudinal design, models
have to be developed that incorporate the role of time in the relationship between
WFC and FWC. As Mitchell and James (2001) note, we need a better specication
when eects manifest in organizations.
The two types of work/family conict had similar but distinct correlations with
eight external variables. Both WFC and FWC had similar correlations with behaviors indicative of organizational withdrawal. This result is consistent with Go et
al.s (1990) nding that both WFC and FWC are predictive of tardiness, absenteeism, family-related interruptions at work, and intent to leave the organization. Past
research has suggested that a supportive work environment (e.g., provision of company-sponsored family-friendly policies) is eective in diminishing feelings of work/
family conict and in decreasing incidence of worker withdrawal behaviors (Flye et
al., 2003; Kossek & Ozeki, 1998). Interestingly, while WFC was negatively correlated
with supportive work environment ( .16), this relationship was not as high as would
have been predicted based on past research ndings (e.g., Flye et al., 2003). Certainly, this result may be a product of the method by which variables of this type were
grouped for examination in this study. Perhaps examining each variable (i.e., perceived workfamily culture, workfamily policies, supervisor support, and co-worker support) separately would yield dierent outcomes. However, given the enormous
expense of workfamily policies to organizations, it is of importance to further

230

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

examine the impact of each of these programs on the reduction in worker feelings of
work/family conict.
Researchers widely agree that stressors associated with the work role (e.g., work
role overload, role ambiguity, and heavy responsibilities) are more likely to increase
feelings of WFC than FWC, and that stressors associated with the family role (e.g.,
family-role demands, ambiguity, and/or overload) are more likely to impact feelings
of FWC than WFC. Therefore, it was not surprising to nd that WFC correlated
more highly than FWC with job stressors (.41 vs. .27, respectively) and FWC correlated higher with non-work stressors than WFC (.23 vs. .17, respectively). These patterns of correlations further lend support to the need to distinguish between the two
types of conicts.
Interestingly, our analysis of partial correlations between WFC and FWC and external correlates revealed an unexpected pattern. As expected, we found that FWC
does not add incremental variance over and above WFC when examining its relation
to job stressors, and that WFC does not add incremental variance over and above
FWC when examining its relation to non-work stressors. However, we were surprised to learn that WFC was less related to job satisfaction than was FWC. Certainly, we would expect that both WFC and FWC would be negatively related to job
satisfaction (e.g., Kossek & Ozeki, 1998), though we assumed that job satisfaction
would be aected more when accomplishment of work roles impeded meeting family
demands (WFC). This nding indicates that repeated interruption of work role performance by family demands may be a stronger instigator of decreased job satisfaction. The possibility remains, however, that this result is moderated by the sex of the
worker. In their meta-analysis of work/family conict and job and life satisfaction,
Kossek and Ozeki (1998) found a much stronger relation between work/family conict and job satisfaction for females than for males. Perhaps females, who typically
bear the burden of child and elder care, experience a greater decrease in job satisfaction when these family-related demands interfere with meeting work requirements.
Further research should address the potential that job satisfaction is dierentially affected by family-to-work conict in males and females.
4.1. Limitations
Theories of work/family conict that incorporate a bi-directional conceptualization of conict between work and family have only recently appeared in the literature. Hence, only a small number of studies that employ scales reecting this
paradigm were available. Consequently, many of the estimated relationships we have
reported rely upon a relatively few studies. While we are unaware of any concrete
guidelines regarding a minimum number of studies necessary to utilize meta-analysis,
we acknowledge that second-order sampling error poses a threat to the validity of
our ndings (Hunter & Schmidt, 1990). Nonetheless, given the claims made in the
literature regarding the distinctiveness of WFC and FWC, we felt an empirical examination of this assumption was crucial. In the absence of such a meta-analysis, researchers are likely to base judgments on the ndings of individual studies, which
would be even more likely to lead to error.

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

231

5. Conclusion
Individual lives are vastly inuenced by the distinct institutions of work and family. Understanding the overlap between the two and attempting to mitigate conicts
between them is an important research endeavor, one with the potential to assist
workers in their search to realize satisfaction and fulllment in both domains. The
call to consider the direction of the conict is a recent development, and we investigated whether the measures in the extant literature demonstrate adequate discriminant validity between the two types of conict. By demonstrating adequate
discriminant validity, this paper provides condence in our continuing eorts for
better understanding and theoretical renement in this research domain.

