Content uploaded by Russell A. Matthews
Author content
All content in this area was uploaded by Russell A. Matthews on Apr 29, 2015
Content may be subject to copyright.
Journal of Occupational Health Psychology
Developing and Investigating the Use of Single-Item
Measures in Organizational Research
Gwenith G. Fisher, Russell A. Matthews, and Alyssa Mitchell Gibbons
Online First Publication, April 20, 2015. http://dx.doi.org/10.1037/a0039139
CITATION
Fisher, G. G., Matthews, R. A., & Gibbons, A. M. (2015, April 20). Developing and
Investigating the Use of Single-Item Measures in Organizational Research. Journal of
Occupational Health Psychology. Advance online publication.
http://dx.doi.org/10.1037/a0039139
Developing and Investigating the Use of Single-Item Measures in
Organizational Research
Gwenith G. Fisher
Colorado State University
Russell A. Matthews
Bowling Green State University
Alyssa Mitchell Gibbons
Colorado State University
The validity of organizational research relies on strong research methods, which include effective
measurement of psychological constructs. The general consensus is that multiple item measures have
better psychometric properties than single-item measures. However, due to practical constraints (e.g.,
survey length, respondent burden) there are situations in which certain single items may be useful for
capturing information about constructs that might otherwise go unmeasured. We evaluated 37 items,
including 18 newly developed items as well as 19 single items selected from existing multiple-item scales
based on psychometric characteristics, to assess 18 constructs frequently measured in organizational and
occupational health psychology research. We examined evidence of reliability; convergent, discriminant,
and content validity assessments; and test–retest reliabilities at 1- and 3-month time lags for single-item
measures using a multistage and multisource validation strategy across 3 studies, including data from
N⫽17 occupational health subject matter experts and N⫽1,634 survey respondents across 2 samples.
Items selected from existing scales generally demonstrated better internal consistency reliability and
convergent validity, whereas these particular new items generally had higher levels of content validity.
We offer recommendations regarding when use of single items may be more or less appropriate, as well
as 11 items that seem acceptable, 14 items with mixed results that might be used with caution due to
mixed results, and 12 items we do not recommend using as single-item measures. Although multiple-item
measures are preferable from a psychometric standpoint, in some circumstances single-item measures can
provide useful information.
Keywords: single-item measures, survey research methods, survey measurement, work–family conflict,
occupational health
The field of occupational health psychology has grown expo-
nentially over the past three decades, contributing a great deal of
knowledge about worker health and well-being. However, re-
searchers have issued a call for more variety in the types of
research designs, including longitudinal research, within-person
designs (e.g., diary studies or experience sampling), triangulation
of qualitative and quantitative research, and research involving
more diverse populations and occupations (Casper, Eby, Bor-
deaux, Lockwood, & Lambert, 2007; Gonzalez-Morales, Tetrick,
& Ginter, 2013; Kossek, Baltes, & Matthews, 2011).
Unfortunately, there are a number of practical barriers to
many research designs important for advancing occupational
health psychology. For example, diary studies involve frequent
repeated measurements, longitudinal research requires sus-
tained contact with participants over time, and sampling work-
ers from a wide variety of occupations (e.g., hourly workers)
can be challenging and costly. Employers may also impose time
constraints when surveying employees on work time. Further-
more, survey methodologists have demonstrated declines in
survey research participation in the population, which can ham-
per response rates, increase attrition, and negatively affect the
generalizability of research results (Groves, 2004). Demands
placed on participants are already high, requiring researchers to
either assess fewer constructs, or assess constructs of interest
with as few items as possible to minimize respondent burden
and attrition and maximize response rates. Therefore, surveys
should be designed to avoid any unnecessary redundancy. Us-
ing valid measures with fewer items not only reduces assess-
ment time, it also reduces the potential for participant fatigue,
frustration, and the likelihood of potential respondents refusing
to participate because they feel the survey is too long and time
consuming (Burisch, 1984). In studies involving many vari-
ables, researchers may consider the use of single-item measures
of some constructs.
Gwenith G. Fisher, Department of Psychology, Colorado State Univer-
sity; Russell A. Matthews, Department of Psychology, Bowling Green
State University; Alyssa Mitchell Gibbons, Department of Psychology,
Colorado State University.
We thank Lauren Cotter for her assistance, Mark Nagy and Steven
Poelmans for their comments and suggestions on a previous version of this
article, and the subject matter experts who provided useful item content
validity ratings.
Correspondence concerning this article should be addressed to Gwenith
G. Fisher, Department of Psychology, Colorado State University, 228
Behavioral Sciences Building, 1876 Campus Delivery, Fort Collins, CO
80523-1876. E-mail: gwen.fisher@colostate.edu
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Occupational Health Psychology © 2015 American Psychological Association
2015, Vol. 20, No. 3, 000 1076-8998/15/$12.00 http://dx.doi.org/10.1037/a0039139
1
The purpose of the present study is to examine the psychometric
properties of single items to assess whether they may be useful for
assessing psychological constructs in some areas of occupational
health research, and if so, to provide recommendations regarding
specific items that may be effective. The majority of research in
occupational health psychology relies on the use of survey data
and correlational research methods. As a result, reliability and
validity are essential psychometric properties to accurately inter-
pret research results and avoid increased risk of Type II error. We
are not advocating that single-item measures are interchangeable
with well-developed multiple-item measures, nor should they be
used as the measures for variables central to a research study, such
as primary measures used to evaluate an intervention. However,
we purport that there are situations in which rigorously developed
single items may be useful for capturing information about con-
structs that might otherwise go unmeasured. Rather than abstaining
from the use of single-item measures at all costs, our aim is to
provide researchers with an empirical evaluation of specific single
items measuring common constructs of interests in occupational
health psychology research. To accomplish this aim we compare
two approaches to the development of such items— choosing the
best-performing item from an existing multiitem measure and
creating a new item to capture the construct of interest in a holistic
way.
We organize this article as follows: first we review criticisms of
single-item measures. Next we present a number of reasons why
researchers may use single-item measures in spite of these criti-
cisms. We discuss the strengths and limitations of the two strate-
gies for identifying single-item measures (i.e., developing new
single-item measures, or adapting single items from existing mea-
sures). We then describe steps undertaken to develop and evaluate
single-item measures derived from each of these strategies. We
present results from multiple methods and multiple samples in
which we evaluate whether and which single-item measures may
be an acceptable or viable alternative for some topics in occupa-
tional health psychology research. We conclude with recommen-
dations regarding single item use in future research.
Criticisms of Single-Item Measures
Wanous, Reichers, and Hudy (1997) indicated that single-item
measures can be classified into two general categories: (a) those
asking about general demographic-type information (e.g., age,
tenure, number of children); and (b) measures of psychological-
type constructs (e.g., organizational commitment, domain satisfac-
tion). Although the use of single-item measures to assess con-
structs within the first category is normally deemed acceptable,
single-item measures of attitudes, knowledge, skills, or abilities are
generally discouraged (Wanous & Hudy, 2001).
Many well-trained researchers have a visceral reaction to even
the mention of using single-item measures. Concern over the use
of single-item measures to assess psychological constructs is a
long-standing issue (e.g., Gardner, Cummings, Dunham, & Pierce,
1998; Loo, 2002; Schriesheim, Hinkin, & Podsakoff, 1991). Most
researchers are taught in graduate school to abstain from using
single-item measures, though we may or may not have learned
about the psychometric issues associated with using single-item
measures (e.g., Nunnally & Bernstein, 1978).
Two criticisms in particular have dominated past discussions.
First, single-item measures may not adequately represent the con-
tent domain of conceptually complex constructs; single-item mea-
sures are not content valid due to having criterion deficiency
(Cronbach & Meehl, 1955; Nunnally & Bernstein, 1978;
Schriesheim et al., 1991). Second, single-item measures are gen-
erally deemed as unreliable because internal consistency estimates
of reliability cannot be calculated (Nagy, 2002; Nunnally & Bern-
stein, 1978). When single items are written ad hoc for a particular
study, it is impossible to separate true score variance from error or
to know the degree to which the item converges with other mea-
sures of the same construct. Thus, the reliability and validity of
single-item measures is generally unknown, making it difficult to
argue that they are accurate representations of the construct of
interest.
Acceptable Single-Item Measures
Although the aforementioned criticisms are neither trivial nor
unfounded, there are nevertheless areas of research relevant to
occupational health psychology in which single item measures
have attained a degree of respectability (e.g., job satisfaction,
Wanous & Hudy, 2001; Wanous, Reichers, & Hudy (1997); over-
all self-rated health, Idler & Benyamini, 1997). In fact, Scarpello
and Campbell (1983) compared single-item measures of global job
satisfaction with facet measures and concluded that a global single
item is more inclusive than summing across the facets. Similarly,
a global single-item measure of overall perceived health is com-
monly used in health, epidemiological, and interdisciplinary re-
search. Prior epidemiological research has shown that this item is
related to physical and mental health, and predictive of mortality
(DeSalvo, Bloser, Reynolds, He, & Muntner, 2006; Idler & Be-
nyamini, 1997; Molarius & Janson, 2002). Other constructs that
have been assessed successfully with single-item measures include
social support (Blake & McKay, 1986), job insecurity (Sverke,
Hellgren, & Näswall, 2002), bullying (Sawyer, Bradshaw, &
O’Brennan, 2008), self-esteem (Robins, Hendin, & Trzesniewski,
2001), and stress symptoms (Elo, Leppänen, & Jahkola, 2003).
The success of these single-item measures is not accidental; these
items were thoughtfully developed by subject matter experts and
there is a substantial base of empirical evidence that can be used to
evaluate their psychometric quality.
More recently, Fuchs and Diamantopoulos (2009) offered gen-
eral recommendations regarding circumstances under which
single-item measures might be more applicable in organizational
research. In particular, they suggested that single-item measures
are more acceptable when constructs are concrete, unidimensional,
have high semantic redundancy, are used to assess moderator or
control variables rather than key dependent or independent vari-
ables, desired precision is low, and the sampled population is
diverse (Fuchs & Diamantopoulos, 2009). However, they did not
present their own empirical data as the basis for their recommen-
dations, nor did they evaluate or recommend specific items. As
such, additional research is warranted.
Reasons for Possible Use of Single-Item Measures
Although multiple item scales are more likely to have superior
psychometric properties, there are a number of compelling reasons
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2FISHER, MATTHEWS, AND GIBBONS
why researchers might consider the use of single items (e.g.,
minimizing respondent burden, reducing criterion contamination,
increasing face validity).
Minimizing Respondent Burden
Minimizing respondent burden in survey research is important
for many reasons, including facilitating research with respondent-
intensive research designs, such as longitudinal research and ob-
taining data from populations that may be difficult to access.
Wanous et al. (1997) argued that single-item measures are a more
efficient use of survey space. Prior research has demonstrated that
perceived survey burden, or survey length, has deleterious effects
on responses rates (e.g., Crawford, Couper, & Lamias, 2001).
Specifically, Woods and Hampson (2005) noted that individuals
who perceive little benefit to participating in a study or do not
think they have time to do so may be more inclined to participate
if the survey is shorter. Shortening the length of the survey can
boost the overall response rate and minimize any negative reper-
cussions of nonresponse bias (Rogelberg & Stanton, 2007; Stan-
ton, Sinar, Balzer, & Smith, 2002). One way to shorten the length
of a survey is to use single-item measures. In longitudinal designs,
the use of single-item measures is particularly advantageous be-
cause responding to them is less demanding, it can encourage more
careful responses from participants, and may alleviate possible
respondent bias (Fu, 2005). Metz et al. (2007) indicated that
individuals are more likely to prefer, and respond to, single-item
measures because of their interpretability, face validity, and gen-
eral lack of repetition, as compared with multiple-item measures.
Furthermore, the success of some complex research designs may
rely upon the use of short or single-item measures. For example,
Fuller et al. (2003) examined work stress using a time series design
that incorporated single-item measures of mood and job satisfac-
tion administered daily over a 4-month period. The use of these
single-item measures was useful for keeping the daily surveys
short to minimize respondent burden and attrition throughout the
duration of the study. In light of these various advantages, Robins,
Hendin, and Trzesniewski (2001) noted that well developed single-
item measures can effectively strike the balance between address-
ing practical concerns (e.g., development costs, survey length) and
obtain acceptable psychometric characteristics.
The use of single-item measures may help increase the variety in
the types of samples and populations that can be studied by
reducing survey length and administration time. Increasingly, re-
searchers are acknowledging that to advance the field we must
work to expand our focus of interest (e.g., Eby, Casper, Lockwood,
Bordeaux, & Brinley, 2005). Specific to work–family research, it
has been suggested we need to consider a more broad definition of
family (e.g., single-parent households, households with adult de-
pendent care responsibilities, extended family/multigenerational
households, households with no children, and homosexual cou-
ples; Parasuraman & Greenhaus, 2002). Others have suggested a
need to examine occupational health psychology issues beyond the
professional workforce (e.g., nonprofessionals, part-time workers,
self-employed workers, hourly workers, contingent workers, indi-
viduals working nontraditional shift schedules; e.g., Grandey, Cor-
deiro, & Michael, 2007). For example, for hourly workers, ques-
tionnaires that include single-item measures may be particularly
advantageous because they can be completed quickly, reducing the
amount of time employees are taken off the line (Stanton et al.,
2002).
Prior research has also suggested that we should incorporate
multiple informants in the data collection process (e.g., Matthews,
Del Priore, Acitelli, & Barnes-Farrell, 2006; Voydanoff, 2007).
For example, in addition to collecting responses from the individ-
ual of interest, data might also be collected from coworkers,
supervisors, and other family members (e.g., spouse, children,
other dependents). In situations where a researcher’s question
might be informed by collecting data from children, asking a series
of short single-item measures may be more feasible and compre-
hendible for young participants compared with adults (Mauthner,
1997). Further, Matthews et al. (2006) have demonstrated the
utility of single-item measures when conducting interviews with
dual-earner couples.
