Journal of Occupational Health Psychology © 2010 American Psychological Association

2010, Vol. 15, No. 1, 75–90 1076-8998/10/$12.00 DOI: 10.1037/a0017443

A Short, Valid, Predictive Measure of Work–Family Conflict:

Item Selection and Scale Validation
Russell A. Matthews Lisa M. Kath
Louisiana State University San Diego State University

Janet L. Barnes-Farrell
University of Connecticut
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The purpose of this research is to develop an abbreviated version of Carlson, Kacmar, and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Williams’s (2000) multidimensional measure of work–family conflict. The abbreviated measure

would have high utility in research situations in which a measure representative of the entire
theoretical construct is desired, but the use of a long measure is precluded, as in diary and
longitudinal designs. Two 3-item abbreviated measures based on Carlson et al.’s multidimen-
sional measures were developed, 1 to assess work-to-family conflict and 1 to assess family-to-
work conflict. Two samples were used to provide concurrent and predictive validity evidence for
the abbreviated measure. Results from Study 1 indicate that the abbreviated measure has the
expected factor structure and exhibited concurrent and predictive validity that replicated results
found with Carlson et al.’s 18-item measure. Results from Study 2 provide additional psycho-
metric and construct validity evidence for the abbreviated measure; the abbreviated measure was
internally consistent, exhibited good test–retest reliability, and was systematically related to
measures of role stressors, work–family balance, and well-being outcomes.

Keywords: work–family conflict, measure development, validation

The purpose of this research is to validate an
abbreviated version of Carlson, Kacmar, and Wil-
liams’s (2000) multidimensional measure of work–
family conflict. The measure developed by Carlson et
al. is strong in terms of its representativeness of the
entire theoretical construct of work–family conflict
and its psychometric properties. The measure has
already been used over 25 times in published re-
search (e.g., Allen & Armstrong, 2006; Spector et al.,
2007; van Steenbergen, Ellemers, & Mooijaart,
2007), and has been used to study such diverse topics
as organizational work–family policies (e.g.,
O’Driscoll et al., 2003; Shockley & Allen, 2007),
social support (Daalen, Willemsen, & Sanders,
2006), health (Allen & Armstrong, 2006), worker
well-being (Hughes & Parkes, 2007), and cross-
cultural work–family conflict (e.g., Spector et al.,
2007). Researchers have used the subscales individ-
Russell A. Matthews, Department of Psychology, Louisiana
State University; Lisa M. Kath, Department of Psychology,
San Diego State University; Janet L. Barnes-Farrell, Depart-
ment of Psychology, University of Connecticut.
Correspondence concerning this article should be ad-
dressed to Russell A. Matthews, Department of Psychology,
Louisiana State University, 213 Audubon Hall, Baton
Rouge, LA 70803. E-mail: Matthews@lsu.edu
ually (e.g., Bruck, Allen, & Spector, 2002) and as an
overall measure of work–family conflict (e.g., Witt &
Carlson, 2006), depending on their research ques-
tions.
However, the overall length (18 items) of the mea-
sure developed by Carlson et al. (2000) limits its
usefulness in research in which survey space is at a
premium. An abbreviated version of the measure
would have significant utility in large scale “organi-
zational pulse” type research in which organizations
are trying to assess a wide range of employee atti-
tudes and behaviors efficiently and frequently. In
such cases, organizations often assess constructs with
the fewest items necessary, to limit the amount of
time employees are “off the line” (Stanton, Sinar,
Balzer & Smith, 2002). Such a measure would also
be useful in longitudinal and diary studies in which
frequent assessment takes place across time. General
advantages to using shorter scales include reduced
assessment time, reduced participant fatigue, and in-
creased response rate (Burisch, 1984). Given the
overall strength of Carlson et al.’s measure, a valid
abbreviated version would be helpful for researchers
who have scale length constraints but who still wish
to ensure they are using a theoretically representative
and psychometrically sound measure of work–family
conflict.