References
*Adams, G. A., King, L. A., & King, D. W. (1996). Relationships of job and family involvement, family
social support, and workfamily conict with job and life satisfaction. Journal of Applied Psychology,
81(4), 411420.
*Boyar, S. L., Maertz Jr., C. P., & Mosley Jr., D. C. (2003). Workfamily conict: The mediating eect of
work and family demand. A paper presented at the meeting of the Society of Industrial Organizational
Psychology, Orlando, FL.
*Brewster, N., Beehr, T., & Wagner, S. (2003). Testing workfamily linkages and a measure of workfamily
conict. A paper presented at the meeting of the Society of Industrial Organizational Psychology,
Orlando, FL.
Burke, R. J., Weir, T., & DuWor, R. E. (1979). Type A behavior of administrators and wives reports of
marital satisfaction and well-being. Journal of Applied Psychology, 64, 5765.
Burley, K. (1989). Workfamily conict and marital adjustment in dual career couples: A comparison of
three time models. Unpublished doctoral dissertation, Claremont Graduate School, Claremont, CA.
Carlson, D. S., Kacmar, K. M., & Williams, L. J. (2000). Construction and initial validation of a
multidimensional measure of work-family conict. Journal of Vocational Behavior, 56, 249
276.
Chen, P. Y., & Spector, P. E. (1991). Negative aectivity as the underlying cause of correlations between
stressors and strains. Journal of Applied Psychology, 76, 398407.
Duxbury, L. E., & Higgins, C. A. (1991). Gender dierences in workfamily conict. Journal of Applied
Psychology, 76(1), 6074.
*Eagle, B. W., Miles, E. W., & Icenogle, M. L. (1997). Interrole conicts and permeability of work and
family domains: Are their gender dierences?. Journal of Vocational Behavior, 50, 168184.
*Flye, L. P., Agars, M. D., & Kottke, J. L. (2003). Organizational approaches to workfamily conict:
Testing an integrative model. Paper presented at the meeting of the Society of Industrial Organizational
Psychology, Orlando, FL.
Frone, M. (2003). Workfamily balance. In J. C. Quick & L. E. Tetrick (Eds.), Handbook of occupational
health psychology (pp. 143162). Washington, DC: American Psychological Association.
*Frone, M. R., Russell, M., & Cooper, M. L. (1992). Antecedents and outcomes of workfamily conict:
Testing a model of the workfamily interface. Journal of Applied Psychology, 77(1), 6578.
*Frone, M. R., & Yardley, J. K. (1996). Workplace family-supportive programmes: Predictors of
employed parents importance ratings. Journal of Occupational and Organizational Psychology, 69,
351366.
*Frone, M. R., Yardley, J. K., & Markel, K. S. (1997). Developing and testing an integrative model of the
workfamily interface. Journal of Vocational Behavior, 50, 145167.
Go, S. J., Mount, M. K., & Jamison, R. L. (1990). Employer supported child care, work/family conict,
and absenteeism: A eld study. Personnel Psychology, 43, 793809.

232

J.R. Mesmer-Magnus, C. Viswesvaran / Journal of Vocational Behavior 67 (2005) 215232

Greenhaus, J. H., & Beutell, N. J. (1985). Sources of conict between work and family roles. Academy of
Management Review, 10, 7688.
*Grzywacz, J. G., & Bass, B. L. (2003). Work, family, and mental health: Testing dierent models of
workfamily t. Journal of Marriage and Family, 65, 248262.
*Gutek, B. A., Searle, S., & Klepa, L. (1991). Rational versus gender role explanations for workfamily
conict. Journal of Applied Psychology, 76(4), 560568.
*Hammer, L. B., Bauer, T. N., & Grandey, A. A. (2003). Workfamily conict and work-related
withdrawal behaviors. Journal of Business and Psychology, 17(3), 419436.
Harman, H. H. (1976). Modern factor analysis (3rd rev.). Chicago, IL: University of Chicago Press.
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in research
ndings. New York City: Sage Publications.
*Judge, T. A., Boudreau, J. W., & Bretz, R. D. Jr., (1994). Job and life attitudes of male executives.
Journal of Applied Psychology, 79(5), 767782.
Kopelman, R. E., Greenhaus, J. H., & Connolly, T. F. (1983). A model of work, family, and interrole
conict: A construct validation study. Organizational Behavior and Human Performance, 32, 198215.
Kossek, E. E., & Ozeki, C. (1999). Bridging the workfamily policy and productivity gap: A literature
review. Community, Work and Family, 2(1), 733.
Kossek, E. E., & Ozeki, C. (1998). Workfamily conict, policies, and the job-life satisfaction relationship:
A review and directions for organizational behavior-human resources research. Journal of Applied
Psychology, 83(2), 139149.
*Laster, K. M. (2002). An examination of the relationships between select psychological dimensions and
work-to-family and family-to-work conict in men and women. Unpublished doctoral dissertation,
University of Oklahoma, Norman, OK.
*Leiter, M. P., & Durup, M. J. (1996). Work, home, and in-between: A longitudinal study of spillover.
Journal of Applied Behavioral Science, 32(1), 2947.
Mitchell, T. R., & James, L. R. (2001). Building better theory: Time and the specication of when things
happen. Academy of Management Review, 26(4), 530547.
*Netemeyer, R. G., Boles, J. S., & McMurrian, R. (1996). Development and validation of workfamily
conict and family-work conict scales. Journal of Applied Psychology, 81(4), 400410.
Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York: McGraw Hill.
*ODriscoll, M. P., Ilgen, D. R., & Hildreth, K. (1992). Time devoted to job and o-job activities, interrole
conict, and aective experiences. Journal of Applied Psychology, 77(3), 272279.
*Parasuraman, S., Purohit, Y. S., & Godshalk, V. M. (1996). Work and family variables, entrepreneurial
career success, and psychological well-being. Journal of Vocational Behavior, 48, 275300.
Schmidt, F. L. (1971). The relative eciency of regression and simple unit predictor weights in applied
dierential psychology. Educational and Psychological Measurement, 31(3), 699714.
*Thompson, C. A., & Blau, G. (1993). Moving beyond traditional predictors of job involvement:
Exploring the impact of workfamily conict and overload. Journal of Social Behavior and Personality,
8(4), 635646.
*Wiley, D. L. (1987). The relationship between work/nonwork role conict and job-related outcomes:
Some unanticipated ndings. Journal of Management, 13(3), 467472.
*Williams, K. J., & Alliger, G. M. (1994). Role stressors, mood spillover, and perceptions of workfamily
conict in employed parents. Academy of Management Journal, 37(4), 837868.