Reducing Criterion Contamination
Scarpello and Campbell (1983) indicated that a valid concern
with multiple item measures is that they are more likely to contain
construct-irrelevant items, and may also omit construct-relevant
items. This is an ongoing issue of concern within occupational
health psychology research (Bellavia & Frone, 2005; MacDermid,
2005; Tetrick & Buffardi, 2006). To this end, scholars have rec-
ommended that researchers, when appropriate, consider assessing
constructs as generally as possible (Bellavia & Frone, 2005). The
rationale is that using scale items that are worded more generally
may improve the clarity and reduce respondents’ confusion regard-
ing what is actually being operationalized, and therefore facilitate
more accurate interpretations of predictor-outcome relations. In
other words, there may be a cost associated with developing and
using measures that are too specific, particularly if we cannot
define nor measure constructs well if they become too specific.
Thus, for example, it is possible that a global work–family conflict
measure may accurately reflect overall perceptions of work–family
conflict when it is not important to the research to delve into
specific sources of conflict (e.g., time, strain, and behavior; Green-
haus & Beutell, 1985).
Increasing Face Validity
In addition to reducing survey length, there are other advantages
of using single-item measures. For example, single-item measures
are often viewed as having increased face validity (Wanous et al.,
1997). Often, multiple-item measures, particularly those with the
highest internal consistency estimates, consist of items that are
very similar in focus. However, this similarity can be interpreted as
redundancy on the part of respondents, resulting in resentment and
a reduced willingness to provide accurate responses (Wanous et
al., 1997).
Finally, the implementation of valid single-item measures may
allow researchers to test more holistic or thorough models of
relations among constructs in organizational research. For exam-
ple, Voydanoff (2007) indicated that a significant limitation of
many extant work–family models is that these models fail to assess
important conceptual constructs. The reason is relatively straight-
forward: in any given study, researchers are often limited by the
number of questions they can ask. Researchers therefore tend to
focus on core constructs of interest, and hope that excluded con-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
SINGLE-ITEM MEASURES
structs do not have an instrumental effect on the nature of the
relationship under consideration. Primary constructs may be mea-
sured using traditional multiitem scales, but perhaps secondary
constructs may be measured with single items.
Present Study
Based on past conceptual and empirical work, we developed and
evaluated single-item measures to assess several key constructs
studied frequently in organizational research, including occupa-
tional health psychology. We focused on constructs relevant to the
assessment of the work–family interface: work–family conflict and
work–family balance. Work–family conflict (including work-to-
family conflict and family to-work conflict) is by far one of the
most commonly studied constructs in the work–family literature,
and the concept of balance has been gaining popularity (Greenhaus
& Allen, 2011). We also examined single-item measures of five
key predictors of work–family outcomes: domain centrality, con-
trol within the domain, role overload, social support, and role
clarity (specifically for the work domain). These constructs repre-
sent some of the most consistent and commonly studied predictors
within the work–family literature (Byron, 2005; Eby et al., 2005).
Single-item measures of these constructs might prove particularly
advantageous for researchers who are not explicitly examining
these constructs, but wish to include one or more of these con-
structs within their analyses to ensure more accurate model spec-
ification (e.g., to rule out the third variable problem). Finally, we
included four outcomes: domain and overall life satisfaction, burn-
out, and depression.
We began by developing conceptual definitions for each of the
constructs, and then identified single-item measures in two ways.
First, based on recommendations from the scale development
literature to use a deductive approach to develop items based on
construct definitions (e.g., Clark & Watson, 1995; Hinkin, 1995),
we wrote new items to assess each of the 18 different constructs.
We sought to create these new single items for the specific purpose
of being a global, inclusive single-item indicators of the given
construct with high levels of content validity. However, new items
are subject to the criticism of unknown reliability and validity, and
may also be perceived as “reinventing the wheel” when a high-
quality, established multiitem measure already exists. Therefore,
we also identified popular and well-established multiitem scales
for each construct and used factor analysis, the most popular and
well-established data reduction technique in applied psychological
and organizational research, to choose single-item indicators of
each construct (Hinkin, 1995). Specifically, we followed the rec-
ommendation by Stanton, Sinar, Balzer, and Smith (2002) to select
single items with the highest factor loadings. We then empirically
examined both sets of single-item measures (i.e., generated in two
different ways) to determine which, if any, single-item measures
may be effective for research.
There are both advantages and disadvantages of selecting
single-item measures from existing scales. For example, if prior
data are available about existing measures, it might also be pos-
sible to estimate reliability and relations with other constructs.
Wanous and Hudy (2001) have argued that the communality of an
item can be interpreted as an estimate of the item’s internal
consistency, or the Spearman-Brown prophecy formula can be
used to estimate the reliability of a “typical” single-item from a
multiitem scale (McDonald, 1999). However, single items selected
from a multiple-item measure may be subject to the criticism of
criterion deficiency, as the individual items in longer scales are
seldom written to be comprehensive in themselves. However, choos-
ing items based on psychometric evidence is somewhat atheoretical; it i s
possible that the item with (e.g.) the highest factor loading may
represent only part of the construct domain, but the quantitative
evidence alone does not take this into account. It is our view that
there is uncertainty whether using newly developed or psychomet-
rically chosen existing items to derive single-item measures will
yield higher quality single items: The psychometric properties of
new items and the construct representativeness of psychometri-
cally chosen items are both unknown. If researchers had evidence
to address these unknowns, they could make informed decisions
about the quality of specific items, and could in time develop a
body of evidence regarding the properties and nomological net of
those individual items. Atkinson and Lennox (2006) described five
different types of measurement models (a multiple effect indicator
model, a multiple cause indicator model, a single effect indicator
model, a single cause indicator model, and a mixed multiple
indicator model). The assumptions underlying each type of model
and specifications of relations between items, scales, and con-
structs influences item selection and validation.
In the present study, we contribute to the literature by comparing
items derived from both approaches (i.e., new and existing items)
for 18 constructs common in organizational and occupational
health psychology research. This allowed us to more directly
compare items based on multiple effect indicator models and
single effect indicator models (Atkinson & Lennox, 2006). Next
we describe the process by which we selected and compared items,
present the results, and conclude with a set of recommendations
about the use of the single-item measures we developed and
evaluated.
First, in Study 1, we examined the convergence of the newly
written single-item measures (hereafter referred to as “new single
items”) with psychometrically chosen (hereafter referred to as
“existing”) items from multiple-item measures of the same con-
struct (Wanous et al., 1997). By definition, single items chosen
from multiitem measures on the basis of high factor loadings
should show high convergent correlations with their original
scales. However, it is an empirical question whether new single
items, written to capture a construct broadly, will show similarly
high convergent correlations with a multiitem measure of the same
construct, which may have been developed with a slightly different
definition in mind. If the new single-item measures the same
construct as the original scale, however, we would expect to see
the new item load onto the same common factor as the original
items (a standardized factor loading can be interpreted as the
correlation between the item and the common factor; McDonald,
1999).
Hypothesis 1: New single-item measures will show conver-
gence with multiple-item measures of the same construct, as
evidenced by substantial loadings on the same common factor.
We examined the internal consistency reliability of each single-
item measure by calculating the communality of each with the
corresponding multiitem scale, following Wanous and Hudy
(2001). It is well known that, other factors being equal, internal
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4FISHER, MATTHEWS, AND GIBBONS
consistency increases with the length of a test (Cortina, 1993).
Thus, we do not expect that single-item measures will be as
internally consistent as multiple-item measures. However, a well-
developed or well-chosen single item might still show acceptable
levels of internal consistency reliability.
Hypothesis 2a: Psychometrically chosen single-item measures
will show at least moderate levels of internal consistency
reliability.
Hypothesis 2b: New single-item measures will show at least
moderate levels of internal consistency reliability.
Further, we examined and compared correlations among con-
structs using new single items, psychometrically chosen single
items, and the original multiitem scales. We used a multitrait–
multimethod matrix approach (Campbell & Fiske, 1959) to eval-
uate convergent and discriminant validity:
Hypothesis 3: New and psychometrically chosen single items
measuring the same construct should correlate more highly
with one another than with other single items intended to
assess other constructs.
One argument against the use of single-item measures is that,
because single-item measures are less reliable and unreliability
attenuates correlations (i.e., validity), using single-item measures
will weaken the observed correlations and reduce the overall
power of the study to find expected relationships (cf. Schmidt &
Hunter, 2014). However, the degree of attenuation in any one
situation depends on the reliability of the particular measures used.
Thus, to estimate the likely impact of using single-item measures
on researchers’ conclusions, we identified several common
predictor-outcome pairs from among the constructs we measured
and compared the predictor-outcome correlations obtained using
new single items, existing single items, and multiitem scales.
Hypothesis 4: Bivariate correlations among single-item mea-
sures are consistent with correlations among multiple item
measures of the same constructs.
Next, in Study 2, we evaluated content validity as well as
perceived research utility of single item measures for each con-
struct using ratings from occupational health subject matter ex-
perts. We expect new single items to have higher ratings of content
validity and research utility compared with items selected from
existing scales for two reasons. First, the new single item measures
we developed were written with the specific objective of being a
global, inclusive indicator of the construct. Second, a single item
selected from a multiple item measure may be criterion deficient
(i.e., individual items in longer scales are seldom written to be
comprehensive in themselves). Therefore, we hypothesize that:
Hypotheses 5– 6: Newly developed single items will have
higher ratings of content validity (Hypothesis 5) and research
utility (Hypothesis 6) based on subject matter expert ratings
than items chosen from multiple item scales.
Lastly, based on results from Study 2, as recommended by Nagy
(2002) as well as Gosling, Rentfrow, and Swann (2003), in Study
3 we conducted an exploratory investigation to examine test–retest
reliability estimates of the new single items across both 1-and
3-month lags. We examined test–retest reliability estimates across
more than one time lag because some inconsistency has been
observed in the work–family literature when using different lags
between assessments (for discussion see Matthews, Wayne, &
Ford, 2014), and other occupational health research has high-
lighted the importance of examining time lags (Ford et al., 2014).
We examined test–retest reliability as a research question (i.e., to
what extent are the single-item measures stable over time?) rather
than stating an a priori hypothesis about the level of stability
because some constructs included in our study are believed to be
more stable or enduring traits (e.g., domain centrality) and others
assess more transient states (e.g., work–family conflict).
Study 1: Item Development
We began by using prior reviews of the literature (e.g., Eby et
al., 2005; Greenhaus & Allen, 2011) to develop initial conceptual
definitions for 18 constructs. We then wrote initial single-item
measures of each construct based on these conceptual definitions
and sent both the definitions and initial new single items to four
PhD level subject matter experts (SMEs). SMEs were identified
based on recent publications relevant to occupational health psy-
chology; all held doctoral degrees. As a group, these SMEs had
published over 35 related articles and book chapters. Each subject
matter expert was e-mailed a document, which included the con-
ceptual definitions and new single items and was asked to evaluate
each conceptual definition as well as provide feedback on the new
single-item operationalizations of the constructs. Based on pro-
vided feedback we reworded several conceptual definitions and
single-item measures to increase clarity. The final conceptual
definitions and new single-item operationalizations of these 18
constructs appear in Table 1. As previously explained, we also
examined single items selected from existing multiple item scales,
and compare single versus multiple item measures of the same
constructs. The single items chosen from existing scales are also
listed in Table 1.
Method
Participants and procedure. A total of 384 individuals par-
ticipated in an online survey to obtain evidence about the relations
among the single-item measures and related multiple item mea-
sures. Using a method similar to Matthews, Bulger, and Barnes-
Farrell (2010), a total of 53 trained undergraduate student recruit-
ers from advanced psychology classes from four universities
assisted with the data collection process. Two of the universities
were large public institutions in the United States with one located
in the South, and one located on the west coast. The two remaining
school were smaller regional public university in the Midwest.
Students were trained on the data collection methodology and
ethics in research. They were provided with an e-mail invitation
that they distributed to working adults they personally knew who
met the eligibility requirements for the study (at least 18 years old
and working at least 15 hours per week). Recipients of the invi-
tation e-mails were asked to follow the Web-link supplied in the
e-mail and complete the online survey. The survey took approxi-
mately 15 min to complete; participation was voluntary. Student
recruiters received nominal course extra credit for their involve-
ment.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
SINGLE-ITEM MEASURES
Table 1
Conceptual Definitions, Single Items, and Subject Matter Expert Ratings of Item Content Validity (Studies 1 and 2)
Construct Definition Item
SME content
validity
represents
definition
c
SME content
validity
captures all
content
d
SME research
utility
assessment
a
MSDMSDMSD
Work centrality Degree to which an individual
perceives that work is an
important component of
their current self-identity.
1. My work is one of the
most important things
in my life right now.
a
4.12 .86 4.06 .83 4.00 .94
2. To me, my job is a
very large part of who
I am.
b
3.94 .97 3.94 .97 3.94 .97
Family centrality Degree to which an individual
perceives that family is an
important component of
their current self-identity.
3. My family is one of
the most important
things in my life right
now.
a
4.00
ⴱ
1.17 4.00
ⴱ
1.17 4.00
ⴱ
1.17
4. I am very much
personally involved
with my family.
b
3.00 1.12 3.00 1.06 2.82 1.13
Coworker support An interpersonal transaction
involving emotional
concern, instrumental aid,
information, or appraisal
from an individual’s
coworkers.
5. I can count on my
coworkers/work
colleagues for support
when I need it.
a
4.18
ⴱ
0.73 3.94
ⴱ
0.83 4.12
ⴱ
0.86
6. (To what extent can
you...)Count on
your coworkers to
back you up at work?
b
2.71 0.92 2.65 0.70 2.53 0.94
Supervisor
support
An interpersonal transaction
involving emotional
concern, instrumental aid,
information, or appraisal
from an individual’s
supervisor.