75
 76 MATTHEWS, KATH, AND BARNES-FARRELL
Although there are a variety of short measures of
work–family conflict available (e.g., Boyar, Carson,
Mosley, Maertz, & Pearson, 2006; Frone, Russell, &
Cooper, 1992; Gutek, Searle, & Klepa, 1991), we feel
none match the measure developed by Carlson et al.
(2000) in terms of theoretical representativeness or
psychometric support. In an effort to assess work–
family conflict with as few items as possible, work–
family researchers often either (a) create their own
short measure (e.g., Kinnunen & Mauno, 1998), or
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
(b) trim items from existing measures in a way that is
This document is copyrighted by the American Psychological Association or one of its allied publishers.
often not specified (e.g., Kovner, Brewer, Wu,
Cheng, & Suzuki, 2006). These abbreviated mea-
sures typically have not been subjected to much
psychometric or validity evaluation. As a result, we
applied the guidelines set forth by Stanton et al.
(2002) to create an abbreviated measure of work–
family conflict that we then evaluate for psychomet-
ric properties and validity evidence. Thus, the current
study addresses the need for an abbreviated measure
of work–family conflict that is drawn from an estab-
lished measure and has been demonstrated to have
strong psychometric characteristics.
Background
Work–family conflict is a form of interrole conflict
in which the demands of functioning in the two
domains of work and family are incompatible in
some respect (Greenhaus & Beutell, 1985). As such,
theoretical and empirical work recognizes that work–
family conflict functions in a bidirectional manner:
work can interfere with family, and family can inter-
fere with work (Frone, 2003). The majority of mea-
sures currently used by work–family researchers rec-
ognize the bidirectional nature of the conflict, and
several relatively standard measures of work–family
conflict now exist. These include measures devel-
oped by Bohen and Viveros-Long (1981), Carlson et
al. (2000), Frone et al. (1992), Gutek et al. (1991),
Kopelman, Greenhaus, and Connolly (1983), as well
as Netermeyer, Boles, and McMurrian (1996).
In addition to proposing that work–family conflict
is bidirectional, Greenhaus and Beutell (1985) pro-
posed that three types of pressures contribute to the
experience of work–family conflict: time-, strain-,
and behavior-based pressures. Time-based pressure
is a function of devoting time to one role, making it
difficult to devote time to another role (e.g., having to
take a sick child to the doctor during work hours).
Strain-based pressure is a function of strain resulting
from one role negatively affecting participation in
another role (e.g., being in a bad mood because of a
difficult work day, which leads to a short temper with
one’s family). Behavior-based pressure is a function
of specific behaviors required in one role being in-
compatible with behavioral expectations in another
role (e.g., being authoritative at work may facilitate
getting work done, but being authoritative at home
may lead to conflicts with one’s family).
The measure developed by Carlson et al. (2000) is
perhaps the best example of a measure intended to
account for the three bidirectional types of pressures.
These authors used five independent samples in the
development and validation of their measure, which
includes six subscales accounting for the three types
of pressures for both directions of work–family con-
flict. Overall, we consider it to be one of the most
theoretically and psychometrically sound measures
of work–family conflict available to researchers
today.
Although theoretical and psychometric properties
must be considered when selecting measures for re-
search purposes, most researchers are constrained by
another real-world factor: survey length. This is the
primary drawback of the Carlson et al. (2000) measure:
The full measure is 18 items long. In a study in which
a researcher has space to include 75 survey items, this
measure would account for approximately 25% of the
items. Even more restrictive are diary studies (e.g.,
Butler, Grzywacz, Bass, & Linney, 2005), which
often include no more than 15 items in any one
wave of data collection. As a result, researchers
typically use shorter measures of work–family
conflict to maximize the number of constructs that
can be measured.
Indeed, many work–family researchers have used
shorter measures in their research, and these mea-
sures tend to recognize the bidirectional nature of
work–family conflict. One of the more commonly
used measures was developed by Netemeyer et al.
(1996). This 10-item measure was subjected to inten-
sive development and validation procedures, but the
authors acknowledged the measure does not capture
all three types of pressure as outlined by Greenhaus
and Beutell (1985). Boyar et al. (2006) recently
sought to develop an abbreviated, and more psycho-
metrically valid, version of Netemeyer et al.’s mea-
sure. However, as the authors noted, there remain
concerns about the cross-sectional data used for val-
idation purposes. Other common work–family con-
flict measures include Bohen and Viveros-Long
(1981), Frone et al. (1992), Gutek et al. (1991), and
Kopelman et al. (1983). The general issue with cur-
rently available abbreviated measures of work–
 WORK–FAMILY CONFLICT
family conflict is that these measures have not been
subjected to extensive development and validation.
Thus, we propose that it would be useful to have an
abbreviated measure of work–family conflict that is
both psychometrically sound and representative of
the full construct space of work–family conflict as
theorized by Greenhaus and Beutell (1985), includ-
ing the three types of pressures (i.e., time-, strain-,
and behavior-based pressures). Such a measure
would have great utility when survey length is of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
concern, the research question focuses on the full
This document is copyrighted by the American Psychological Association or one of its allied publishers.
construct space of work–family conflict and the re-
search question is concerned with overall work–
family conflict instead of differential relations with
specific types of pressures.
Validation of an Abbreviated Measure
The most common methodology for item selection
in scale reduction is to select items that maximize
internal consistency, that is, selecting items with high
interitem correlations (Clark & Watson, 1995; Stan-
ton et al., 2002). However, as discussed by these and
other authors (e.g., Boyle, 1991; Cortina, 1993;
Kline, 1986), using an internal consistency maximiz-
ing approach can have some drawbacks such as item
redundancy (Smith & Stanton, 1998). After all, the
most internally consistent measure would likely oc-
cur when items are simply paraphrases of each other.
This approach can also lead to factor structure limi-
tations such that, for example, only positively worded
items are selected (Miller & Cleary, 1993). Finally,
focusing on maximizing internal consistency can lead
to reduced validity (Kline, 1986) when researchers
inadvertently trim items that serve to represent a
greater portion of the theoretical construct space
(Smith & Stanton, 1998). Thus, the resulting abbre-
viated scale may have lower construct validity.
To avoid these problems, Stanton et al. (2002)
recommended selecting items based on three criteria:
judgmental qualities (e.g., subjective assessment of
face validity and other nonstatistical considerations),
internal qualities (e.g., item level statistics), and ex-
ternal qualities (e.g., relationship between the scale,
or item, and other related constructs). Based on a
judgmental evaluation of the original measures, we
selected three items for each of the abbreviated
bidirectional measures; one item for each proposed
work–family pressure. To identify the six items
to be included, we examined internal qualities (factor
loadings) of the original measure. To ensure that
these items operated well, we examined the exter-
nal qualities (concurrent and predictive validities)
77
of the abbreviated measure. Utilizing these three
item-selection criteria, our goal was to develop an
abbreviated measure of work–family conflict that
generally replicates the pattern of relationships
observed with Carlson et al.’s (2000) multidimen-
sional measure.
To achieve this goal, we conducted two studies. In
Study 1, we used cross-sectional and multiwave data
from a single data set to examine scale properties,
concurrent validity, and predictive validity of Carl-
son et al.’s (2000) original measure and our abbrevi-
ated measure. In Study 2, we used cross-sectional and
multiwave data from a different dataset to test the
psychometric properties and validity of the abbrevi-
ated measure when used alone.
Study 1
Method
Participants. Study participants were recruited
from a broad range of industries and levels using a
peer-nomination web-based survey. The Time 1 sur-
vey included 735 participants. Of these, 74 were
removed because they did not work at least 15 hr a
week. Another 5 participants were removed for ex-
cessive missing data (completed less than 80% of the
survey). This resulted in a final sample of 656 par-
ticipants. Of these, 272 (41.5%) indicated they were
interested in participating in the Time 2 survey, of
which 95 completed the Time 2 survey (34.9%).
Two subsamples were created: (a) a cross-
sectional sample (n ⫽ 561) consisting of those indi-
viduals who only completed the Time 1 survey, and
(b) a multiwave sample (n ⫽ 95) with those individ-
uals who participated in both waves of data collec-
tion. Participants were included in only one sub-
sample, and all 656 participants were accounted for
across the two subsamples.
In the cross-sectional sample, 59.4% were women,
the mean age was 36.84 (SD ⫽ 13.48), 51.3% indi-
cated they were married or living with life partner,
and 39.6% had at least one child under the age of 18
living at home. The mean tenure with current em-
ployer was 7.34 years (SD ⫽ 8.24) and average work
hours was 38.76 per week (SD ⫽ 11.62). Approxi-
mately 19% reported working in management, busi-
ness, or financial operations related occupations, 18%
reported working in professional and related occupa-
tions, and another 16.2% reported working in more
traditional blue-collar (e.g., production, installation,
maintenance) or service occupations.
 78 MATTHEWS, KATH, AND BARNES-FARRELL
In the multiwave sample, 67.4% were women, the
mean age was 41.24 years (SD ⫽ 12.35), 65.3%
indicated they were married or living with life part-
ner, and 49.5% had at least one child under the age of
18 living at home. The mean tenure with current
employer was 8.75 years (SD ⫽ 8.75) and average
work hours was 38.78 a week (SD ⫽ 10.26). Approx-
imately 17% reported working in management, busi-
ness, or financial operations related occupations,
23.2% reported working in professional and related
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
occupations, and another 13.7% reported working in
This document is copyrighted by the American Psychological Association or one of its allied publishers.
more traditional blue-collar (e.g., production, instal-
lation, maintenance) or service occupations.
Participants in the multiwave sample were older
than the cross-sectional sample (M ⫽ 41.2 and 36.8,
respectively), t(649) ⫽ 2.08, p ⬍ .05, and had a
higher average number of children under the age of
18 living at home (M ⫽ .88 and .66, respectively),
t(649) ⫽ 2.98, p ⬍ .01, but the two samples did not
differ in terms of gender, t(648) ⫽ 1.36, p ⬎ .05,
average number of hours worked, t(648) ⫽ .02, p ⬎
.05, or job tenure, t(648) ⫽ 1.53, p ⬎ .05.
Procedure. Participant recruitment and survey
procedure. Trained undergraduate student recruit-
ers (N ⫽ 163) from advanced psychology classes
from five universities assisted with the data collection
process. These five universities included both public
and private universities that ranged in student body
size from 4,000 to nearly 40,000. Students enrolled at
these universities came from a wide variety of socio-
economic classes and include varying percentages of
first-generation college students. Students were
trained on the data collection methodology and ethics
in research. They were provided with an e-mail in-
vitation that they distributed to working adults they
personally knew who met the eligibility requirements
for the study (at least 18 years old, working 15 or
more hours a week). Students typically recruited 4 to
5 participants each. Recipients of the invitation
emails were asked to follow the web-link supplied in
the email to complete the online survey. The survey
took approximately 15 min to complete; participation
was voluntary. Student recruiters received nominal
course extra credit for their involvement. The nature
of the initial recruitment procedure does not allow us
to calculate a survey response rate.1
At the end of the Time 1 survey, participants were
asked if they would be interested in completing a
10-min follow-up survey in about a month. A
1-month lag was used to allow sufficient time for the
stressor-strain (i.e., work-family conflict— outcome)
relationships to unfold. Respondents interested in
participating were asked to provide their email ad-
dress, the last four digits of their home phone num-
ber, and their postal code (phone number and postal
code information was used to link participants’ two
sets of responses). Those participants who provided
an email address were contacted by the primary in-
vestigator 4 weeks after their initial participation. As
part of this follow-up email, participants were re-
minded of the nature of the study and provided with
a second Web link. They were asked to follow the
Web link and complete the Time 2 survey. Reminder
e-mails were sent 1 week after the follow-up e-mail.
The Time 2 survey took approximately 10 min to
complete; participation was voluntary.
Item selection strategy for the abbreviated mea-
sure. To select items for the abbreviated measure
of work–family conflict, we used the method pro-
posed by Stanton et al. (2002), who advocated the
consideration of judgmental, internal, and external
qualities of the scale. Focusing first on the judg-
mental qualities of the items, we sought to ensure
the abbreviated measure would represent all three
types of pressure for each direction of work–family
conflict. Thus, one item from each of the six pres-
sure subscales developed by Carlson et al. (2000)
was to be selected.
We then focused on internal qualities, choosing
the highest loading items from the final confirma-
tory factor analyses (CFA) reported by Carlson et
al. (2000; see Figure 2, p. 264). Specifically, the
top-loading items for each of the three types of
pressure (time-, strain-, and behavior-based) were
selected, for work-to-family and family-to-work
conflict, respectively (see Table 1 for items se-
lected). The items were aggregated across pressure
types to create two abbreviated subscales of work–
family conflict: Three items assessed work-to-
family conflict and three items assessed family-to-
work conflict. By selecting the highest loading
items from each dimension we essentially selected
the items with the strongest relationship with their
underlying dimensions (Stanton et al., 2002). This
approach of selecting the top-loading items on
each factor to create an abbreviated version of a
measure has been applied successfully before (cf.
Clark & Watson, 1995; Eisenberger, Stinglhamber,
Vandenberghe, Sucharski, & Rhoades, 2002). In
addition, the items selected were not redundant and
1
A portion of these data was used in another study
(Matthews, Bulger, & Barnes-Farrell, in press), but the
research questions and subset of respondents differ from the
current study.
 WORK–FAMILY CONFLICT 79