7. I can count on my
supervisor/manager for
support when I need
it.
a
4.35
ⴱ
0.86 4.35
ⴱ
0.86 4.29
ⴱ
0.99
8. My supervisor tries to
make my job as
interesting as
possible.
b
1.94 0.83 1.71 0.59 1.59 0.62
Personal/family
support
An interpersonal transaction
involving emotional
concern, instrumental aid,
information, or appraisal
from an individual’s
personal or family life.
9. I can count on my
friends/family
members for support
when I need it.
a
3.94
ⴱ
0.83 3.82
ⴱ
0.81 3.76
ⴱ
0.90
10. Please rate the extent
to which....youfeel
that [your family
members] were really
trying to understand
your problems?
b
2.59 0.80 2.59 0.80 2.06 0.90
Work role clarity The degree to which
individuals feel they have
clear guidance about
expected roles job-related
and behaviors associated.
11. I have a clear
understanding of what
is expected of me in
my job.
a
4.53
ⴱ
0.51 4.35
ⴱ
0.70 4.41
ⴱ
0.62
12.(Inmyjob...)I
know what my
responsibilities are.
b
3.71 1.10 3.71 0.85 3.53 1.12
Job control The extent to which an
individual has autonomy to
make decisions about how
and when to complete job-
related tasks.
13. I feel I have a lot of
control over how I do
my job.
a
3.88
ⴱ
0.99 3.76
ⴱ
0.97 3.76
ⴱ
1.09
14. (To what extent do
you...)Plan your
own work?
b
2.94 0.83 2.82 0.88 2.76 0.83
Personal/family
life control
The extent to which an
individual has autonomy to
make decisions about how
and when to complete
personal life-related tasks.
15. I feel I have a lot of
control over things in
my personal/family
life.
a
4.00
ⴱ
0.87 3.94
ⴱ
0.90 3.82
ⴱ
1.01
16. (To what extent do
you...)Plan your
own [personal life]
activities?
b
2.29 1.31 2.24 1.15 2.06 1.30
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6FISHER, MATTHEWS, AND GIBBONS
Table 1 (continued)
Construct Definition Item
SME content
validity
represents
definition
c
SME content
validity
captures all
content
d
SME research
utility
assessment
a
MSDMSDMSD
Work role
overload
Defined as occurring when an
individual perceives they do
not have the necessary
resources to meet all the
role demands within the
work domain.
17. I often feel I am
unable to meet all the
demands related to my
work.
a
3.65 1.06 3.76
ⴱ
0.97 3.59 1.12
18. I cannot ever seem to
catch up (at work).
b
3.29 0.99 3.18 0.88 3.35 1.00
Family role
overload
Defined as occurring when an
individual perceives they do
not have the necessary
resources to meet all the
role demands within the
family domain.
19. I often feel I am
unable to meet all the
demands in my
personal/family life.
a
3.88
ⴱ
1.27 3.71 1.16 3.71
ⴱ
1.21
20. I need more hours in
the day to do all the
things that are
expected of me (at
home).
b
3.18 1.19 3.12 1.11 3.06 1.20
Work-to-family
conflict
A form of inter-role conflict
where the demands of
functioning in the work
domain are incompatible
with functioning in the
family domain.
21. In the past month my
WORK life frequently
interfered with my
PERSONAL/FAMILY
life.
a
4.12
ⴱ
0.60 4.12
ⴱ
0.86 3.94
ⴱ
0.90
22. I have to miss family
activities due to the
amount of time I must
spend on work
responsibilities.
b
3.24 1.09 3.18 1.01 2.88 1.22
23. I am often so
emotionally drained
when I get home from
work that it prevents
me from contributing
to my family.
b
3.00 1.00 2.88 0.78 2.82 1.07
Family-to-work
conflict
A form of inter-role conflict
where the demands of
functioning in the family
domain are incompatible
with functioning in the
work domain.
24. In the past month my
PERSONAL/
FAMILY life
frequently interfered
with my WORK life.
b
3.65 1.06 3.76
ⴱ
1.15 3.71
ⴱ
1.10
25. I have to miss work
activities due to the
amount of time I must
spend on family
responsibilities.
a
3.06 1.30 3.06 1.09 3.06 1.30
26. Because I am often
stressed from family
responsibilities, I have
a hard time
concentrating on my
work.
b
3.12 1.17 2.88 0.99 2.71 1.05
Work/family
balance
An individual’s global
assessment that work and
family demands are
adequately met with
existing resources such that
participation is effective in
both domains.
27. In general I feel that I
have an adequate
balance between my
work and personal/
family life.
a
3.71
ⴱ
1.16 3.82
ⴱ
1.29 3.65
ⴱ
1.32
28. It is clear to me, based
on feedback from
coworkers and family
members, that I am
accomplishing both my
work and family
responsibilities.
b
2.71 1.10 2.71 1.05 2.59 1.18
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
SINGLE-ITEM MEASURES
To be included in the present study participants had to be
employed by an organization and not self-employed, working at
least 15 hours a week, and have complete data at the item level.
Based on these criteria, 44 participants were identified for exclu-
sion. Descriptive statistics for demographic variables for the re-
maining sample of n⫽340 are presented in Table 2.
Measures.
Single-item measures. Participants were asked to respond to
the 18 new single-item measures presented in Table 1 based on a
7-point Likert-type response scale (from 1 ⫽strongly disagree to
7⫽strongly agree).
Multiple item measures. Unless otherwise indicated, all re-
sponses to the measures were made on a 7-point Likert-type
response scale (from 1 ⫽strongly disagree to 7 ⫽strongly agree).
Example items from each scale appear in Table 1.
Centrality. Work centrality (␣⫽.88) was assessed with five
items of a modified version of a measure adapted by Matthews,
Swody, and Barnes-Farrell (2012). Family centrality (␣⫽.91)
was assessed using five items (Matthews et al., 2012).
Social support. Coworker support (␣⫽.90) was assessed
with a 4-item measure (Haynes, Wall, Bolden, Stride, & Rick,
1999). Supervisor support (␣⫽.88) was assessed with a 3-item
measure (Eisenberger, Stinglhamber, Vandenberghe, Sucharski, &
Rhoades, 2002). Personal/family support (␣⫽.91) was assessed
with 6-item measure (Winefield, Winefield, & Tiggemann, 1992).
Respondents were given the prompt, “Please think about your
family and close friends, especially the 2–3 who are most impor-
tant to you. Thinking about the past month, how often did...”
Responses were made on a 5-point frequency scale (1 ⫽never to
5⫽always).
Table 1 (continued)
Construct Definition Item
SME content
validity
represents
definition
c
SME content
validity
captures all
content
d
SME research
utility
assessment
a
MSDMSDMSD
Life satisfaction An individual’s overall
assessment of feelings and
attitudes about one’s life at
a particular point in time.
29. As a whole, I am
satisfied with my life.
a
4.71
ⴱ
0.47 4.65
ⴱ
0.49 4.65
ⴱ
0.49
30. In most ways my life
is close to ideal.
b
3.71 0.99 3.71 1.05 3.65 1.17
Personal/family
life satisfaction
An individual’s global
assessment that they enjoy
their personal/family life
31. Overall, I am satisfied
with my
personal/family life.
a
4.47 0.87 4.41 0.71 4.35 0.93
Job satisfaction An individual’s global
assessment that they enjoy
their job, they do it well,
and that they are suitably
rewarded for their efforts.
32. Overall, I am satisfied
with my job.
a
4.35 1.06 4.29 0.99 4.29 1.10
33. All in all I am satisfied
with my job.
b
4.59 0.87 4.41 0.94 4.47 0.80
Burnout An individual’s personal
assessment of emotional
exhaustion,
depersonalization of others,
and a feeling of reduced
personal accomplishment.
34. Feelings of burnout
occur when an
individual feels
emotional exhausted,
cynical, and feels a
lack of personal
accomplishment. In the
past month, how often
have you experienced
feelings of burnout?
a
3.94
ⴱ
1.20 4.06
ⴱ
1.09 3.53 1.23
35. I feel burned out.
b
3.18 1.24 3.06 1.09 3.00 1.22
Perceived
depression
An individual’s perception
that they are suffering from
a psychological disorder
affecting their mood,
physical functions, and
social interactions.
36. Depression is
considered to exist
when an individual
feels sad, has trouble
sleeping, lacks
motivation, feels
worthless, is
withdrawn, and is
generally fatigued. In
the past month, how
often have you felt
depressed?
a
4.24
ⴱ
0.97 4.35
ⴱ
0.79 4.00
ⴱ
1.22
Perceived
depression
37. Much of the time
during the past week
you felt depressed.
b
3.24 1.20 3.06 1.14 3.00 1.22
Note. N ⫽17.
a
Newly written single-item.
b
Single-item from existing scale.
c
Evaluations made using a 5-point scale (1 ⫽poor to5⫽excellent).
d
Evaluations
made using a 5-point scale (1 ⫽not at all to 5 ⫽to a great extent).
ⴱ
Indicates statistically significant difference (p⬍.05) in paired t-test comparing SME ratings of new(
a
) vs. existing(
b
) single-items.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8FISHER, MATTHEWS, AND GIBBONS
Work role clarity. Work role clarity (␣⫽.85) was assessed
with a six-item measure by Rizzo, House, and Lirtzman (1970).
Control. Job control (␣⫽.88) was assessed with a six-item
measure by Haynes et al. (1999). Respondents were given the
prompt, “At work, to what extent do you...”Personal/personal/
family life control (␣⫽.92) was assessed with a parallel set of six
items, but the target of the prompt was changed to ask about their
personal/family life.
Role overload. Work overload (␣⫽.88) was assessed with a
five-item measure by Matthews, Kath, and Barnes-Farrell (2010).
Family overload (␣⫽.89) was assessed with a parallel set of five
items. Participants were first directed to respond to the items
thinking about the work domain, and then again considering the
family/home-life domain.
Work–family interface. Work-to-family conflict (␣⫽.72) was
assessed with a three-item measure (Matthews, Kath, et al., 2010).
A parallel set of items was used to assess family-to-work conflict
(␣⫽.67). Work–family balance (␣⫽.94) was assessed with a
six-item measure (Carlson, Grzywacz, & Zivnuska, 2009).
Satisfaction. Life satisfaction (␣⫽.89) was assessed with a
five-item measure (Diener, Emmons, Larsen, & Griffin, 1985). Job
satisfaction (␣⫽.91) was assessed with three items (Cammann,
Fichman, Jenkins, & Klesh, 1983). Personal/family life satisfac-
tion (␣⫽.81) was measured with a five-item measure (Matthews,
Kath, et al., 2010).
Burnout. Burnout (␣⫽.92) was measured with a 14-item
measure (Shirom & Melamed, 2006) that assessed physical fatigue
(six items), emotional exhaustion (three items), and cognitive
weariness (five items). For the purposes of this study, we report
results using the physical fatigue dimension.
Perceived depression. Perceived depression (␣⫽.83) was
assessed with an eight-item version of the Center for Epidemio-
logical Studies of Depression scale (Radloff, 1977; Steffick, 2000).
Responses were made on a 5-point frequency scale (1 ⫽never to
5⫽always).
Results
Single items were selected from existing measures by conduct-
ing exploratory factor analyses with maximum likelihood extrac-
tion for each scale and choosing the item with the highest loading.
1
To test Hypothesis 1 regarding the convergence of the new single-
item measures with the corresponding multiple-item measures, we
fit single factor models to each of the original multiitem scales,
including the corresponding new single item. This allowed us to
obtain standardized factor loadings for the new items, which
estimate the correlation of the item with the common factor (Mc-
Donald, 1999). Standardized loadings can also be used to calculate
the communality of the item (i.e., the proportion of variance in the
item that is shared with the other items), which can be interpreted
as an estimate of the reliability (i.e., shared or true-score
variance) of the item (Wanous & Hudy, 2001). We also com-
pared the loading of each single-item measure with the loadings
of the original items as another indicator of how well the single
1
For work-to-family conflict and family-to-work conflict, both time-
and strain-based conflict items are shown in Table 1 because they were
both included in Study 2 (content validation). However, the strain-based
conflict items (Items 23 and 26 in Table 1) had higher factor loadings and
were therefore used as the existing single items used in Study 1.
Table 2
Demographic Characteristics for Study 1 and Study 3 Samples
Study 1 Study 3
Full sample
(n⫽340)
Primary sample
(n⫽992)
Test–retest sample
(n⫽302)
M SD M SD M SD
Age 37.13 13.64 36.03 12.91 38.06 12.16
Number of children under 18 .45 .87 .68 .99 .74 1.02
Organizational tenure 6.84 8.36 6.61 8.16 7.16 8.01
Work hours per week 39.87 12.96 40.33 12.33 41.91 12.01
Female 69.1% 58.1% 67.7%
Marital status
Married/living with partner 52.1% 54.0% 62.9%
Single 39.1% 33.4%
Education
High school/GED or less 7.1% 10.9% 6.9%
Some college or associate’s 45.6% 32.2% 28.1%
Bachelor’s 35.8% 38.5% 40.1%
Master’s or beyond 11.6% 18.4% 24.9%
Occupation
Management, business, or financial operations
related occupations 13.8% 21.4% 21.2%
Professional and related occupations 34.4% 23.1% 30.1%
Production, installation, maintenance, repair,
and service occupations 18.5% 17.1% 12.4%
Household income
Under $50,000 38.7% 39.3% 31.2%
$50,001–$100,000 37.1% 33.2% 36.9%
$100,001 or more 24.2% 28.3% 31.9%
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
SINGLE-ITEM MEASURES
items fit into the scales. Results of these analyses are summa-
rized in Table 3.
Overall, the single-factor models including the new single items
showed acceptable to good fit for most of the scales, with most
CFIs above .90 and many above .95, and standardized root mean
square residuals (SRMR) below .06 in all but one case. The
RMSEA values for most scales were higher than is typically
desired (observed range: .05–.23); however, RMSEA is often
inflated in models with small numbers of variables (Kenny &
McCoach, 2003), as was the case for the short scales used here.
Another set of factor analyses, which excluded the new single
items from each measure, suggested that the new single items were
not the primary sources of the high RMSEA values.
2
There was considerable variability in the standardized loadings
of the new single items on their respective scales. The highest
loadings were for the new single items measuring job satisfaction
(⫽.90) and life satisfaction (⫽.86). In contrast, the loading
for the new single item measuring family control (⫽.29) did not
reach the threshold of .30; in applied research loadings equal to or
greater than .30 and .40 are generally interpreted as “salient”
(Brown, 2006). As shown in Table 3, 10 of the single-item mea-
sures had factor loadings greater than the smallest loading among
the multiple item measures, suggesting that these single items
performed comparably to the items of the multiple-item measure.
However, in no case did the new single item outperform the
existing item with the highest factor loading on the original scale.
Altogether we conclude that results indicated partial support for
Hypothesis 1, which proposed that new single-item measures
would load highly onto the same common factor as multiple-item
measures of the same construct.
Reliability. The communalities of both new and existing sin-
gle items showed considerable variability (see Table 4). For the
existing single items, the highest communality was .88 (work–
family balance) and the lowest was .56 (family-to-work conflict).
For the new single items, communalities ranged from a high of .81
(job satisfaction) to a low of .08 (personal/family control). If we
interpret communalities as an index of single-item reliability
(Wanous & Hudy, 2001), only four of the new single-item mea-
sures met the traditional .70 threshold for “acceptable” reliability
in a research context (Nunnally & Bernstein, 1978). Fourteen of
the existing single items met this threshold. However, five new
single items and four existing single items had communalities
between .50 and .70, and researchers in some contexts (e.g.,
Fleeson, 2001) have made the argument that reliabilities in this
range are meaningful. As reliability coefficients represent the
proportion of total variance that can be attributed to true score,
rather than error (Raykov & Marcoulides, 2011), a reliability
coefficient above .50 might reasonably be interpreted as indicating
more signal than noise. The results presented in Table 4 suggest
that most of the existing single items showed acceptable reliability,
supporting Hypothesis 2a, and some of the new single items did as
well, partially supporting Hypothesis 2b.
Convergent and discriminant validity. Next, we examined
the correlations among different types of measures of the same
construct (multiple item, existing single item, and new single item)
and among different constructs measured using the same method
of item selection (see Table 5). The convergent correlations be-
tween multiple-item measures and existing single items were quite
high, as might be expected given that the existing items were
drawn from the multiple-item measures. The convergent correla-
tions between multiple-item measures and new single items were
more varied.
2
The fit statistics for the models with and without the new single items
were highly similar; for brevity, we report results for only the models
including the new items. Full results are available upon request from the
first author.
Table 3
Study 1 Model Fit and Factor Loadings for New Single-Items With Multiple-Item Measures
Model fit including single-item New single
item
Original items
rangeConstruct MSD# Items
2
df p CFI SRMR RMSEA
Work centrality 4.03 1.76 6 136.79 9 .00 .89 .06 .20 .71 .70–.86
Family centrality 6.30 1.07 6 71.20 9 .00 .96 .03 .14 .79 .76–.92
Coworker support 5.25 1.42 5 55.97 5 .00 .95 .03 .17 .72 .77–.88
Supervisor support 5.17 1.56 4 30.94 2 .00 .97 .03 .21 .84 .80–.90
Personal/family support 6.06 1.14 7 148.15 14 .00 .92 .05 .17 .50 .67–.92
Work role clarity 5.89 1.14 7 40.42 14 .00 .98 .03 .08 .70 .48–.88
Job control 5.44 1.38 7 97.39 14 .00 .93 .04 .13 .51 .56–.83
Personal/family control 5.18 1.45 7 108.64 14 .00 .94 .03 .14 .29 .71–.87
Work overload 3.22 1.65 6 50.76 9 .00 .96 .04 .12 .50 .66–.88
Family overload 3.71 1.73 6 20.60 9 .01 .99 .02 .06 .52 .67–.88
Work-to-family conflict 3.35 1.81 4 10.61 2 .01 .98 .03 .11 .70 .54–.79
Family-to-work conflict 2.87 1.66 4 3.93 2 .14 .99 .02 .05 .71 .52–.75
Balance 5.06 1.47 7 150.21 14 .00 .93 .05 .17 .62 .73–.94
Life satisfaction 5.59 1.25 6 65.88 9 .00 .96 .03 .14 .86 .60–.92
Personal/family satisfaction 5.68 1.36 6 58.00 9 .00 .94 .04 .13 .68 .53–.84
Job satisfaction 5.14 1.56 4 13.72 2 .00 .99 .02 .13 .90 .81–.93
Burnout 3.50 1.20 6 65.72 14 .00 .96 .04 .11 .58 .68–.90
Perceived depression 3.26 1.84 9 117.35 27 .00 .91 .05 .10 .66 .43–.83
Note. N ⫽331–340; # Items ⫽number of items in original multiple-item scale;
2
⫽-square test of model fit; df ⫽degrees of freedom; CFI ⫽
comparative fit index; SRMR ⫽standardized root mean square residual; RMSEA ⫽root mean squared error of approximation; ⫽factor loading for the
single-item measure; range ⫽minimum and maximum values for items in multiple item scale.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 FISHER, MATTHEWS, AND GIBBONS
For all items, the convergent correlations were higher than the
average discriminant correlations. For the existing single items, the
convergent correlations were always higher than any discriminant
correlation. This was also true for most of the new single items;
exceptions were coworker support, supervisor support, work con-
trol, personal/family control, life satisfaction, family satisfaction,
and burnout. Many of these exceptions can be explained by the
similarity among constructs; for example, coworker support and
supervisor support were correlated at r⫽.70, and life satisfaction
and family satisfaction were correlated at r⫽.83. Convergent and
discriminant correlations obtained in Study 1 provide support for
Hypothesis 3a and partial support for Hypothesis 3b.
Relationships with other variables. Finally, we compared
relations among specific pairs of constructs to examine whether
and how the relations changed with the use of single or multiple
item measures. To reduce the number of comparisons to a
manageable set, we identified a set of seven constructs that are
commonly treated as predictors in occupational health research
and a set of five constructs that are commonly treated as
outcomes (see Table 6). We calculated the correlation for each
predictor-outcome combination in three different ways: (a) us-
ing the original multiple-item measure of the predictor; (b)
using the existing single-item measure; and (c) using the new
single-item measure. We used the multiple-item measure of the
outcome in all cases to reduce the effects of attenuation due to
unreliability. We used Steiger’s (1980) test for the difference
between two dependent correlations (i.e., the correlations of
two separate variables with the same third variable) to test
whether one single-item measure was a better predictor than the
other. In 22 of the 35 predictor-outcome correlations we tested,
there were no significant differences between the correlations
involving the new and existing single items.
There were no significant differences between the correla-
tions for multiple item measures and existing single items for
19 of the 35 cases (see Table 6); multiple item measures were
superior in all but one of the rest (family-to-work conflict
predicting life satisfaction). Although the pattern of results
differed somewhat for different outcomes, there were some
Table 4
Study 1 Estimates of Internal Consistency and Communalities
for Multiple-Item and Single-Item Measures
Construct
Existing
multiple-
item scale
Existing single
item
New single
item
# Items ␣h
2
h
2
Work centrality 5 .87 .77 .50
Family centrality 5 .91 .84 .62
Coworker support 3 .88 .80 .51
Supervisor support 4 .90 .74 .70
Personal/family support 6 .91 .85 .25
Work role clarity 6 .85 .75 .50
Job control 6 .88 .70 .26
Personal/family control 6 .92 .74 .08
Work overload 5 .88 .77 .25
Family overload 5 .89 .77 .27
Work-to-family conflict 3 .72 .60 .49
Family-to-work conflict 3 .68 .56 .50
Balance 6 .93 .88 .70
Life satisfaction 5 .89 .78 .74
Personal/family satisfaction 5 .81 .73 .47
Job satisfaction 3 .89 .76 .81
Burnout 14 .92 .71 .27
Perceived depression 8 .83 .64 .44
Note. N ⫽331–340.
Table 5
Study 1 Convergent and Average Discriminant Correlations
Existing single item New single item
Convergent r
Discriminant r
Convergent r
Discriminant r
Mean Min Max Mean Min Max
Work centrality .88 .11 .01 .35 .67 .09 .01 .32
Family centrality .91 .16 .01 .47 .76 .16 .01 .38
Coworker support .91 .18 .06 .39 .56 .25 .10 .70
Supervisor support .91 .19 .00 .45 .48 .26 .08 .70
Personal/family support .88 .18 .00 .33 .49 .21 .01 .41
Work role clarity .85 .19 .08 .34 .63 .18 .05 .38
Job control .84 .15 .00 .46 .48 .19 .03 .51
Personal/family control .87 .18 .02 .46 .28 .26 .03 .64
Work overload .86 .16 .01 .45 .49 .12 .01 .24
Family overload .88 .16 .00 .45 .51 .19 .05 .36
Work-to-family conflict .83 .22 .01 .44 .58 .14 .02 .41
Family-to-work conflict .80 .19 .01 .37 .56 .17 .03 .41
Balance .91 .27 .01 .38 .65 .24 .03 .41
Life satisfaction .89 .25 .03 .42 .79 .30 .04 .83
Personal/family satisfaction .82 .21 .02 .47 .61 .29 .03 .83
Job satisfaction .93 .25 .01 .45 .85 .26 .07 .55
Burnout .74 .23 .05 .46 .56 .19 .01 .58
Perceived depression .81 .23 .02 .46 .58 .20 .05 .58
Average across constructs .86 .19 .00 .47 .59 .21 .01 .83
Note. Convergent r⫽correlation between single item and original multi-item measure. Discriminant r⫽absolute correlation between constructs
measured with the same type of item (existing single item or new single item).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
SINGLE-ITEM MEASURES
trends. For example, the multiple item measures for work
overload and job clarity were better predictors than the existing
single items for at least four of the five outcomes. The existing
single items for work-to-family conflict, family-to-work con-
flict, balance, and job control did not show significantly differ-
ent correlations compared with multiitem measures for most
outcomes.
Similarly, there were no significant differences between the
correlations for multiple item scales and new single items in 21 of
35 cases; multiple item measures were superior in all of the rest.
The pattern of results was different from that observed for existing
single items. Again, the multiitem measure of work-to-family
conflict was a better predictor than the new single item for four of
the five cases, but this pattern did not hold for family-to-work
conflict. Multiitem measures for balance and work overload had
stronger correlations than new single items for three of the five
outcomes; the new single item for job control consistently showed
no difference. Taken together, these results provide moderate
support for Hypothesis 4; single-item measures can show similar
correlations compared to multiitem measures, but it depends on the
construct and, to some extent, the outcome.
Brief Discussion
The purpose of Study 1 was to develop and evaluate the reli-
ability, convergent, and discriminant validity of single item mea-
sures. Our results provided partial support for our hypotheses. For
example, Hypotheses 2a and 3a were generally supported in that
single items selected from existing measures showed acceptable
internal consistency and good convergent and discriminant valid-
ity. Hypotheses 1, 2b, and 3b were supported for some new single
items but not for others, suggesting that more holistically oriented
single items can load highly on and show good internal consis-
tency and convergence with multiple-item measures, as well as
good discriminant validity, but that this is not always the case.
Further, there was mixed support for Hypothesis 4 for both new
and existing single items; items of both types showed correlations
with other variables that were sometimes consistent and sometimes
Table 6
Study 1 Predictor-Outcome Correlations Using Multiple-Item, Existing, and New
Single-Item Measures
Outcome Predictor Multi-item r
Existing single
item r
New single
item r
Life satisfaction Work-to-family conflict ⫺.23
a
⫺.23
a
⫺.06
Family-to-work conflict ⫺.28
a
⫺.35 ⫺.18
a
Balance .40
a
.38
a
.39
a
Job control .16
a
.16
a
.16
a
Work overload ⫺.16 ⫺.09
a
.00
a
Work role clarity .26 .15
a
.09
a
Family overload ⫺.33
a
⫺.21
b
⫺.29
a,b
Family satisfaction Work-to-family conflict ⫺.15
a
⫺.13
a
⫺.06
a
Family-to-work conflict ⫺.28
a
⫺.31
a
⫺.21
Balance .44
a
.46
a
.31
Job control .09
a
.03
a
.15
a
Work overload ⫺.11
a
⫺.12
a
⫺.14
a
Work role clarity .28
a
.22
a
.26
a
Family overload ⫺.23
a
⫺.14 ⫺.28
a
Job satisfaction Work-to-family conflict ⫺.30
a
⫺.25
a
⫺.09
Family-to-work conflict ⫺.18
a
⫺.12
a
⫺.12
a
Balance .40
a
.42
a
.35
a
Job control .42
a
.32 .48
a
Work overload ⫺.18
a
⫺.05
b
⫺.13
a,b
Work role clarity .47 .35
a
.32
a
Family overload ⫺.11
a
⫺.06
b
⫺.16
a,b
Burnout Work-to-family conflict .49
a
.49
a
.30
Family-to-work conflict .37
a
.36
a,b
.26
b
Balance ⫺.40
a
⫺.38
a,b
⫺.32
b
Job control ⫺.28
a,b
⫺.23
a
⫺.36
b
Work overload .42 .34
a
.26
a
Work role clarity ⫺.39
a
⫺.29
b
⫺.32
a,b
Family overload .40 .26
a
.27
a
Perceived depression Work-to-family conflict .44
a
.44
a
.25
Family-to-work conflict .37
a
.42
a
.29
Balance ⫺.44 ⫺.39
a
⫺.34
a
Job control ⫺.17
a,b
⫺.12
a
⫺.24
b
Work overload .30 .22
a
.16
a
Work role clarity ⫺.31
a
⫺.22
b
⫺.23
a,b
Family overload .43
a
.30
b
.34
a,b
Note. Entries in the same row with matching
a
and
b
superscripts are not significantly different from one
another using Steiger’s test for the difference between two dependent correlations, p⬍.05. All outcomes were
measured using multiple-item scales.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 FISHER, MATTHEWS, AND GIBBONS
inconsistent with the results obtained from multiple-item mea-
sures.