Table 1
Items Selected for Abbreviated Work–Family Conflict Measures
Item Pressure
Work-to-family
I have to miss family activities due to the amount of time I must spend on work responsibilities. Time based
I am often so emotionally drained when I get home from work that it prevents me from
contributing to my family. Strain based
The behaviors I perform that make me effective at work do not help me to be a better parent
and spouse. Behavior based
Family-to-work
I have to miss work activities due to the amount of time I must spend on family responsibilities. Time based
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Because I am often stressed from family responsibilities, I have a hard time concentrating on
This document is copyrighted by the American Psychological Association or one of its allied publishers.

my work. Strain based

Behavior that is effective and necessary for me at home would be counterproductive at work. Behavior based

they were collectively representative of the larger
construct space of work–family conflict.
Validation strategy for the abbreviated measure.
We validated our abbreviated measure using a con-
struct validation approach (focusing on external qual-
ities), as recommended by Stanton et al. (2002). We
used recent work–family research reviews (Allen,
Herst, Bruck, & Sutton, 2000; Byron, 2005) to iden-
tify traditional predictors (role conflict and role am-
biguity) and outcomes (satisfaction, health, depres-
sion, turnover intentions, and stress) of work–family
conflict to examine concurrent and predictive valid-
ities of the abbreviated measure.
Measures. Unless otherwise indicated, respon-
dents responded on a 5-point Likert scale ranging
from 1 (strongly disagree) to 5 (strongly agree).
Time 1, work–family conflict. Participants com-
pleted the full 18-item version of Carlson et al.’s
(2000) work–family conflict measure. Scale scores
were computed for each of the six subscales (one
dimension each for time-, strain-, and behavior-based
conflict for both work-to-family and family-to-work
directions). In addition, the three respective pressure
subscales were averaged to produce two overall
scores, one for work-to-family conflict and one for
family-to-work conflict. Finally, the three items se-
lected for the abbreviated work-to-family conflict
measure and the abbreviated family-to-work conflict
measure (see Table 1 for items) were averaged to
produce a score for each direction of the conflict.
Time 1 predictors of work–family conflict.
Work role conflict (␣ ⫽ .86) was assessed with a
modified 4-item measure developed by Ivancevich
and Matteson (1976 –2005). A sample item is, “When
at work, I seem to receive conflicting requests from
different people (e.g., coworkers, bosses).” A parallel
set of items was used to assess family role conflict
(␣ ⫽ .87). A sample item is, “When at home, I seem
to receive conflicting requests from different people
(e.g., spouse, parent, child).” Work role ambiguity
(␣ ⫽ .83) was assessed with four items selected from
the measure developed by Rizzo, House, and Lirtz-
man (1970). A sample item is, “When at work, I
know what my responsibilities are.” A parallel set of
items was used to assess family role ambiguity (␣ ⫽
.84). A sample item is, “When at home, I know
exactly what is expected of me.” Items were scored
such that higher scores indicate higher levels of
the predictors.
Time 1, consequences of work–family conflict.
Job satisfaction (␣ ⫽ .84) was assessed with three
items (Cammann, Fichman, Jenkins, & Klesh, 1983).
A sample item is, “All in all, I am satisfied with my
job.” Family satisfaction (␣ ⫽ .79) was assessed with
five items selected from the Family Satisfaction Scale
(Carver & Jones, 1992). A sample item is, “I am
happy with my family just the way it is.” Depression
(␣ ⫽ .92) was assessed with the three-item Psycho-
logical General Well-Being questionnaire (Dupuy,
1984). The following instructions were given: “Be-
low is a list of some of the ways you may have felt or
behaved in the past month. Please indicate how often
you have felt this way during the past month.” Par-
ticipants responded on a 4-point Likert scale ranging
from 1 (rarely or none of the time, less than 1 day) to
4 (all of the time, 5 to 7 days). Overall health was
assessed with a single item, “Would you say that in
general your health is:” Participants responded on a
5-point Likert scale ranging from 1 (poor) to 5 (ex-
cellent).
Time 2, consequences of work–family conflict.
Turnover intentions (␣ ⫽ .95) were assessed with
 80 MATTHEWS, KATH, AND BARNES-FARRELL
four items (Kelloway, Gottlieb, & Barham, 1999). A
sample item is, “I am thinking about leaving this
organization.” Life satisfaction (␣ ⫽ .82) was as-
sessed with five items (Diener, Emmons, Larsen, &
Griffin, 1985). A sample item is, “In most ways my
life is close to my ideal.” Life stress (␣ ⫽ .90) was
assessed with the 10-item version of the Perceived
Stress Scale (Cohen, Kamarck, & Mermelstein,
1983; Cohen & Williamson, 1988). A sample item is,
“In the last month, how often have you been upset
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
because of something that happened unexpectedly?”
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Depression (␣ ⫽ .92) and overall health were as-
sessed with the same measures as at Time 1.
Demographics. Participants were asked to pro-
vide basic demographic information, including gen-
der (1 ⫽ male, 2 ⫽ female), age, number of children,
marital status (1 ⫽ not married, 2 ⫽ married or
living with partner), hours worked, and organiza-
tional tenure.
Results
Factor structure of abbreviated measure. To
confirm the factor structure of the abbreviated work–
family conflict measure (i.e., three-item work-to-
family scale and three-item family-to-work scale),
the cross-sectional data were analyzed using AMOS
5 software (Arbuckle, 2003). Item level list-wise
deletion resulted in a working sample of 552 partic-
ipants. A standard latent variable CFA was con-
.61**
Work-to-Family .67**
Conflict
.65**
d1
.44**
d1
.65**
Family-to-Work .81**
Conflict
.46**
Figure 1. Results of the confirmatory factor analyses to confirm the factor structure of the
abbreviated measure from Study 1. Square multiple correlations reported in italics. ⴱ p ⬍ .05.
ⴱⴱ
p ⬍ .01.
ducted, with one latent factor for work-to-family
conflict and a second latent factor for family-to-work
conflict. The three items for each of the abbreviated
measure’s subscales were included as indicators of
the two latent factors. Following a standard CFA
approach, the disturbances between the two latent
factors were set free to correlate. In addition, given
the nature of the constructs being assessed, the three
types of pressures were set free to correlate across the
two directions of conflict (e.g., the error term for
strain-based work-to-family conflict item was set free
to correlate with strain-based family-to-work conflict
item error term).
Three measures of model fit were calculated: chi-
square, standard-root-mean residual (SRMR), and
comparative fit index (CFI). A nonsignificant chi-
square indicates good model fit; however, chi-square
is sensitive to sample size. An SRMR value of .08 or
lower and a CFI value of 0.95 or higher are also
indicative of good model fit (Hu & Bentler, 1999).
Although the chi-square statistic was significant,
␹2(5) ⫽ 23.72, p ⬍ .05, both the CFI (0.97) and
SRMR (.04) were indicative of good fit for the two-
factor model. Results for these analyses are reported
in Figure 1. As shown, all items had standardized
factor loadings greater than .40. In addition, the cor-
relation between the two latent factors (r ⫽ .44, p ⬍
.01) was not so high as to imply an issue of discrimi-
nant validity (Kline, 1998). This two-factor model
also fit the data significantly better than a model in
.38
Strain-based
e1
work-to-family conflict
.45
Time-based
e2
work-to-family conflict
.17 .12*
Behavior-based
e3
work-to-family conflict
-.05
.43
Strain-based
e4
family-to-work conflict .34**
.65
Time-based
e5
family-to-work conflict
.21
Behavior-based
e6
family-to-work conflict
 WORK–FAMILY CONFLICT 81

which all six items were hypothesized to load on a Table 3

single factor, ⌬␹2(1) ⫽ 98.13, p ⬍ .01. Descriptive Statistics for Validation Measures
Descriptive statistics. Next, as reported in Table 2, Variable M SD ␣
we examined the means, standard deviations, and inter-
nal consistency measures for the original and abbrevi- Cross-sectional samplea
Work role conflict 2.50 .94 .86
ated measures in both the cross-sectional and multiwave Work role ambiguity 2.09 .75 .83
samples. For the original pressure subscales, the orig- Family role conflict 2.28 .89 .87
inal overall measure, and our proposed abbreviated Family role ambiguity 2.10 .77 .84
measure of work–family conflict, the means are Job satisfaction 3.72 .89 .84
Family satisfaction 4.20 .69 .79
around the midpoint of the 5-point scale for both the Depression 1.55 .76 .92
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

samples. There were no significant mean differences Overall Health 3.68 .81 —
This document is copyrighted by the American Psychological Association or one of its allied publishers.