One explanation for these varying results is that the construct
definitions used to create the new single items were not necessarily
the same as the definitions used to create the existing multiple item
scales. For example, the new single items with the lowest factor
loadings when added to the multiple item scales all assessed
constructs in the nonwork domain (e.g., personal/family support,
personal/family control, and family overload). Perhaps these re-
sults are indicative of the complex nature of life outside of work
such that a single new item designed to globally capture that
criterion space does not correlate highly with multiple items in-
tended to assess that construct. The analyses presented above
compare the new single items with the existing scales; although
these analyses are informative, they only address one portion of the
puzzle. For example, the existing single-item measures generally
appear more internally consistent and often have higher conver-
gent validity than the new single measures. However, we have a
biased yardstick in that we are comparing them with a measure that
includes the existing item, resulting in an inflated relationship. The
existing single items we tested here might have performed quite
differently in comparison to a different multiple-item scale. Thus,
it is important to understand not only how the single items com-
pare to existing scales, but whether they are valid representations
of the constructs they purport to measure. Therefore, our next step
was to systematically assess content validity of these items.
Study 2: Content Validation
In addition to the construct validity evidence sought in Study 1,
we also gathered data to assess the content validity and research
utility of the new and existing single items from Study 1.
Subject Matter Experts and Rating Procedure
An independent sample of 17 subject matter experts (i.e., not the
same individuals who reviewed the items and constructs described
in Study 1) reviewed and rated a total of 37 single-item measures,
including the 18 items developed in Study 1 and 19 items selected
from multiple-item scales.
3
As a group, all N⫽17 experts earned
PhDs in relevant fields (psychology, management), collectively
have published 500 articles (an average of 29 articles per SME),
and approximately two thirds (65%) are members of editorial
boards for occupational health psychology, applied psychology or
management journals and the other third all serve as reviewers of
relevant journals. Each subject matter expert was e-mailed a link to
an online qualtrics survey and asked to make three assessments
about each item: (a) “Please rate how well the single item above
adequately represents the conceptual definition provided” using a
scale of 1 ⫽poor,2⫽fair,3⫽good,4⫽very good,5⫽
excellent; (b) “To what extent does this single item capture all
important aspects of the construct?” using a scale of 1 ⫽not at all
to5⫽to a great extent; and (c) to what degree would the item be
useful to researchers (i.e., “Please rate the utility of each item, on
its own, for research purposes. In other words, how useful would
this item be to researchers as a single-item measure of the intended
construct?” using a 5-point scale of 1 ⫽poor,2⫽fair,3⫽good,
4⫽very good,5⫽excellent). SMEs were blind regarding
whether items were newly developed or derived from existing
measures. The order in which the 37 items were presented to SMEs
was randomized by the qualtrics survey in order to avoid system-
atically presenting the new or existing items first.
Results
Results from the assessments provided by the SMEs are re-
ported in Table 1. Content validity ratings were consistent in terms
of whether items represented the conceptual definition and the
extent to which the items captured all aspects of the construct.
Among the new single items, the item with the lowest content
validity assessment was the burnout item (M⫽3.53, SD ⫽1.40 on
a 5-point scale), whereas the item evaluated to have the most
content validity was the life satisfaction item (M⫽4.71, SD ⫽
.47). All 18 new items were also evaluated as having good to very
good research utility. The item evaluated to have the most research
utility was the life satisfaction item (M⫽4.65, SD ⫽.70). The
items with the lowest research utility assessment relative to other
items were the work–family balance (M⫽3.07, SD ⫽1.22), and
the burnout items (M⫽3.07, SD ⫽1.39).
As we expected, SME ratings indicated that all new single-item
measures were rated as good or very good and content validity and
research utility ratings varied as a function of item origin. We
conducted paired sample ttests to test for statistical significance
between means of SME ratings for new versus existing items.
Results indicated that 15 items were rated as having significantly
higher content validity than the existing single item and 13 items
had higher research utility (see Table 1), providing support for
Hypotheses 5 and 6. SME ratings for work centrality, family–work
conflict and job satisfaction all indicated that the new items had
good or very good ratings, but were not significantly different from
ratings of existing items. Job satisfaction was the one item for
which the existing item ratings were slightly, but not significantly,
higher than the new item.
Brief Discussion
As part of this study, we wrote new single items to assess the
constructs included in this study because we sought to overcome
the criticism in the psychometric literature that single-item mea-
sures lack content validity and may be criterion deficient by
intentionally writing broader, more global items. We also obtained
data to evaluate the content validity of single items selected from
multiple item measures to test our hypothesis that the new items
would have higher levels of content validity. Our results provided
support for our prediction that new single item measures will
generally demonstrate evidence of content validity research and
research utility compared with single items chosen psychometri-
cally from existing measures. Both new and existing job satisfac-
tion items were rated highly and results did not suggest that the
new work centrality item was much better than an item selected
from an existing scale.
3
Two psychometrically chosen existing items were included to measure
work to family conflict as well as family to work conflict (i.e., one item
each for time-based and strain-based conflict). No existing single items
were used to assess personal/family life satisfaction because there was no
such existing validated measure at the time all the data here were collected.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
SINGLE-ITEM MEASURES
Study 3: Further Evidence of Relations With Other
Variables and Stability
In Study 1, we found that the internal consistency of existing
single items was consistently higher than that of new single items;
however, we noted that using the communality as an index of an
item’s internal consistency was somewhat biased in favor of the
existing items because those items are part of the original scale on
which the communality is based. The new items were not written
to fit the definitions of these particular existing scales, but rather to
fit broad definitions informed by the consensus of subject matter
experts. To the extent that the underlying definitions differ be-
tween the new single items and the multiitem scales, we might
expect that the new single item will show lower “internal consis-
tency” with a scale that is, after all, measuring something slightly
different. Accordingly, we were also interested in assessing test–
retest reliability for the new single items across two different time
lags. Because the examination of stability was conducted on a very
exploratory basis, and in an effort to minimize respondent burden
by keeping the survey very short, we only examined test–retest
reliability for the new single items.
Method
Participants and procedure. Using the same method as re-
ported in Study 1, an independent sample of N⫽1,294 individuals
participated in a basic assessment survey to obtain evidence about
the relations among the new single-item measures. Of these, 302
completed a second assessment at a later point in time to provide
evidence about the stability of the items over time. The primary
analysis sample (n⫽992) was used to estimate descriptive sta-
tistics and correlations among the new single-item measures. The
302 participants who completed the test–retest reliability assess-
ment are excluded from the other analyses reported in this section.
Demographic characteristics for both samples (n⫽992, n⫽302)
are reported in Table 2.
A total of 126 trained undergraduate student recruiters from
advanced psychology classes from two universities assisted with
the data collection process. Both universities were large public
institutions in the United States; one located in the South, and one
located on the West Coast. The survey took approximately 5
minutes to complete; participation was voluntary.
Stability assessment. Of the total number of 1,294 partici-
pants, 581 (44.9%) agreed to participate in a second (Time 2)
survey. They provided their e-mail address to the primary
investigator and were randomly assigned to one of two time
lags for the Time 2 survey: 1 versus 3 months. Due to concerns
over lower response rates for individuals in the 3-month lag
condition, a higher proportion (i.e., 60%; n⫽349) of respon-
dents were randomly assigned to the longer 3-month lag; 169
respondents (48.4%) completed the survey. The remaining 40%
of respondents were assigned to the 1-month lag condition; 133
respondents (56.9%) completed the survey. Nonresponse analyses
were conducted to examine if those who completed the Time 2
survey (n⫽302, regardless of time lag) differed from nonrespon-
dents (n⫽279, regardless of time lag). Respondents did not
significantly differ from nonrespondents in terms of age, number
of children, education level, organizational tenure, hours worked,
or household income level.
Participants were reminded of the nature of the study and
provided with a second Web-link. Reminder e-mails were sent 1
week after the follow-up e-mail. The Time 2 survey took approx-
imately 5 min to complete and participation was voluntary. In
return for participating respondents were entered into a drawing
for one of six $25 cash prizes.
Measures. Survey content in both the Time 1 and Time 2
surveys was the same. Participants were asked to respond to the 18
newly developed single-item measures presented in Table 1 based
on a 7-point Likert-type response scale (from 1 ⫽strongly dis-
agree to 7 ⫽strongly agree). Participants also provided basic
demographic information, including: age, gender, marital status,
number of children, household income, hours worked, tenure, and
occupation.
Results
Descriptive statistics and relations among variables. Using
the primary analysis sample (n⫽992), means and standard devi-
ations for the 18 new single items are reported in Table 7. Gen-
erally speaking, the 18 items assessed in this large sample dem-
onstrated means and standard deviations that might be expected
based on prior research. For example, respondents indicated both
work (M⫽4.22, SD ⫽1.75) and family (M⫽6.34, SD ⫽1.01)
are central to them, with a higher mean score for family centrality.
Respondents experienced moderate levels of work overload (M⫽
3.30, SD ⫽1.8) and family overload (M⫽3.84, SD ⫽1.74).
Additionally, reports of work-to-family (M⫽3.73, SD ⫽1.93)
and family-to-work conflict (M⫽3.09, SD ⫽1.73) are consistent
with past research, with respondents indicating higher occurrences
of work-to-family than family-to-work conflict. On a 7-point re-
sponse scale, all 18 items had a standard deviation of 1.0 or more,
indicating that variability in responses occurred within the sample.
Bivariate correlations that show relations among constructs using
the single-item measures are presented in Table 7.
Stability over time. Test–retest reliability estimates for each
of the single-item measures based on those respondents who
completed the Time 2 survey (1-month lag N⫽133; 3-month lag
N⫽169) are presented in Table 8. Test–retest reliabilities for the
1-month lag were relatively high. The work-to-family and family-
to-work conflict items had the lowest overall test–retest reliabili-
ties for the 1-month lag (.47, .46, respectively). The work central-
ity measure had the highest test–retest reliability (.78) for the
1-month lag. Overall, these constructs were stable over a 1-month
lag. Among participants in the 3-month lag condition, the family
to-work conflict measure had the lowest overall test-retest reli-
abilities (.44). The personal/family support measure had the high-
est test–retest reliability (.77) for the 3-month lag. These results
suggest these constructs are generally stable, even over a 3-month
lag. Overall test-retest reliabilities were lower for the 3-month lag
compared with the 1-month lag. However, the personal/family
support measure was actually more stable in the 3-month condition
compared with the 1-month condition.
Of particular interest was how the single-item measures would
function when examined on more than a single occasion. In Table
8 correlations between the 18 single-item measures for the Time1–
Time 2 samples (1-month lag N⫽132; 3-month lag N⫽162) are
reported. Because the within-lag correlations are similar to those
reported in Table 7, only cross-lagged correlations are reported (a
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
14 FISHER, MATTHEWS, AND GIBBONS
full report of correlations is available from the second author).
These results answer our research question regarding the test–
retest reliability of the single-item measures.
Discussion
Although there is a strong consensus that multiple-item mea-
sures are preferred over single-item measures, we suggest that
situations exist wherein it may not be feasible or practical to
include longer measures in surveys. Using single-item measures
may be an effective way to reduce the length of a survey, which
may increase response rates, increase the feasibility of survey
research in longitudinal research designs, and reduce survey de-
velopment and administration costs. Preserving survey response
rates is important to minimize nonresponse bias (Rogelberg &
Stanton, 2007) as well as minimize attrition in longitudinal re-
search.
The aim of this study was to develop and evaluate two types of
single-item measures for commonly studied constructs in organi-
zational research: a set of new items written as single-item indi-
cators, and a set of single items selected on the basis of psycho-
metric properties from existing measures. We examined 37 items
for measuring 18 constructs to assess the work–family interface as
well as common predictors (domain centrality, control within the
domain, role overload, work–family conflict, work–family bal-
ance, social support, work role clarity) and important outcomes
(job, family, and life satisfaction, burnout, and depression) of
work–family issues. We gathered data and tested hypotheses to
evaluate these items.
In general our results provided support for the reliability, con-
tent, and construct validity of many of the items. However, some
items performed differently across the various criteria we evalu-
ated. Specifically, single items selected from multiple-item mea-
sures based on the highest factor loading showed higher commu-
nalities and better convergence with the multiple-item scales
compared with the new single items, though this result is not
surprising given that the items were selected from those scales.
Our results also showed evidence of convergent and discriminant
validity regardless of origin. Relations between single-item mea-
sures and other constructs were generally similar upon comparison
of new and psychometrically chosen single items. The primary
exception regarding convergent validity evidence was for work–
family conflict, primarily because the existing item that shows the
highest correlation is strain-based conflict. However, SMEs indi-
cated lower levels of content validity and research utility for the
strain-based conflict items and wrote specific comments about
potential problems of using source-based work–family conflict
items (e.g., time- and strain-based items) to assess work–family
and family–work conflict using single-item measures. In most
cases, new items were rated by subject matter experts as having
higher content validity and research utility. Lastly, as suggested by
Nagy (2002) and Gosling et al. (2003), we examined test–retest
reliability estimates across both 1-and 3-month lags and found
stability across shorter lags for some new single items (work
centrality, family centrality, supervisor support, job control, work–
family balance, job satisfaction, and perceived depression) and
lower levels of stability across both lags for other items (personal/
family control, work overload, family overload, both directions of
work–family conflict, life satisfaction, and burnout). The test–
retest correlations were higher for enduring traits (e.g., domain
centrality) or most stable states (e.g., job satisfaction) and lower
for more transient conditions (e.g., work–family conflict, over-
load).