in the work–family conflict measures between the Multiwave sampleb

cross-sectional and multiwave samples. Turnover intentions 2.34 1.19 .95
With the exception of the abbreviated measure of Life satisfaction 3.50 .67 .82
Life stress 2.62 .59 .90
work–family conflict, all measures demonstrated tra- Depression 1.49 .70 .92
ditionally acceptable internal reliability levels (␣ ⱖ Overall health 3.72 .77 —
.70). Because we did not adopt an internal consis- a
Data collected at Time 1. b
Data collected at Time 2.
tency maximizing approach in abbreviating the full
measure, it is not surprising that, by selecting items
that tap into theoretically distinct aspects of work–
family conflict, our internal consistency estimates are
somewhat lower than the original measure (Stanton In Table 4, correlations between all variations of
et al., 2002). We will return to the issue of the low the work–family conflict measures are reported. Cor-
internal consistency estimates in Study 2. relations among variables in the cross-sectional sam-
In Table 3, descriptive statistics for the validation ple are reported below the diagonal and correlations
measures are reported. Data reported for the cross- for the multiwave sample are reported above the
sectional sample were collected at Time 1 (n ⫽ 561). diagonal. The pattern of relationships between mea-
For the multiwave sample, descriptive statistics were sures was relatively similar across the two samples.
only reported for Time 2 (n ⫽ 95), as those were the Because three items from the abbreviated measure
measures used in validation analyses. All measures overlap with the full nine-item measures and one
demonstrated acceptable internal reliabilities (.79 item from the abbreviated measure overlaps with
or greater). each subscale from the full-length measure, this may

Table 2
Descriptive Statistics for Work–Family Conflict Measures
Cross-sectional samplea Multiwave sampleb
Variable M SD ␣ M SD ␣ tc
Work-to-family
Time subscale 2.56 .95 .80 2.49 1.01 .85 ⫺.62
Strain subscale 2.67 .93 .83 2.69 1.03 .87 .25
Behavior subscale 2.47 .76 .69 2.36 .78 .72 ⫺1.24
Overall, nine items 2.57 .71 .85 2.52 .72 .83 ⫺.61
Abbreviated, three items 2.50 .76 .60 2.46 .81 .55 ⫺.45
Family-to-work
Time subscale 2.01 .71 .81 1.87 .72 .86 ⫺1.75
Strain subscale 2.06 .80 .87 1.95 .84 .91 ⫺1.26
Behavior subscale 2.34 .81 .84 2.27 .79 .8 ⫺.74
Overall, nine items 2.14 .84 .88 2.03 .62 .87 ⫺1.50
Abbreviated, three items 2.09 .66 .67 2.03 .69 .62 ⫺.84
Note. All data collected as part of the initial data collection (Time 1).
a
n ⫽ 561. b n ⫽ 95. c df ⫽ 654; no significant differences observed between the cross-sectional and multiwave samples
at p ⬍ .05.
 82 MATTHEWS, KATH, AND BARNES-FARRELL

lead to inflated correlations. Therefore, we recalcu-

Note. Correlations corrected for inflation reported in parentheses. Cross-sectional sample below the diagonal, r ⬎ |.17| ⫽ p ⬍ .01; multiwave sample above the diagonal, r ⬎ |.19|,
.17 (.12)
.31 (.25)
.38 (.37)
.37 (.31)
.42 (.42)

.72 (.62)
.80 (.74)
.68 (.59)
.93 (.83)
10

—
lated each correlation, removing the overlapping item
from each measure for that specific correlation. These
corrected correlations are reported in Table 1 in pa-
rentheses. Across the 45 cross-sectional correlations,
.28 (.23)
.33 (.25)
.41 (.41)
.44 (.38)
.47 (.46)

.81 (.81)
.85 (.83)
.71 (.72)

.92 (.83)
the average difference between the original and cor-

—
9

rected correlations was .04 (ranging from .00 to .13).

For the 45 multiwave correlations, the average dif-
ference between the original and corrected correla-
.27 (.26)
.30 (.21)
.67 (.64)
.52 (.46)
.53 (.50)

.32 (.37)
.35 (.34)

.77 (.77)
.69 (.56)
tions was also .04 (ranging from ⫺.05 to .12). Based
—
8
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

on these results we do not feel that the correlations

This document is copyrighted by the American Psychological Association or one of its allied publishers.

were unduly inflated because of overlapping items;

for all further analyses we report results based on the
complete measures (i.e., overlapping items are
.21 (.14)
.28 (.25)
.19 (.24)
.30 (.27)
.34 (.34)

.64 (.57)

.47 (.45)
.87 (.85)
.80 (.74)
—

not excluded).
7

Relationships among original and abbreviated

measures of conflict. As expected, the abbreviated
measure for work-to-family conflict correlated
.17 (.18)
.18 (.13)
.10 (.10)
.20 (.19)
.23 (.23)

.67 (.62)
.41 (.39)
.83 (.81)
.79 (.72)

strongly with Carlson et al.’s (2000) directional pres-

—
6

sure subscales (correlations for the two samples

ranged between .65 and .78). The correlations be-
tween the abbreviated measure for work-to-family
.72 (.66)
.68 (.56)
.71 (.60)
.92 (.81)

.29 (.29)
.32 (.32)
.46 (.44)
.44 (.44)
.38 (.38)

conflict and the nine-item overall version of Carlson

—
5

et al.’s original measure was also very strong (.91 for

the cross-sectional sample and .92 for the multiwave
sample). Similarly, the abbreviated measure for fam-
ily-to-work conflict correlated strongly with Carlson
.81 (.81)
.79 (.77)
.68 (.63)

.91 (.80)

.26 (.23)
.32 (.31)
.48 (.42)
.43 (.40)
.38 (.35)
—

et al.’s directional pressure subscales (correlations for

4

the two samples ranged between .68 and .80). The

correlation between the abbreviated measure of fam-
ily-to-work conflict and the nine-item overall version
.38 (.34)
.30 (.23)

.72 (.71)
.65 (.52)

.22 (.20)
.28 (.27)
.60 (.52)
.46 (.42)
.40 (.35)

of Carlson et al.’s original measure was also very

—
3

strong (.92 for the cross-sectional sample and .93 for

Bivariate Correlations for Work–Family Conflict Measures

the multiwave sample). These correlations provide

partial evidence that our abbreviated measure func-
.42 (.40)

.47 (.44)
.87 (.85)
.77 (.68)

.18 (.18)
.29 (.29)
.36 (.34)
.34 (.34)
.29 (.27)

tions similarly to the Carlson et al. measure.

—
2

Concurrent validation. Using the cross-sec-

tional data to further validate the abbreviated mea-
sure of work–family conflict, the observed relation-
ships between the abbreviated measure and various
.60 (.54)
.35 (.32)
.83 (.81)
.78 (.69)

.23 (.18)
.21 (.18)
.24 (.18)
.27 (.22)
.25 (.22)
—

predictors and outcomes of work-to-family and fam-

1

ily- to-work conflict were calculated. These results

with the abbreviated measures were compared to
corresponding relationships of predictors and out-
5. Abbreviated, three items

10. Abbreviated, three items

p ⬍ .05; r ⬎ |.26| ⫽ p ⬍ .01.

comes with the directional pressure subscales and the

nine-item overall measures of work-to-family and
4. Overall, nine items

9. Overall, nine items

3. Behavior subscale

8. Behavior subscale
Work-to-family conflict

Family-to-work conflict

family-to-work conflict. Because we were testing

2. Strain subscale

7. Strain subscale
1. Time subscale

6. Time subscale
Variable

whether two correlations from the same sample sig-

nificantly differ, we used the t test formula for testing
correlations that are not statistically independent
(Blalock, 1972, p. 407). Significant t tests would
Table 4

indicate that the abbreviated measure did not have the

same relationship with validation measures as the
 WORK–FAMILY CONFLICT 83

original measure. Given the large number of compar-
isons being made (64 comparisons), concerns over
inflated Type I error exists. One method to control for
family wise Type I error is to apply a correction for
alpha, such as the Bonferroni correction. However,
this method for controlling for Type I error can
become too conservative when the number of com-
parisons grows large (Jaccard & Wan, 1996); the
Bonferroni correction to maintain an estimated fam-
ily wise Type I error rate of .05 with 64 comparisons
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
As reported in Table 5, the correlations with pre-
dictors and outcomes for the abbreviated measure of
work–family conflict did not significantly differ from
the correlations with predictors and outcomes for the
nine-item overall measures (p ⬍ .001). Full t-test
results are not reported; detailed results can be ob-
tained from the first author. In sum, based on the
cross-sectional data, patterns of relationships for
work-to-family conflict and family-to-work conflict
were equivalent for the three-item and the nine-

would be to set p at .00078. An additional trade-off item measures.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

for attempting to control for Type I error is a poten-
tial increase in Type II errors (Rothman, 1990). To
balance concerns over Type I and Type II error rates,
we used a moderately conservative criterion of p ⬍
.001 for statistical significance.
Relationships with predictors and outcomes of
work–family conflict for our abbreviated measure did
differ somewhat from those observed for the direc-
tional pressure subscales. Two significant differences
were observed for the abbreviated work-to-family

Table 5
Comparison of Correlations of Work–Family Conflict Measures With Construct Validation Measures:
Cross-Sectional Sample
Abbreviated, Overall, Time Strain Behavior
Variable three items nine items subscale subscale subscale
WFC
Gender ⫺.03a ⫺.02 ⫺.04 .07a ⫺.09
Age ⫺.17 ⫺.15 ⫺.18 ⫺.12 ⫺.06
Number of children ⫺.01 ⫺.02 .00 ⫺.05 .01
Marital status ⫺.14 ⫺.10 ⫺.12 ⫺.07 ⫺.04
Work hours .12 .16 .17 .14 .06
Organizational tenure ⫺.13 ⫺.12 ⫺.12 ⫺.07 ⫺.09
Work role conflict .34 .38 .29 .37 .27
Work role ambiguity .31 .31 .24 .30 .20
Family role conflict .28 .28 .25 .23 .20
Family role ambiguity .25a .23 .14a .23 .19
Job satisfaction ⫺.34b ⫺.36 ⫺.29 ⫺.34 ⫺.22b
Family satisfaction ⫺.18 ⫺.17 ⫺.11 ⫺.16 ⫺.16
Depression .26 .28 .18 .33 .15
Overall health ⫺.16 ⫺.17 ⫺.10 ⫺.19 ⫺.13
FWC
Gender .07a .02 .04 .08 ⫺.07a
Age ⫺.12 ⫺.10 ⫺.08 ⫺.11 ⫺.07
Number of children .08 .07 .15 .03 .01
Marital status ⫺.08 ⫺.07 ⫺.01 ⫺.12 ⫺.04
Work hours ⫺.06a ⫺.02 ⫺.07 ⫺.07 .07a
Organizational tenure ⫺.10 ⫺.09 ⫺.04 ⫺.09 ⫺.10
Work role conflict .17 .19 .11 .14 .22
Work role ambiguity .30 .32 .28 .27 .24
Family role conflict .44c .48 .41 .44 .34c
Family role Ambiguity .36 .41 .34 .39 .28
Job satisfaction ⫺.18 ⫺.19 ⫺.10 ⫺.15 ⫺.21
Family satisfaction ⫺.28 ⫺.32 ⫺.22 ⫺.36 ⫺.21
Depression .24a,b .27 .13a .34b .20
Overall health ⫺.21a ⫺.24 ⫺.12a ⫺.25 ⫺.20
Note. Ns ⫽ 555 to 561. Coefficients that share a subscript significantly differ in magnitude at the p ⬍ .001 level based
on dependent samples t tests. Gender: 1 ⫽ male, 2 ⫽ female. Marital status: 1 ⫽ not married, 2 ⫽ married/living with
partner. r ⱖ |.09| significant at p ⬍ .05, and r ⱖ |.12| significant at p ⬍ .01. WFC ⫽ work-to-family conflict; FWC ⫽
family-to-work conflict.
 84 MATTHEWS, KATH, AND BARNES-FARRELL
conflict measure, and four significant differences
were observed for the abbreviated family-to-work
conflict measure. For each of these differences, we
conducted a t test to determine if the correlations
observed for the nine-item overall measure signifi-
cantly differed from the relationship observed using
the individual pressure subscales. In each case, the
relationship observed with the nine-item overall mea-
sure also significantly differed from the relationship
observed with the pressure subscales; the same pat-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
tern of differences was observed for both measures of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
work–family conflict for both directions. This pro-
vides additional support for the assertion that the
abbreviated measure functions similarly to the parent
overall measure. We noted that additional differences
were observed between the abbreviated measure and
the directional pressure subscales for gender and
hours worked. However, none of the bivariate corre-
lations themselves were significant and thus not par-
ticularly meaningful.2
Predictive validation. Predictive validation was
conducted by correlating Carlson et al.’s (2000) sub-
scales, the overall scale, and the abbreviated measure
of work–family conflict (both work-to-family and
family-to-work conflict at Time 1) with the outcome
measures (at Time 2). To test whether the magnitude
of the bivariate correlation differed between the ab-
breviated measure compared to Carlson et al.’s pres-
sure subscales or composite nine-item overall mea-
sures, the same t-test approach as described above
was used. Given the number of comparisons (40
total) and the sample size (n ⫽ 95), only those
correlations that differed significantly at p ⬍ .01 were
considered meaningful, that is, t values had to exceed
t(92) ⫽ 2.63.
For the multiwave data, the observed correlations
between the two overall measures (at Time 1) and the
outcome measures (at Time 2) were equivalent in all
but one case, as reported in Table 6. For life stress,
the correlation observed for the abbreviated measure
of family-to-work conflict (r ⫽ .29, p ⬍ .01) was not
as strong as the correlation observed for the full
nine-item overall measure (r ⫽ .43, p ⬍ .01). To
better understand these results we correlated each of
the nine family-to-work conflict items with life stress.
The strongest predictors of life stress were the three
strain items. As such, although the strain item in-
cluded in the abbreviated family-to-work conflict
measure had a strong correlation with life stress (r ⫽
.42), the other two items, representing time- and
behavior-based pressure, had only small correlations
with life stress (average r ⫽ .13). Thus, the variance
accounted for by the items in the abbreviated mea-
sure was lower, driving the observed differences in
the correlations between the measures. For all other
constructs, the pattern of observed relationships was
similar for the abbreviated measure and the
longer measure.
As with the cross-sectional sample, some relation-
ships using the subscales significantly differed from
those using the abbreviated scale. We followed up
with an analysis of relationships using the full overall
scale as well. The full overall measure exhibited the
same pattern of differences in relationships with out-
comes as the abbreviated measure did, with the ex-
ception of family-to-work conflict relationships with
life stress and overall health outcomes. For those two
outcomes, the correlation observed for the nine-item
overall measure did not significantly differ from the
correlation observed for the pressure subscales.3
Discussion
The purpose of Study 1 was to develop an abbre-
viated measure of Carlson et al.’s (2000) measure of
work–family conflict. The framework proposed by
Stanton et al. (2002) was used to identify appropriate
representative items for the abbreviated measure. We
then demonstrated that these items fit a two-factor
structure corresponding to the two work–family con-
flict directions. The abbreviated measure showed
strong correlations with the original pressure sub-
scales and the nine-item overall measures of work-
to-family conflict and-family-to-work conflict.
Further validation evidence for the abbreviated
measure was provided using the cross-sectional sam-
ple to examine the relationships of the measure with
known predictors and outcomes of work–family con-
flict. Relationships of the abbreviated measure with
predictors and outcomes were generally of the same
magnitude as those using the original measure. Also,
using a predictive validation strategy with our mul-
2
Analyses reported in Table 5 were also recalculated
correcting for attenuation such that overlapping items were
excluded from the work–family measures. The results of
these additional analyses replicate those reported in Table 5
with the following exception: Correlations observed with
the abbreviated work-to-family conflict scale significantly
differed from the correlations observed with the work-to-
family time-based pressure subscale for work role ambigu-
ity (r ⫽ .31 versus .19), family satisfaction (r ⫽ ⫺.18
versus ⫺.07), and depression (r ⫽ .26 versus .14).
3
Analyses reported in Table 6 were also recalculated
correcting for attenuation such that overlapping items were
excluded from the work–family measures. The results of
these additional analyses replicate those reported in Table 6.
 WORK–FAMILY CONFLICT 85

Table 6
Correlations for the Abbreviated Measure and the Original Measure With Construct Validation Measures
and Testing for Significant Differences in Magnitudes of Correlations Using Dependent Sample t Test:
Longitudinal Sample
Time 1
Abbreviated, Overall, Time Strain Behavior
Time 2 three items nine items subscale subscale subscale
WFC
Turnover intentions .38 .40 .22 .40 .28
Life satisfaction ⫺.25 ⫺.30 ⫺.17 ⫺.39 ⫺.09
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Life stress .45b .44 .35 .46 .17b

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Depression .09a .09 ⫺.07 .34a ⫺.10