Recommendations
Based on results across three studies to evaluate both new single
item measures as well as existing, psychometrically chosen single
items, we offer the following recommendations. First, if a re-
searcher’s goal is to obtain results that are consistent with or
directly comparable with a specific established measure, then
based on results regarding reliability, convergent, and discriminant
validity in Study 1, we recommend choosing an item from that
(existing) measure. However, if one aims to assess constructs
Table 7
Study 3 Bivariate Correlations Among the Single-Item Measures
MSD 1234567891011121314151617
1. Work centrality 4.22 1.75
2. Family centrality 6.34 1.01 ⫺.07
3. Coworker support 5.33 1.48 .17 .10
4. Supervisor support 5.29 1.59 .17 .07 .61
5. Personal/family support 5.95 1.17 ⫺.04 .31 .24 .18
6. Work role clarity 5.89 1.31 .15 .05 .32 .38 .14
7. Job control 5.56 1.46 .23 .08 .36 .44 .15 .40
8. Personal/family control 4.97 1.44 ⫺.02 .23 .17 .15 .42 .16 .20
9. Work overload 3.30 1.85 .08 .02 ⫺.17 ⫺.16 ⫺.10 ⫺.21 ⫺.19 ⫺.14
10. Family overload 3.84 1.74 .12 ⫺.11 ⫺.13 ⫺.14 ⫺.29 ⫺.15 ⫺.16 ⫺.40 .42
11. Work-to-family conflict 3.73 1.93 .13 ⫺.05 ⫺.16 ⫺.20 ⫺.07 ⫺.14 ⫺.16 ⫺.09 .34 .35
12. Family-to-work conflict 3.09 1.73 .09 ⫺.01 ⫺.03 ⫺.06 ⫺.21 ⫺.12 ⫺.06 ⫺.17 .26 .29 .30
13. Work–family balance 5.10 1.51 .01 .20 .29 .32 .26 .22 .28 .28 ⫺.24 ⫺.41 ⫺.45 ⫺.09
14. Life satisfaction 5.68 1.20 .04 .26 .27 .27 .34 .23 .27 .34 ⫺.14 ⫺.28 ⫺.09 ⫺.07 .46
15. Personal/family satisfaction 5.68 1.30 ⫺.04 .38 .18 .19 .40 .17 .17 .43 ⫺.08 ⫺.29 ⫺.06 ⫺.09 .40 .68
16. Job satisfaction 5.14 1.48 .33 .11 .35 .44 .16 .32 .40 .16 ⫺.14 ⫺.15 ⫺.16 ⫺.06 .46 .50 .27
17. Burnout 2.55 1.00 ⫺.08 ⫺.08 ⫺.24 ⫺.29 ⫺.21 ⫺.21 ⫺.29 ⫺.26 .24 .33 .29 .20 ⫺.43 ⫺.36 ⫺.24 ⫺.42
18. Perceived depression 2.15 1.03 ⫺.04 ⫺.12 ⫺.22 ⫺.25 ⫺.25 ⫺.19 ⫺.27 ⫺.34 .16 .30 .15 .19 ⫺.30 ⫺.43 ⫺.36 ⫺.32 .52
Note. N ⫽963–992; rⱖ|.07| significant at p⬍.05; rⱖ|.10| significant at p⬍.01.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
15
SINGLE-ITEM MEASURES
Table 8
Study 3 Test–Retest Reliabilities and Cross-Lagged Correlations at 1- and 3-Month Time Lags
Construct assessed
at Time 1
Construct assessed at Time 2 (1-month lag/3-month lag)
123456789101112131415161718
1. Work centrality .78/.59 .04/⫺.11 .11/.06 .25/.06 .07/⫺.04 .16/.01 .34/.17 .05/.00 .04/.13 .07/.04 .12/.04 .06/.02 .05/.23 .17/.07 .06/.00 .38/.31 .01/⫺.01 ⫺.05/.02
2. Family
centrality .02/⫺.09 .71/.57 .21/.08 .07/.02 .32/.28 .12/⫺.03 .06/⫺.08 .14/.23 ⫺.05/⫺.02 .03/⫺.09 .10/⫺.01 ⫺.01/⫺.07 .08/.06 .21/.21 .28/.24 .17/.19 .07/⫺.02 ⫺.04/⫺.09
3. Coworker
support .38/.16 .32/.14 .61/.46 .51/.35 .21/.18 .33/.16 .44/.24 .07/.09 ⫺.23/⫺.02 .01/⫺.02 .00/⫺.18 .02/⫺.10 .27/.27 .31/.31 .27/.17 .47/.35 ⫺.21/⫺.22 ⫺.15/⫺.21
4. Supervisor
support .36/.13 .17/⫺.02 .36/.43 .69/.61 .24/.12 .33/.23 .49/.45 .08/.21 ⫺.32/⫺.14 .07/⫺.16 ⫺.16/⫺.19 .00/.01 .32/.37 .32/.34 .14/.23 .47/.43 ⫺.20/⫺.35 ⫺.06/⫺.28
5. Personal/family
support .03/⫺.14 .33/.38 .23/.08 .14/.14 .56/.77 .09/.04 .06/.00 .29/.41 ⫺.08/⫺.19 ⫺.18/⫺.40 .01/⫺.12 ⫺.22/⫺.27 .09/.24 .20/.36 .24/.42 .05/.08 ⫺.11/⫺.26 ⫺.16/⫺.40
6. Work role
clarity .30/.07 .11/.12 .17/.14 .31/.09 .09/.17 .54/.35 .36/.26 .10/.17 ⫺.35/⫺.33 ⫺.01/⫺.14 ⫺.11/⫺.04 ⫺.14/⫺.09 .27/.13 .18/.19 .08/.04 .52/.21 ⫺.16/⫺.16 ⫺.04/⫺.23
7. Job control .47/.19 .26/⫺.10 .32/.36 .48/.43 .31/.07 .47/.33 .72/.58 .19/.25 ⫺.30/⫺.18 .03/⫺.24 ⫺.21/⫺.14 ⫺.12/⫺.01 .33/.38 .44/.30 .30/.22 .54/.47 ⫺.24/⫺.44 ⫺.25/⫺.24
8. Personal/family
control .12/.08 .12/.22 .17/.12 .14/.21 .18/.33 .27/.12 .26/.09 .54/.57 ⫺.27/⫺.14 ⫺.38/⫺.35 ⫺.16/⫺.15 ⫺.26/⫺.13 .36/.33 .39/.31 .43/.37 .25/.23 ⫺.38/⫺.18 ⫺.34/⫺.08
9. Work overload ⫺.07/.11 .06/.07 ⫺.05/⫺.10 ⫺.22/⫺.16 ⫺.03/⫺.12 ⫺.32/⫺.20 ⫺.29/⫺.18 ⫺.15/⫺.17 .55/.57 .16/.36 .47/.27 .24/.09 ⫺.40/⫺.29 ⫺.06/⫺.27 ⫺.03/⫺.19 ⫺.21/⫺.14 .40/.20 .15/.11
10. Family
overload ⫺.10/⫺.03 ⫺.04/⫺.10 ⫺.20/⫺.15 ⫺.12/⫺.21 ⫺.25/⫺.31 ⫺.21/⫺.13 ⫺.20/⫺.13 ⫺.36/⫺.50 .40/.46 .53/.58 .39/.25 .30/.21 ⫺.41/⫺.40 ⫺.31/⫺.35 ⫺.28/⫺.40 ⫺.13/⫺.25 .44/.23 .30/.23
11. Work-to-family
conflict ⫺.07/⫺.01 .02/.05 ⫺.17/⫺.16 ⫺.19/⫺.10 .00/⫺.10 ⫺.33/⫺.09 ⫺.30/⫺.12 ⫺.14/⫺.18 .35/.41 .09/.40 .47/.52 .02/.20 ⫺.43/⫺.36 ⫺.16/⫺.22 ⫺.02/⫺.11 ⫺.27/⫺.23 .22/.31 .03/.17
12. Family-to-work
conflict .01/⫺.02 ⫺.06/⫺.07 ⫺.12/.05 ⫺.07/.00 ⫺.26/⫺.17 ⫺.14/⫺.08 .00/⫺.08 ⫺.22/⫺.20 .10/.20 .23/.31 ⫺.01/.09 .46/.44 .01/⫺.11 ⫺.07/⫺.15 ⫺.11/⫺.21 ⫺.17/⫺.03 .16/.06 .20/.13
13. Work–family
balance .19/.14 .03/.00 .25/.14 .27/.18 .10/.21 .38/.06 .36/.23 .27/.27 ⫺.45/⫺.32 ⫺.16/⫺.37 ⫺.44/⫺.37 .01/⫺.05 .69/.67 .34/.52 .23/.42 .40/.40 ⫺.48/⫺.35 ⫺.17/⫺.25
14. Life
satisfaction .09/.10 .18/.26 .17/.10 .28/.06 .25/.32 .15/.16 .23/.22 .38/.33 ⫺.05/⫺.06 ⫺.24/⫺.22 ⫺.19/⫺.19 ⫺.23/⫺.17 .35/.46 .59/.56 .44/.42 .34/.45 ⫺.29/⫺.30 ⫺.38/⫺.38
15. Personal/family
satisfaction ⫺.02/⫺.04 .29/.33 .19/.17 .11/.10 .38/.33 .08/.13 .13/.08 .43/.40 ⫺.07/⫺.09 ⫺.30/⫺.20 ⫺.17/⫺.10 ⫺.35/⫺.20 .25/.43 .48/.57 .49/.64 .18/.33 ⫺.19/⫺.20 ⫺.36/⫺.23
16. Job satisfaction .46/.39 .10/.01 .39/.21 .49/.35 .13/.12 .38/.17 .49/.34 .16/.19 ⫺.16/⫺.11 .04/⫺.19 ⫺.14/⫺.16 ⫺.02/⫺.04 .33/.48 .53/.47 .24/.31 .70/.60 ⫺.34/⫺.39 ⫺.32/⫺.30
17. Burnout ⫺.18/⫺.20 ⫺.14/⫺.03 ⫺.30/⫺.23 ⫺.31/⫺.31 ⫺.21/⫺.23 ⫺.29/⫺.21 ⫺.28/⫺.31 ⫺.31/⫺.31 .29/.27 .26/.38 .23/.34 .25/.23 ⫺.35/⫺.46 ⫺.38/⫺.35 ⫺.29/⫺.26 ⫺.39/⫺.37 .64/.54 .42/.38
18. Perceived
depression ⫺.07/⫺.11 ⫺.18/⫺.18 ⫺.27/⫺.06 ⫺.20/⫺.10 ⫺.30/⫺.30 ⫺.19/⫺.06 ⫺.25/⫺.20 ⫺.34/⫺.24 .06/.09 .26/.24 .15/.19 .24/.14 ⫺.23/⫺.20 ⫺.44/⫺.32 ⫺.42/⫺.26 ⫺.28/⫺.22 .48/.35 .72/.63
Note. For the 1-month lag, N⫽133, rⱖ|.18| significant at p⬍.05, rⱖ|.23| significant at p⬍.01; for the 3-month lag, N⫽169, rⱖ|.15| significant at p⬍.05, rⱖ|.20| significant at p⬍.01.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
16 FISHER, MATTHEWS, AND GIBBONS
using single-item measures but with a primary focus on content
validity, then we generally recommend the new items developed in
this study (see Table 1).
If one’s goal is to identify single-item measures that may be
acceptable taking all of the results we evaluated into account, see
Table 9. We synthesized the data presented thus far and provide
concrete recommendations to researchers based on results across
these three studies. Specifically, we classified 11 items as accept-
able because reliability, convergent validity, content validity, re-
search utility ratings, and test-retest results were consistently fa-
vorable; 14 items as use with caution for items that yielded mixed
results, and the remaining 12 items as not recommended because
they lacked favorable evidence to endorse use or possible use of
the item. Our classifications are based on our review across all
forms of empirical evidence we evaluated in this study (internal
consistency reliability, factor loadings in relation to multiple-item
measures, correlations to examine evidence of convergent and
discriminant validity, SME ratings of content validity and research
utility, and test–retest reliability across 1- and 3-month lags).
However, it is important to note that some forms of evidence were
missing for some of the items (e.g., test–retest reliability for
existing single items), and internal consistency reliability was
likely to be higher for items derived from the multiple item scales
used for comparison. We encourage researchers to review the
empirical results presented in this paper and draw their own
conclusions by weighting the evidence that is most important for
the purpose of their research study. If a researcher is quite con-
cerned about reliability, then we recommend a more conservative
approach and suggest avoiding the use of single-item measures
altogether.
Contributions
The present study is the first to systematically develop and
assess multiple single item measures applicable to organizational
and occupational health research. Primary criticisms of single-item
measures include a lack of content validity due to criterion defi-
ciency, and unreliability (Cronbach & Meehl, 1955; Nunnally &
Bernstein, 1978; Schriesheim et al., 1991). Our study empirically
examined data across three studies to evaluate evidence of content
validity and reliability, and our results indicated that many, though
not all, single-item measures may be useful when it is not feasible
nor practical to use multiple item measures. We provided specific
recommendations in Table 9 regarding the items we investigated,
but also noted that which items we recommend depend on re-
searchers’ objectives and criteria that are most important for their
research.
Although Stanton et al. (2002) developed and validated methods
for reducing the length of psychological measures, we extended
this work by shortening measures to the extreme (i.e., using single
items) and evaluating the reliability and validity of single items.
Prior organizational research has evaluated single-item measures
of job satisfaction (Wanous & Hudy, 2001; Wanous et al., 1997).
Our study makes an important contribution to the literature by
developing and evaluating single-item measures of many addi-
tional constructs frequently used in organizational and occupa-
tional health psychology research.
The results of this study may facilitate future research, particu-
larly when researchers or practitioners face very similar and rigid
survey development guidelines regarding survey length. Although
we agree that multiple item measures are preferred, there may be
valid situations when there is only room for one item on a survey
to assess a particular construct. Further, including single-item
measures as complements to existing multiple-item measures may
help address concerns over common method variance that are
pervasive in the literature. In other words, single items could be
used alongside multiple-item measures to provide a multiindicator
perspective, perhaps reducing concerns that study results are a
function of the particular measure used, and not the actual latent
construct of interest (Huffman, Youngcourt, Payne, & Castro,
2008).