Overall health ⫺.23b ⫺.22 ⫺.21 ⫺.26 .01b
FWC
Turnover intentions .22 .20 .12 .13 .22
Life satisfaction ⫺.24 ⫺.30 ⫺.24 ⫺.31 ⫺.15
Life stress .29a,b .43a .30 .50b .20
Depression .20 .24 .21 .35 .01
Overall health ⫺.02b ⫺.08 ⫺.11 ⫺.21b .13
Note. Ns ⫽ 89 to 95. Correlation coefficients that share a subscript significantly differ in magnitude at the p ⬍ .001 level
based on dependent-samples t tests. r ⱖ |.21| significant at p ⬍ .05, r ⱖ |.28| significant at p ⬍ .01. WFC ⫽ work-to-family
conflict; FWC ⫽ family-to-work conflict.

tiwave sample, we again demonstrated that relation- Study 2

ships of the abbreviated measure of work–family
conflict with additional outcomes of work–family
conflict were generally of the same magnitude as
those using the original measure.
One noteworthy exception was the result for the
relationship between family-to-work conflict and life
stress. The original overall scale had a significantly
higher correlation with life stress than did our abbre-
viated measure. It appears, based on available data,
the primary driver of the .43 correlation observed
with the overall measures was a function of the strain
based items (the correlation between the three-item
family-to-work strain-based measure with life stress
was .50). Because of these results, we suggest that
when life stress is the primary outcome of interest,
researchers should examine family-to-work conflict
as a function of the three components rather than
using an abbreviated global measure of work-to-
family conflict.
A significant concern that remained was the rela-
tively low internal consistency reliability estimates
for our abbreviated measure. We reasoned that this
might be due to the presence of the other items from
the original measure (Harrison & McLaughlin,
1996). In addition, Study 1 did not address other
aspects of the reliability, such as stability over time.
Thus, Study 2 was conducted to examine the reliabil-
ity and validity of the abbreviated measure when
used by itself.
Study 2 was conducted to provide additional in-
formation about the psychometric characteristics of
the abbreviated measure developed in Study 1. The
rationale for Study 2 was twofold. First, we examined
reliability in two ways. We expected that internal
consistency estimates would increase when the ab-
breviated measure was completed as it was intended
to be used in practice: As a single, brief measure of
work–family conflict, rather than in the context of
other items that represent and highlight the distinct
facets of work–family conflict. We also took advan-
tage of the opportunity to gather information about a
second aspect of reliability for this measure, that of
test–retest reliability. Second, we obtained additional
construct validity evidence for the abbreviated mea-
sure when it was used as the only measure of work–
family conflict.
Method
Participants. The Time 1 survey had 202 re-
spondents (response rate ⫽ 21%). Of the 161 in-
dividuals who met the inclusion criteria, 69.4%
were women, 67.7% reported being married or
living with their partner, the mean age was 39.1
(SD ⫽ 10.34), 50.9% had at least one child under
the age of 18 living at home, and 68.9% were
White. The mean tenure with their current employ-
 86 MATTHEWS, KATH, AND BARNES-FARRELL

ers was 7.39 years (SD ⫽ 7.61) and the average
hours worked per week was 39.89 (SD ⫽ 10.60).
Approximately 9.3% reported working in manage-
ment, business, or financial operations related oc-
cupations, 14.3% reported working in professional
and related occupations, and another 15.4% re-
ported working in more traditional blue-collar
(e.g., production, installation, maintenance) or ser-
vice occupations. The response rate for the Time 2
survey was 63% (N ⫽ 101). No significant differ-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
otherwise indicated, respondents were asked to con-
sider the last month when responding to each item
and they responded based on a 5-point Likert scale
ranging from 1 (strongly disagree) to 5 (strongly
agree). Internal consistency reliability estimates for
all measures are reported in Table 7.
Participants completed the abbreviated measure
of work–family conflict developed in Study 1.
Work–family balance was assessed with a single
item; “In general I feel that I have an adequate

ences were observed between respondents to the balance between my work and personal life.” Work
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Time 2 survey and nonrespondents in terms of
gender, age, number of children, hours worked, or
organizational tenure.
Procedure. Participants were recruited using
an online participant recruitment panel that is man-
aged as part of an IRB-approved university-based
research study intended to provide diverse samples
for a wide range of research studies (www.studyre-
sponse.org). An invitation for the Time 1 survey
was sent to 952 individuals registered with the
panel who also met the following criteria: a U.S.
resident and working at least 15 hr per week in an
organizational setting. Respondents to the Time 1
survey were invited back to participate in the Time
2 survey 1 month later.
Measures. All constructs were assessed with the
same measures at both Time 1 and Time 2. Unless
role overload was assessed a five-item version of a
measure reported on by Thiagarajan, Chakrabarty,
and Taylor (2006), excluding the item that refer-
enced being a parent. A sample item is, “I have to
do things that I do not really have the time and
energy for.” Family role overload was assessed
with a parallel set of items, but participants were
asked to consider their family/home life when re-
sponding to the items. Job satisfaction and family
satisfaction were assessed with the same three- and
five-item measures used in Study 1. General well-
being was assessed with the 12-item version of the
General Health Questionnaire (Banks et al., 1980).
A sample item is, “Over the past month I have lost
much sleep over worry.” Respondents used a
5-point response scale ranging from 1 (never) to
5 (always).

Table 7
Convergent Validity Evidence for the Abbreviated Work–Family Conflict Measure
Time 1
Work-to-family Family-to-work
Variable M SD ␣ conflict conflict
Time 1 (N ⫽ 160–161)
Work-to-family conflict 2.74 0.95 .75 — .52ⴱⴱ
Family-to-work conflict 2.39 0.83 .71 .52ⴱⴱ —
Work–family balance 3.52 1.09 — ⫺.59ⴱⴱ ⫺.26ⴱⴱ
Work overload 3.06 1.02 .90 .60ⴱⴱ .47ⴱⴱ
Family overload 3.28 1.03 .92 .62ⴱⴱ .55ⴱⴱ
Job satisfaction 3.66 0.96 .84 ⫺.40ⴱⴱ ⫺.19ⴱ
Family satisfaction 4.06 0.76 .85 ⫺.37ⴱⴱ ⫺.37ⴱⴱ
General well-being 3.61 0.69 .90 ⫺.53ⴱⴱ ⫺.43ⴱⴱ
Time 2 (N ⫽ 98–101)
Work-to-family conflict 2.53 0.97 .80 .75ⴱⴱ .45ⴱⴱ
Family-to-work conflict 2.31 0.86 .72 .46ⴱⴱ .61ⴱⴱ
Work–family balance 3.74 1.01 — ⫺.56ⴱⴱ ⫺.35ⴱⴱ
Work overload 2.95 1.13 .93 .42ⴱⴱ .31ⴱⴱ
Family overload 3.18 1.12 .92 .38ⴱⴱ .42ⴱⴱ
Job satisfaction 3.75 0.96 .85 ⫺.42ⴱⴱ ⫺.20ⴱ
Family satisfaction 4.06 0.77 .82 ⫺.43ⴱⴱ ⫺.31ⴱⴱ
General well-being 3.73 0.74 .91 ⫺.50ⴱⴱ ⫺.49ⴱⴱ
ⴱ ⴱⴱ
p ⬍ .05. p ⬍ .01.
 WORK–FAMILY CONFLICT
Results
Basic descriptive statistics and reliability estimates
for the abbreviated measure are reported in Table 7.
Internal consistency estimates for the abbreviated
measure at Time 1 and Time 2, were above standard
cut-off levels (␣ ⱖ .70). Test–retest reliabilities were
also high (work-to-family conflict ⫽ .75, family-to-
work conflict ⫽ .61), and consistent with past re-
search. Levels of both work-to-family conflict,
t(159) ⫽ ⫺1.01, p ⬎ .05, and family-to-work con-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
flict, t(159) ⫽ ⫺.44, p ⬎ .05, did not differ between
samples.
The CFA model tested in Study 1 (see Figure 1)
was cross-validated using all respondents from the
Time 1 survey and had good model fit, ␹2(5) ⫽ 8.43,
p ⬎ .05, CFI ⫽ .99, root-mean-square error of ap-
proximation (RMSEA) ⫽ .07. The results of Study 2
confirm the structure of the abbreviated measure us-
ing an independent sample and demonstrate reliabil-
ity using two different estimates of reliability.
Additional construct validity evidence. To
provide additional construct validity evidence for the
abbreviated measure, a series of correlations were
calculated (see Table 7), using both the cross-
sectional subsample (N ⫽ 161) and the multiwave
subsample (N ⫽ 101). In the top portion of the Table
7, we report correlations of work-to-family and fam-
ily-to-work conflict with relevant constructs, all as-
sessed at Time 1. The lower portion of Table 7
reports on correlations of work-to-family and family-
to-work conflict, assessed at Time 1, and the same set
of relevant constructs, assessed at Time 2.
As reported in Table 7, the work-to-family and
family-to-work conflict scales were negatively re-
lated to perceptions of work–family balance, both at
Time 1 and Time 2, as should be expected. Both
measures were also systematically related to experi-
ences of role stressors, in the form of role overload,
at both Time 1 and Time 2. In addition, both of the
abbreviated scales were systematically related to a
series of well-being measures (i.e., job and family
satisfaction, as well as general well-being), at Time 1
and at Time 2.
Discussion
The purpose of Study 2 was to provide additional
reliability information and construct validity evi-
dence for our abbreviated measure. One notable fea-
ture of Study 2 is that it more faithfully represents an
“in use” context for the measure, in the sense that the
only work–family conflict items offered to respon-
87
dents for the Time 1 and Time 2 surveys were the six
items of the abbreviated measure. This is consistent
with the circumstances for which the measure was
intended. In this context, we found that the internal
consistency estimates exceeded standard cutoff levels
of .70. The most likely explanation for the higher
coefficient alphas observed in Study 2 relative to
Study 1 is reduced error of measurement in Study 2.
We speculate that this occurred because respondents
in Study 2 were not distracted by the larger set of
conceptually similar items that was presented to par-
ticipants in Study 1 (for relevant discussions, see
Cortina, 1993; Harrison & McLaughlin, 1996). We
also found the test–retest reliabilities to be fairly
stable over the course of a month, as is consistent
with past research of work–family conflict (e.g.,
Hammer, Neal, Newsom, Brockwood, & Colton,
2005; Kelloway et al., 1999; Rantanen, Kinnunen,
Feldt, Pulkkinen, 2008).
We also observed additional construct validity ev-
idence. Of particular note, the observed relationships
between role overload and work-to-family and fam-
ily-to-work conflict are in line with past meta-
analytic work (Byron, 2005). In addition, our results
are consistent with past meta-analytic work examin-
ing the relationship between work-to-family and fam-
ily-to-work conflict and domain satisfaction (Allen et
al., 2000; Ford, Heinen, & Langkamer, 2007).
General Discussion
The purpose of the present research was to develop
a valid abbreviated version of Carlson et al.’s (2000)
measure of work–family conflict for use in research
situations in which a theoretically representative
measure is desired, but the original 18-item measure
is too long. The current research was successful in
achieving this goal such that two 3-item abbreviated
measures were developed, one to assess work-to-
family conflict and one to assess family-to-work con-
flict. The two studies we conducted provide evidence
that our abbreviated measure demonstrates consis-
tent, interpretable, and meaningful relationships with
other constructs, both concurrently and predictively.
We feel confident that our measure is sufficiently
reliable to be used when a measure that includes
broad coverage of the conceptual construct space is
required and circumstances dictate the need for an
abbreviated measure.
There are several limitations that should be con-
sidered when interpreting study results. First, both
studies utilized convenience samples. If these sam-
ples are overly homogeneous with respect to study
 88 MATTHEWS, KATH, AND BARNES-FARRELL
variables, the magnitude and pattern of observed
relationships may differ from population values.
Nonetheless, both samples include adults from mul-
tiple-geographical regions, organizations, organiza-
tional levels, employment sectors, and family situa-
tions.
In addition, all of our measures were self-reported,
which raises the possibility that common method
variance may have spuriously produced the relation-
ships that we observed among variables. However,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
we note that common method variance should only
This document is copyrighted by the American Psychological Association or one of its allied publishers.
be considered a serious issue if there appears to be a
systematic and pervasive inflation of observed rela-
tionships (James, Gent, Hater, & Corey, 1979). Al-
though we cannot completely discount that common
method bias could be at play, we believe it does not
significantly drive results because we have a wide
range of correlations, little evidence of multicol-
linearity, and no nonintuitive relationships.
We suggest that our abbreviated measure will have
the greatest utility for researchers who are seeking to
assess overall, or global, work–family conflict. In
such situations, having a comprehensive, psychomet-
rically sound measure is critical. On the other hand,
when the primary concern is development of theoret-
ical distinctions or diagnosis of particular aspects of
work–family conflict that can be used to tailor spe-
cific kinds of interventions, overall/global measures
of work–family conflict will not be as informative as
scales designed to measure theoretically distinguish-
able subdimensions of the work–family conflict con-
struct space. This measure is not appropriate for those
kinds of situations. In such instances the original
measure developed by Carlson et al. (2000) may be
more appropriate.
As such, based on the results of the current set of
studies, we see two promising lines of future re-
search. First, it will be important for work–family
conflict researchers to identify which outcomes are
best predicted by the pressure-based subscales as
opposed to overall/global measures. For example, our
findings indicated that the overall measures of work-
to-family and family-to-work conflict (nine- or three-
item versions) were largely equivalent in predicting a
wide number of known outcomes of work–family
conflict. Yet there were some notable differences,
such as depression, overall health, and life stress.
Further study to identify and understand these differ-
ences will help researchers choose the most appro-
priate work–family conflict measures for their re-
search.
Second, in Study 2 we observed that experiences
of work-to-family and family-to-work conflict are
relatively stable over a 1-month lag, which is consis-
tent with past longitudinal work–family research that
have used even longer intervals between assessments
(e.g., Hammer et al., 2005; Kelloway et al., 1999;
Rantanen et al., 2008). However, other researchers,
using diary-type designs (e.g., Butler et al., 2005)
have found work–family conflict to be highly vari-
able across days, with considerable within-person
variation. Thus, future research might seek to exam-
ine the manner in which individuals take daily expe-
riences of work–family conflict and seek to aggregate
them when making assessments about overall (e.g.,
over the past month) assessment of work–family con-
flict. For example, if an individual experiences con-
siderable variance in day-to-day experiences of
work-family conflict; how does that affect the stabil-
ity of their overall levels of work–family conflict, and
in turn how does that influence important predictor-
outcome relationships?
Conclusions
As discussed by Stanton et al. (2002), there cur-
rently exists little guidance in the literature concern-
ing how researchers should attempt to develop ab-
breviated version of multi-item measure. Without a
well thought-out, systematic approach to reducing
the number of items in a measure, a researcher might
achieve a shorter measure, but at the same time, this
abbreviated measure may be lacking in terms of
content or construct validity. Following procedures
discussed by Stanton et al., we developed an abbre-
viated measure of work–family conflict based on
Carlson et al.’s (2000) multidimensional measure.
This theoretically meaningful and empirically valid
abbreviated measure should be particularly advanta-
geous for researchers conducting longitudinal and
diary-type designs, as well as for any researcher for
whom survey space is limited.
References
Allen, T. D., & Armstrong, J. (2006). Further examination
of the link between work-family conflict and physical
health the role of health-related behaviors. American
Behavioral Scientist, 49, 1204 –1221.
Allen, T. D., Herst, D. E. L., Bruck, C. S., & Sutton, M.
(2000). Consequences associated with work-to-family
conflict: A review and agenda for future research. Jour-
nal of Occupational Health Psychology, 5, 278 –308.
Arbuckle, J. L. (2003). AMOS 5 (Computer software).
Chicago, IL: Smallwaters.
Banks, M. J., Clegg, C. W., Jackson, P. R., Kemp, N. J.,
Stafford, E. M., & Wall, T. D. (1980). The use of the
General Health Questionnaire as an indication of mental
 WORK–FAMILY CONFLICT
health in occupational settings. Journal of Occupational
Psychology, 53, 187–194.
Blalock, H. M. (1972). Social statistics (2nd ed.). New
York: McGraw-Hill.
Bohen, H. H., & Viveros-Long. (1981). Balancing jobs and
family life: Do flexible work schedules really help? Phil-
adelphia: Temple University Press.
Boyar, S. L., Carson, C. M., Mosley, D. C., Maertz, C. P.,
& Pearson, A. W. (2006). Assessment of the validity of
Netemeyer et al.’s (1996) WFC and FWC scales. Inter-
national Journal of Conflict Management, 17, 34 – 44.
Boyle, G. J. (1991). Does item homogeneity indicate inter-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
nal consistency or item redundancy in psychometric
This document is copyrighted by the American Psychological Association or one of its allied publishers.
scales? Personality and Individual Differences, 12, 291–
294.
Bruck, C. S., Allen, T. D., & Spector, P. E. (2002). The
relation between work-family conflict and job satisfac-
tion: A finer-grained analysis. Journal of Vocational Be-
havior, 60, 336 –353.
Burisch, M. (1984). Approaches to personality inventory
construction: A comparison of merits. American Psy-
chologist, 39, 214 –227.
Butler, A. B., Grzywacz, J. G., Bass, B. L., & Linney, K. D.
(2005). Extending the demands-control model: A daily
diary study of job characteristics, work-family conflict
and work-family facilitation. Journal of Occupational
and Organizational Psychology, 78, 155–169.
Byron, K. (2005). A meta-analytic review of work-family
conflict and its antecedents. Journal of Vocational Be-
havior, 67, 169 –198.
Cammann, C., Fichman, M., Jenkins, G. D., & Klesh, J. R.
(1983). Assessing the attitudes and perceptions of orga-
nizational members. In S. E. Seashore, E. Lawler, P. H.
Mirvis, & C. Cammann (Eds.), Assessing organizational
change (pp. 71–138). New York: Wiley.
Carlson, D. S., Kacmar, K. M., & Williams, L. J. (2000).
Construction and initial validation of a multidemensional
measure of work-family conflict. Journal of Vocational
Behavior, 56, 249 –276.
Carver, M. D., & Jones, W. H. (1992). The Family
Satisfaction Scale. Social Behavior and Personality,
20, 71– 84.
Clark, L. A., & Watson, D. (1995). Constructing validity:
Basic issues in objective scale development. Psycholog-
ical Assessment, 7, 309 –319.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A
global measure of perceived stress. Journal of Health
and Social Behavior, 24, 386 –396.
Cohen, S., & Williamson, G. (1988). Perceived stress in a
probability sample of the United States. In S. Spacapam
& S. Oskamp (Eds.). The social psychology of health:
Claremont Symposium on Applied Psychology (pp. 31–
67). Newbury Park, CA: Sage.
Cortina, J. M. (1993). What is coefficient alpha? An exam-
ination of theory and applications. Journal of Applied
Psychology, 78, 98 –104.
Daalen, G. V., Willemsen, T. M., & Sanders, K. (2006).
Reducing work—family conflict through different
sources of social support. Journal of Vocational Behav-
ior, 69, 462– 476.
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S.
(1985). The satisfaction with life scale. Journal of Per-
sonality Assessment, 49, 71–75.
Dupuy, H. J. (1984). The Psychological General Well-
89
Being Index (PGWB). In N. K. Wenger, M. E. Mattson,
C. D. Furburg, & J. Elinson (Eds.), Assessment of quality
of life in clinical trials of cardiovascular therapies (pp.
170 –183). New York: Le Jacq.
Eisenberger, R., Stinglhamber, F., Vandenberghe, C.,
Sucharski, I. L., & Rhoades, L. (2002). Perceived super-
visor support: Contributions to perceived organizational
support and employee retention. Journal of Applied Psy-
chology, 87, 565–573.
Ford, M. T., Heinen, B. A., & Langkamer, K. L. (2007).
Work and family satisfaction and conflict: A meta-
analysis of cross-domain relations. Journal of Applied
Psychology, 92, 57– 80.
Frone, M. R. (2003). Work-family balance. In J. C. Quick &
L. E. Tetrick (Eds.), Handbook of occupational health
psychology (pp. 143–162). Washington, DC: American
Psychological Association.
Frone, M. R., Russell, M., & Cooper, M. L. (1992). Ante-
cedents and outcomes of work-family conflict: Testing a
model of the work-family interface. Journal of Applied
Psychology, 77, 65–78.
Greenhaus, J. H., & Beutell, N. J. (1985). Sources of con-
flict between work and family roles. Academy of Man-
agement Review, 10, 76 – 88.
Gutek, B. A., Searle, S., & Klepa, L. (1991). Rational versus
gender role explanations for work-family conflict. Jour-
nal of Applied Psychology, 76, 560 –568.
Hammer, L. B., Neal, M. B., Newsom, J. T., Brockwood,
K. J., & Colton, C. L. (2005). A longitudinal study of the
effects of dual-earner couples’ utilization of family-
friendly workplace supports on work and family out-
comes. Journal of Applied Psychology, 90, 799 – 810.
Harrison, D. A., & McLaughlin, M. E. (1996). Structural
properties and psychometric qualities of organizational
self-reports: Field tests of connections predicted by cog-
nitive theory. Journal of Management, 22, 313–338.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit
indices in covariance structure analysis: Conventional
criteria versus new alternatives. Structural Equation
Modeling, 6, 1–55.
Hughes, E. L., & Parkes, K. R. (2007). Work hours and
well-being: The roles of work-time control and work-
family interference. Work and Stress, 21, 264 –278.
Ivancevich, J. M., & Matteson, M. T. (1976 –2005). Stress
diagnostic survey. Houston, TX: University of Houston.
Jaccard, J., & Wan, C. K. (1996). LISREL approaches to
interaction effects in multiple regression. Thousand
Oaks, CA: Sage.
James, L., Gent, M., Hater, J., & Corey, K. (1979). Corre-
lates of psychology influence: An illustration of the psy-
chological climate approach to work environment. Per-
sonnel Psychology, 32, 563–588.
Kelloway, E. K., Gottlieb, B. H., & Barham, L. (1999). The
source, nature, and direction of work and family conflict:
A longitudinal investigation. Journal of Occupational
Health Psychology, 4, 337–346.
Kinnunen, U., & Mauno, S. (1998). Antecedents and out-
comes of work-family conflict among employed women
and men in Finland. Human Relations, 51, 157–177.
Kline, P. (1986). A handbook of test construction: Introduc-
tion to psychometric design. New York: Methuen.
Kopelman, R. E., Greenhaus, J. H., & Connolly, T. F.
(1983). A model of work, family, and inter-role conflict:
 90 MATTHEWS, KATH, AND BARNES-FARRELL