Having a priori evidence about the quality of a single-item
measure allows researchers to make informed decisions about the
likely costs and benefits of using the item in their research designs.
For example, the National Health Interview Study (NHIS), a U.S.
national population-based health survey conducted by the National
Center for Health Statistics, included an occupational health mod-
ule in 2010 (Alterman, Luckhaupt, Dahlhamer, Ward, & Calvert,
2013) and a second in 2014. Numerous occupational health re-
searchers provided input regarding the design of these modules,
and when faced with strict and severe limitations regarding survey
length, these researchers decided to pursue breadth of measure-
ment across multiple occupational health constructs rather than
depth of measurement among a few constructs. The end result was
the use of a series of single-item occupational health measures, and
not necessarily items that had been previously validated. This
example illustrates our argument that researchers can and do
consider such tradeoffs in their research designs.
Limitations
One limitation of the current study was the selection of multiple-
item measures. First, we compared each single-item measure with
only one multiple item measure of the same construct. Therefore,
the results we obtained upon comparing the single- and multiple-
item measures may have varied if we had selected a different
multiple-item measure. Second, some of the multiple-item mea-
sures we used in this study may not be the best representation for
assessing the construct of interest. As such, our evaluation of the
single-item measure is dependent on the quality of the multiple-
item measure. This study was an important first step in assessing
the validity of these single-item measures. Additional research
should be conducted to further validate the items using alternative
measures as comparison.
A second limitation worth noting is that not all of the multiple
item scales used for comparison were assessed using the same
response scale. For example, some used response options with a
Likert scale ranging from strongly agree to strongly disagree
whereas others used a frequency-based response scale and the
number of scale points varied. These differences across scales
were based on our efforts to use scales in a manner consistent with
prior research. However, future research should incorporate mea-
sures in a more consistent manner across scales.
Lastly, the use of self-report survey data may lead to artificially
inflated correlations between measures of psychological con-
structs. However, the constructs in the present study necessitated a
self-report methodology in which individuals reported their own
perceptions (Schmitt, 1994). Further, we note that common
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
17
SINGLE-ITEM MEASURES
Table 9
Recommendations for use of Single-Item Measures
Construct Item Acceptable
Use with
caution
Not
recommended
Work centrality 1. My work is one of the most important things in my life right now.
a
X
2. To me, my job is a very large part of who I am.
b
X
Family centrality 3. My family is one of the most important things in my life right
now.
a
X
4. I am very much personally involved with my family.
b
X
Coworker support 5. I can count on my coworkers/work colleagues for support when I
need it.
a
X
6. (To what extent can you...)Count on your coworkers to back you
up at work?
b
X
Supervisor support 7. I can count on my supervisor/manager for support when I need it.
a
X
8. My supervisor tries to make my job as interesting as possible.
b
X
Personal/family support 9. I can count on my friends/family members for support when I need
it.
a
X
10. Please rate the extent to which....youfeel that [your family
members] were really trying to understand your problems?
b
X
Work role clarity 11. I have a clear understanding of what is expected of me in my job.
a
X
12.(Inmyjob...)Iknow what my responsibilities are.
b
X
Job Control 13. I feel I have a lot of control over how I do my job.
a
X
14. (To what extent do you...)Plan your own work?
b
X
Personal/family control 15. I feel I have a lot of control over things in my personal/family life.
a
X
16. (To what extent do you...)Plan your own [personal life]
activities?
b
X
Work role overload 17. I often feel I am unable to meet all the demands related to my
work.
a
X
18. I cannot ever seem to catch up (at work).
b
X
Family role overload 19. I often feel I am unable to meet all the demands in my personal/
family life.
a
X
20. I need more hours in the day to do all the things that are expected
of me (at home).
b
X
Work/family conflict 21. In the past month my WORK life frequently interfered with my
PERSONAL/FAMILY life.
a
X
22. I have to miss family activities due to the amount of time I must
spend on work responsibilities.
b
X
23. I am often so emotionally drained when I get home from work that
it prevents me from contributing to my family.
b
X
Family/work conflict 24. In the past month my PERSONAL/FAMILY life frequently
interfered with my WORK life.
b
X
25. I have to miss work activities due to the amount of time I must
spend on family responsibilities.
a
X
26. Because I am often stressed from family responsibilities, I have a
hard time concentrating on my work.
b
X
Work/family balance 27. In general I feel that I have an adequate balance between my work
and personal/family life.
a
X
28. It is clear to me, based on feedback from coworkers and family
members, that I am accomplishing both my work and family
responsibilities.
b
X
Life satisfaction 29. As a whole, I am satisfied with my life.
a
X
30. In most ways my life is close to ideal.
b
X
Personal/family life satisfaction 31. Overall, I am satisfied with my personal/family life.
a
X
Job satisfaction 32. Overall, I am satisfied with my job.
a
X
33. All in all I am satisfied with my job.
b
X
Burnout 34. Feelings of burnout occur when an individual feels emotional
exhausted, cynical, and feels a lack of personal accomplishment. In
the past month, how often have you experienced feelings of
burnout?
a
X
35. I feel burned out.
b
X
Perceived depression 36. Depression is considered to exist when an individual feels sad, has
trouble sleeping, lacks motivation, feels worthless, is withdrawn,
and is generally fatigued. In the past month, how often have you
felt depressed?
a
X
37. Much of the time during the past week you felt depressed.
b
X
Note. Please see Measures section in Study 1 regarding the source of each item.
a
Newly written single item as part of this study.
b
Single item from existing scale.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
18 FISHER, MATTHEWS, AND GIBBONS
method variance should only be considered a serious issue if there
appears to be a systematic and pervasive inflation of observed
relations (James, Gent, Hater, & Corey, 1979). In our study, the
range of individual-level correlations, absence of multicollinearity
(Kline, 1998), and absence of nonintuitive relations leads us to
believe that common method variance is not a significant concern.
Future Research
The new single items we developed were written by experienced
occupational health researchers, based on a careful and thoughtful
analysis of the literature, and incorporated feedback from knowl-
edgeable independent experts. However, in any scale development
process, even experts may produce items that perform in unex-
pected ways, and it is possible that revising or rewording some of
these items could yield stronger results in future. Future studies
examining alternative single-item measures, comparison with
other multiple item scales, and additional constructs can shed light
on this issue. Additional research is needed to compare the new
and existing items examined in this study with other existing items
and measures, as well as over time. The current study only exam-
ined test–retest reliability among the newly developed items. Test–
retest reliability of the existing single-item measures should also
be assessed.
Consistent with suggestions by Fahrenberg (2006) and Klumb,
Elfering, and Herre (2009), we recommend frequent replication of
the single-item measures included in this study. Although single-
item measures may be useful by offering opportunities to assess a
broader range of topics and measure control variables to assess
multivariate organizational and occupational health issues, they
also have specific limitations. For example, single-item measures
may be prone to reduced variability within and between individ-
uals, skewed distributions, response bias, and context effects
(Fahrenberg, 2006). Furthermore, no measure of any length can be
declared categorically “valid” on the basis of a single study, or
even a small collection of studies as we have presented here.
Rather, the modern view of validity argues that we should avoid
describing a measure as “valid” at all; rather, we should consider
the body of evidence that supports the appropriateness of making
particular inferences about the scores we obtain from that measure
(e.g., Messick, 1995). Such a body of evidence should grow and
develop over time as multiple researchers contribute multiple
studies using the measure. Thus, future research should replicate
the use of single-item measures recommended here and examine
how the current items perform in other contexts (e.g., in relation to
other variables, among other populations) to fully understand what
inferences we can make from these single items. When researchers
create single-item measures ad hoc for each new study, the devel-
opment of cumulative knowledge about the properties of those
items is impossible. If researchers who use single items can at least
be persuaded to use the same items, it becomes possible to situate
those items within a nomological net and to understand their
typical functioning. We hope that by providing single-item mea-
sures about which at least some psychometric information is
available, we can encourage and facilitate such research.
Finally, we suggested that one reason to consider the use of
single-item measures is to increase response rates and reduce
attrition in occupational health psychology research. Although
survey methodologists have demonstrated that the length of survey
instruments is an important factor in relation to response rates and
attrition (Crawford et al., 2001; Groves, 2004), future research
should be conducted to further examine this issue and to better
understand research participants’ perceptions of single-item mea-
sures.
Conclusions
Our aim was to provide organizational and occupational health
researchers and practitioners with single items to assess psycho-
logical constructs as well as an empirical evaluation of those items
to inform decisions regarding the reliability and validity of some
possible single-item measures. We do not advocate that single-
item measures are a panacea for use in research, but rather an
alternative measurement approach when survey length and/or re-
spondent burden concerns preclude the use of multiple item mea-
sures. Furthermore, single items are more appropriate as moderator
or control variables, rather than as a primary construct in a research
study (Fuchs & Diamantapoulos, 2009). In short, we propose that
some data to assess constructs of interest may be better than no
data, and this can be accomplished more effectively for some
constructs than others. The use of single-item measures in survey
research is one strategy for reducing the length of a survey, which
may result in more desirable response rates and/or lower attrition
in longitudinal studies. The data we presented from multiple
sources and at multiple time points illustrate that single-item
measures can be used effectively to assess many relevant con-
structs, and we offered specific recommendations about the items
we evaluated. The use of single-item measures, if done systemat-
ically using concrete and validated items, should not be viewed as
a fatal flaw within a study; whether single- or multiple-item
measures are used, the study as a whole must be examined in terms
of the contribution it makes in advancing the field.
References
Alterman, T., Luckhaupt, S. E., Dahlhamer, J. M., Ward, B. W., & Calvert,
G. M. (2013). Job insecurity, work–family imbalance, and hostile work
environment: Prevalence data from the 2010 National Health Interview
Survey. American Journal of Industrial Medicine, 56, 660 – 669. http://
dx.doi.org/10.1002/ajim.22123
Atkinson, M. J., & Lennox, R. D. (2006). Extending basic principles of
measurement models to the design and validation of Patient Reported
Outcomes. Health and Quality of Life Outcomes, 4, 65–77. http://dx.doi
.org/10.1186/1477-7525-4-65
Bellavia, G. M., & Frone, M. R. (2005). Work–family conflict. In J.
Barling, E. K. Kelloway, & M. R. Frone (Eds.), Handbook of work stress
(pp. 113–147). Thousand Oaks, CA: Sage. http://dx.doi.org/10.4135/
9781412975995.n6
Blake, R. L., Jr., & McKay, D. A. (1986). A single-item measure of social
supports as a predictor of morbidity. The Journal of Family Practice, 22,
82– 84.
Brown, T. A. (2006). Confirmatory factor analysis for applied research.
New York, NY: Guilford Press.
Burisch, M. (1984). Approaches to personality inventory construction: A
comparison of merits. American Psychologist, 39, 214 –227. http://dx
.doi.org/10.1037/0003-066X.39.3.214
Byron, K. (2005). A meta-analytic review of work–family conflict and its
antecedents. Journal of Vocational Behavior, 67, 169 –198. http://dx.doi
.org/10.1016/j.jvb.2004.08.009
Cammann, C., Fichman, M., Jenkins, G. D., & Klesh, J. (1983). Michigan
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
19
SINGLE-ITEM MEASURES
Organizational Assessment Questionnaire. In S. E. Seashore, E. E.
Lawler, P. H. Mirvis, & C. Cammann (Eds.), Assessing organizational
change: A guide to methods, measures, and practices (pp. 71–138). New
York, NY: Wiley-Interscience.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant
validation by the multitrait-multimethod matrix. Psychological Bulletin,
56, 81–105. http://dx.doi.org/10.1037/h0046016
Carlson, D. S., Grzywacz, J. G., & Zivnuska, S. (2009). Is work–family
balance more than conflict and enrichment? Human Relations, 62,
1459 –1486. http://dx.doi.org/10.1177/0018726709336500
Casper, W. J., Eby, L. T., Bordeaux, C., Lockwood, A., & Lambert, D.
(2007). A review of research methods in IO/OB work–family research.
Journal of Applied Psychology, 92, 28 – 43. http://dx.doi.org/10.1037/
0021-9010.92.1.28
Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in
objective scale development. Psychological Assessment, 7, 309 –319.
http://dx.doi.org/10.1037/1040-3590.7.3.309
Cortina, J. M. (1993). What is coefficient alpha? An examination of theory
and applications. Journal of Applied Psychology, 78, 98 –104. http://dx
.doi.org/10.1037/0021-9010.78.1.98
Crawford, S. D., Couper, M. P., & Lamias, M. J. (2001). Web surveys:
Perception of burden. Social Science Computer Review, 19, 146 –162.
http://dx.doi.org/10.1177/089443930101900202
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological
tests. Psychological Bulletin, 52, 281–302. http://dx.doi.org/10.1037/
h0040957
DeSalvo, K. B., Bloser, N., Reynolds, K., He, J., & Muntner, P. (2006).
Mortality prediction with a single general self-rated health question. A
meta-analysis. Journal of General Internal Medicine, 21, 267–275.
http://dx.doi.org/10.1111/j.1525-1497.2005.00291.x
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The
satisfaction with life scale. Journal of Personality Assessment, 49, 71–
75. http://dx.doi.org/10.1207/s15327752jpa4901_13
Eby, L. T., Casper, W. J., Lockwood, A., Bordeaux, C., & Brinley, A.
(2005). Work and family research in IOOB: Content analysis and review
of the literature (1980 –2002). Journal of Vocational Behavior, 66,
124 –197. http://dx.doi.org/10.1016/j.jvb.2003.11.003
Eisenberger, R., Stinglhamber, F., Vandenberghe, C., Sucharski, I. L., &
Rhoades, L. (2002). Perceived supervisor support: Contributions to
perceived organizational support and employee retention. Journal of
Applied Psychology, 87, 565–573. http://dx.doi.org/10.1037/0021-9010
.87.3.565
Elo, A. L., Leppänen, A., & Jahkola, A. (2003). Validity of a single-item
measure of stress symptoms. Scandinavian Journal of Work, Environ-
ment & Health, 29, 444 – 451. http://dx.doi.org/10.5271/sjweh.752
Fahrenberg, J. (2006). Self-reported subjective state—Single-item or scales
like AD-ACL and PANAS? Retrieved from http://www.jochen-fahrenberg
.de/uploads/media/Self_report_of_Subjective_State_01.pdf
Fleeson, W. (2001). Toward a structure- and process-integrated view of
personality: Traits as density distribution of states. Journal of Person-
ality and Social Psychology, 80, 1011–1027. http://dx.doi.org/10.1037/
0022-3514.80.6.1011
Ford, M. T., Matthews, R. A., Wooldridge, J. D., Mishra, V., Kakar, U. M.,
& Strahan, S. R. (2014). How do occupational stressor-strain effects
vary with time? A review and meta-analysis of the relevance of time lags
in longitudinal studies. Work and Stress, 28, 9 –30. http://dx.doi.org/
10.1080/02678373.2013.877096
Fu, Y. C. (2005). Measuring personal networks with daily contacts: A
single-item survey question and the contact diary. Social Networks, 27,
169 –186. http://dx.doi.org/10.1016/j.socnet.2005.01.008
Fuchs, C., & Diamantopoulos, A. (2009). Using single item measures for
construct measurement in management research. Die Betriebswirtschaft,
69, 195–210.
Fuller, J. A., Stanton, J. M., Fisher, G. G., Spitzmuller, C., Russell, S. S.,
& Smith, P. C. (2003). A lengthy look at the daily grind: Time series
analysis of events, mood, stress, and satisfaction. Journal of Applied
Psychology, 88, 1019 –1033. http://dx.doi.org/10.1037/0021-9010.88.6
.1019
Gardner, D. G., Cummings, L. L., Dunham, R. B., & Pierce, J. L. (1998).
Single-item versus multiple-item measurement scales: An empirical
comparison. Educational and Psychological Measurement, 58, 898 –
915. http://dx.doi.org/10.1177/0013164498058006003
Gonzalez-Morales, M. G., Tetrick, L. E., & Ginter, R. (2013). Measure-
ment issues in work–family research. In R. R. Sinclair, M. Wang, &
L. E. Tetrick (Eds.), Research methods in occupational health psychol-
ogy: Measurement, design, and data analysis (pp. 31– 48). New York,
NY: Routledge.
Gosling, S. D., Rentfrow, P. J., & Swann, W. B., Jr. (2003). A very brief
measure of the Big-Five personality domains. Journal of Research in
Personality, 37, 504 –528. http://dx.doi.org/10.1016/S0092-
6566(03)00046-1
Grandey, A. A., Cordeiro, B. L., & Michael, J. H. (2007). Work–family
supportiveness organizational perceptions: Important for the well-being
of male blue-collar hourly workers? Journal of Vocational Behavior, 71,
460 – 478. http://dx.doi.org/10.1016/j.jvb.2007.08.001
Greenhaus, J. H., & Allen, T. D. (2011). Work–family balance: A review
and extension of the literature. In J. C. Quick & L. E. Tetrick (Eds.),
Handbook of occupational health psychology (2nd ed., pp. 165–183).
Washington, DC: American Psychological Association.
Greenhaus, J. H., & Beutell, N. J. (1985). Sources of conflict between work
and family roles. The Academy of Management Review, 10, 76 – 88.
Groves, R. M. (2004). Survey errors and survey costs. Hoboken, NJ:
Wiley.
Haynes, C. E., Wall, T. D., Bolden, R. I., Stride, C., & Rick, J. E. (1999).
Measures of perceived work characteristics for health services research:
Test of a measurement model and normative data. British Journal of
Health Psychology, 4, 257–275. http://dx.doi.org/10.1348/
135910799168614
Hinkin, T. R. (1995). A review of scale development practices in the study
of organizations. Journal of Management, 21, 967–988. http://dx.doi
.org/10.1177/014920639502100509
Huffman, A. H., Youngcourt, S. S., Payne, S. C., & Castro, C. A. (2008).
The importance of construct breadth when examining inter-role conflict.
Educational and Psychological Measurement, 68, 515–530. http://dx.doi
.org/10.1177/0013164407308472
Idler, E. L., & Benyamini, Y. (1997). Self-rated health and mortality: A
review of twenty-seven community studies. Journal of Health and
Social Behavior, 38, 21–37. http://dx.doi.org/10.2307/2955359
James, L., Gent, M., Hater, J., & Corey, K. (1979). Correlates of psychol-
ogy influence: An illustration of the psychological climate approach to
work environment. Personnel Psychology, 32, 563–588. http://dx.doi
.org/10.1111/j.1744-6570.1979.tb02154.x
Kenny, D. A., & McCoach, D. B. (2003). Effect of the number of variables on
measures of fit in structural equation modeling. Structural Equation Modeling,
10, 333–351. http://dx.doi.org/10.1207/S15328007SEM1003_1
Kline, R. B. (1998). Principles and practices of structural equation mod-
eling. New York, NY: Guilford Press.
Klumb, P., Elfering, A., & Herre, C. (2009). Ambulatory assessment in I/O
Psychology: Fruitful examples and methodological issues. European
Psychologist, 14, 120 –131. http://dx.doi.org/10.1027/1016-9040.14.2
.120
Kossek, E. E., Baltes, B. B., & Matthews, R. A. (2011). How work–family
research can finally have an impact in organizations. Industrial and
Organizational Psychology: Perspectives on Science and Practice, 4,
352–369. http://dx.doi.org/10.1111/j.1754-9434.2011.01353.x
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
20 FISHER, MATTHEWS, AND GIBBONS
Loo, R. (2002). A caveat on using single-item versus multiple-item scales.
Journal of Managerial Psychology, 17, 68 –75. http://dx.doi.org/
10.1108/02683940210415933
MacDermid, S. M. (2005). (Re)considering conflict between work and
family. In E. E. Kossek & S. Lambert (Eds.), Work and life integration
in organizations: Organizational, cultural, and individual perspectives
(pp. 19 –40). Mahwah, NJ: Erlbaum.
Matthews, R. A., Bulger, C. A., & Barnes-Farrell, J. L. (2010). Work social
supports, role stressors, and work–family conflict: The moderating effect
of age. Journal of Vocational Behavior, 76, 78 –90. http://dx.doi.org/10
.1016/j.jvb.2009.06.011
Matthews, R. A., Del Priore, R. E., Acitelli, L. K., & Barnes-Farrell, J. L.
(2006). Work-to-relationship conflict: Crossover effects in dual-earner
couples. Journal of Occupational Health Psychology, 11, 228 –240.
Matthews, R. A., Kath, L. M., & Barnes-Farrell, J. L. (2010). A short,
valid, predictive measure of work–family conflict: Item selection and
scale validation. Journal of Occupational Health Psychology, 15, 75–90.
http://dx.doi.org/10.1037/a0017443
Matthews, R. A., Swody, C. A., & Barnes-Farrell, J. L. (2012). Work hours
and work–family conflict: The double-edged sword of involvement in
work and family. Stress and Health, 28, 234 –247. http://dx.doi.org/
10.1002/smi.1431
Matthews, R. A., Wayne, J. H., & Ford, M. T. (2014). A work–family
conflict/subjective well-being process model: A test of competing the-
ories of longitudinal effects. Journal of Applied Psychology, 99, 1173–
1187. http://dx.doi.org/10.1037/a0036674
Mauthner, M. (1997). Methodological aspects of collecting data from
children: Lessons from three research projects. Children & Society, 11,
16 –28. http://dx.doi.org/10.1111/j.1099-0860.1997.tb00003.x
McDonald, R. P. (1999). Test theory: A unified approach. Mahwah, NJ:
Erlbaum.
Messick, S. (1995). Validity of psychological assessment: Validation of
inferences from persons’ responses and performances as scientific in-
quiry into score meaning. American Psychologist, 50, 741–749. http://
dx.doi.org/10.1037/0003-066X.50.9.741
Metz, S. M., Wyrwich, K. W., Babu, A. N., Kroenke, K., Tierney, W. M.,
& Wolinsky, F. D. (2007). Validity of patient-reported health-related
quality of life global ratings of change using structural equation mod-
eling. Quality of Life Research, 16, 1193–1202. http://dx.doi.org/
10.1007/s11136-007-9225-1
Molarius, A., & Janson, S. (2002). Self-rated health, chronic diseases, and
symptoms among middle-aged and elderly men and women. Journal of
Clinical Epidemiology, 55, 364 –370. http://dx.doi.org/10.1016/S0895-
4356(01)00491-7
Nagy, M. S. (2002). Using a single-item approach to measure facet job
satisfaction. Journal of Occupational and Organizational Psychology,
75, 77– 86. http://dx.doi.org/10.1348/096317902167658
Nunnally, J. C., & Bernstein, I. H. (1978). Psychometric theory (2nd ed.).
New York, NY: McGraw-Hill.
Parasuraman, S., & Greenhaus, J. H. (2002). Toward reducing some critical
gaps in work–family research. Human Resource Management Review,
12, 299 –312. http://dx.doi.org/10.1016/S1053-4822(02)00062-1
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for
research in the general population. Applied Psychological Measurement,
1, 385– 401. http://dx.doi.org/10.1177/014662167700100306
Raykov, T., & Marcoulides, G. A. (2011). Introduction to psychometric
theory. New York, NY: Taylor & Francis.
Rizzo, J. R., House, R. J., & Lirtzman, S. I. (1970). Role conflict and
ambiguity in complex organizations. Administrative Science Quarterly,
15, 150 –163. http://dx.doi.org/10.2307/2391486
Robins, R. W., Hendin, H. M., & Trzesniewski, K. H. (2001). Measuring
global self-esteem: Construct validation of a single-item measure and the
Rosenberg Self-Esteem Scale. Personality and Social Psychology Bul-
letin, 27, 151–161. http://dx.doi.org/10.1177/0146167201272002
Rogelberg, S. G., & Stanton, J. M. (2007). Introduction: Understanding and
dealing with organizational survey nonresponse. Organizational Re-
search Methods, 10, 195–209. http://dx.doi.org/10.1177/
1094428106294693
Sawyer, A. L., Bradshaw, C. P., & O’Brennan, L. M. (2008). Examining
ethnic, gender, and developmental differences in the way children report
being a victim of “bullying” on self-report measures. Journal of Ado-
lescent Health, 43, 106 –114. http://dx.doi.org/10.1016/j.jadohealth.2007
.12.011
Scarpello, V., & Campbell, J. P. (1983). Job satisfaction: Are all the parts
there? Personnel Psychology, 36, 577– 600. http://dx.doi.org/10.1111/j
.1744-6570.1983.tb02236.x
Schmidt, F. L., & Hunter, J. E. (2014). Methods of meta-analysis: Cor-
recting error and bias in research findings. Thousand Oaks, CA: Sage.
Schmitt, N. (1994). Method bias: The importance of theory and measure-
ment. Journal of Organizational Behavior, 15, 393–398. http://dx.doi
.org/10.1002/job.4030150504
Schriesheim, C. A., Hinkin, T. R., & Podsakoff, P. M. (1991). Can ipsative
and single-item measures produce erroneous results in field studies of
French and Raven’s (1959) five bases of power? An empirical investi-
gation. Journal of Applied Psychology, 76, 106 –114. http://dx.doi.org/
10.1037/0021-9010.76.1.106
Shirom, A., & Melamed, S. (2006). A comparison of the construct validity
of two burnout measures in two groups of professionals. International
Journal of Stress Management, 13, 176 –200. http://dx.doi.org/10.1037/
1072-5245.13.2.176
Stanton, J. M., Sinar, E. F., Balzer, W. K., & Smith, P. C. (2002). Issues
and strategies for reducing the length of self-report scales. Personnel
Psychology, 55, 167–194. http://dx.doi.org/10.1111/j.1744-6570.2002
.tb00108.x
Steffick, D. E. (2000). Documentation of affective functioning measures in
the Health and Retirement Study. Documentation Report DR-005. Ann
Arbor, MI: Survey Research Center, University of Michigan.
Steiger, J. H. (1980). Tests for comparing elements of a correlation matrix.
Psychological Bulletin, 87, 245–251. http://dx.doi.org/10.1037/0033-
2909.87.2.245
Sverke, M., Hellgren, J., & Näswall, K. (2002). No security: A meta-
analysis and review of job insecurity and its consequences. Journal of
Occupational Health Psychology, 7, 242–264. http://dx.doi.org/10.1037/
1076-8998.7.3.242
Tetrick, L. E., & Buffardi, L. C. (2006). Measurement issues in research on
the work-home interface. In F. Jones, R. J. Burke, & M. Westman (Eds.),
Work–life balance: A psychological perspective (pp. 90 –114). New
York, NY: Psychological Press.
Voydanoff, P. (2007). Work, family, and community: Exploring intercon-
nections. Mahwah, NJ: Erlbaum.
Wanous, J. P., & Hudy, M. J. (2001). Single-item reliability: A replication
and extension. Organizational Research Methods, 4, 361–375. http://dx
.doi.org/10.1177/109442810144003
Wanous, J. P., Reichers, A. E., & Hudy, M. J. (1997). Overall job
satisfaction: How good are single-item measures? Journal of Applied
Psychology, 82, 247–252. http://dx.doi.org/10.1037/0021-9010.82.2.247
Winefield, H. R., Winefield, A. H., & Tiggemann, M. (1992). Social
support and psychological well-being in young adults: The multi-
dimensional support scale. Journal of Personality Assessment, 58, 198 –
210. http://dx.doi.org/10.1207/s15327752jpa5801_17
Woods, S. A., & Hampson, S. E. (2005). Measuring the big five with single
items using a bipolar response scale. European Journal of Personality,
19, 373–390. http://dx.doi.org/10.1002/per.542
Received July 17, 2014
Revision received February 13, 2015
Accepted February 16, 2015 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
21
SINGLE-ITEM MEASURES
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Occupational Health Psychology
This content is subject to copyright. Terms and conditions apply.