A construct validation study. Organizational Behavior
and Human Performance, 32, 198 –215.
Kovner, C., Brewer, C., Wu, Y., Cheng, Y., & Suzuki, M.
(2006). Factors associated with work satisfaction of
registered nurses. Journal of Nursing Scholarship, 38,
71–79.
Matthews, R. A., Bulger, C. A., & Barnes-Farrell, J. L. (in
press). Work social supports, role stressors, and work-
family conflict: The moderating effect of age. Journal of
Vocational Behavior.
Miller, T. R., & Cleary, T. A. (1993). Direction of wording
effects in balanced scales. Educational and Psychologi-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
helps: Another look at the availability of flexible work
arrangements and work—family conflict. Journal of Vo-
cational Behavior, 71, 479 – 493.
Smith, P. C., & Stanton, J. M. (1998). Perspectives on the
measurement of job attitudes: The long view. Human
Resources Management Review, 8, 367–386.
Spector, P. E., Allen, T. D., Poelmans, S., Lapierre, L. M.,
Cooper, C. L., O’Driscoll, M. P., et al. (2007). Cross-
national differences in relationships of work demands,
job satisfaction, and turnover intentions with work-
family conflict. Personnel Psychology, 60, 805– 835.
Stanton, J. M., Sinar, E. F., Balzer, W. K., & Smith,

cal Measurement, 53, 51– 60. P. C. (2002). Issues and strategies for reducing the
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Netemeyer, R. G., Boles, J. S., & McMurrian, R. (1996).
Development and validation of work-family conflict and
family-work conflict scales. Journal of Applied Psychol-
ogy, 81, 400 – 410.
O’Driscoll, M. P., Poelmans, S., Spector, P. E., Kalliath, T.,
Allen, T. D., Cooper, C. L., et al. (2003). Family-
responsive interventions, perceived organizational and
supervisor support, work-family conflict, and psycholog-
ical strain. International Journal of Stress Management,
10, 326 –344.
Rantanen, J., Kinnunen, U., Feldt, T., & Pulkkinen, L.
(2008). Work-family conflict and psychological well-
being: Stability and cross-lagged relations within one-
and six-year follow-ups. Journal of Vocational Behavior,
73, 37–51.
Rizzo, J. R., House, R. J., & Lirtzman, S. I. (1970). Role
conflict and ambiguity in complex organizations. Admin-
istrative Science Quarterly, 15, 150 –163.
Rothman, K. J. (1990). No adjustments are needed for
multiple comparisons. Epidemiology, 1, 43– 46.
Shockley, K. M., & Allen, T. D. (2007). When flexibility
length of self-report scales. Personnel Psychology, 55,
167–194.
Thiagarajan, P., Chakrabarty, S., & Taylor, R. D. (2006). A
confirmatory factor analysis of Reilly’s role overload
scale. Educational and Psychological Measurement, 66,
657.
van Steenbergen, E. F., Ellemers, N., & Mooijaart, A.
(2007). How work and family can facilitate each other:
Distinct types of work-family facilitation and outcomes
for women and men. Journal of Occupational Health
Psychology, 12, 279 –300.
Witt, L. A., & Carlson, D. S. (2006). The work-family
interface and job performance: Moderating effects of
conscientiousness and perceived organizational support.
Journal of Occupational Health Psychology, 11, 343–
357.
Received March 5, 2008
Revision received July 9, 2009
Accepted July 28, 2009 y

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at http://notify.apa.org/ and you will be
notified by e-mail when issues of interest to you become available!