DOI: 10.1002/hrm.

22004

ORIGINAL ARTICLE

How does the availability and use of flexible leave influence

the employer–employee relationship?

Leroy White | Andy Lockett | Graeme Currie

Warwick Business School, University of

Warwick, Coventry, UK Abstract
In this article, we develop, and empirically test, a moderated mediation model of the
Correspondence
Leroy White, Warwick Business School, effects of flexible leave on employees' organizational attachment. Drawing on a social
University of Warwick, Coventry exchange framework and signaling theory, we explore how the material and non-
CV4 7AL, UK.
Email: [email protected] material nature of exchange between the employer and employee shapes their rela-
tionship. First, we show that the relationship between the availability of flexible leave
Funding information
British Academy, Grant/Award Number: and organizational attachment is shaped by two competing mediators, perceived orga-
SG120877 nizational support (POS) and perceived flexibility stigma. Second, we delineate
between availability and use of the policy, to show how the effect of POS is enhanced
and perceived flexibility stigma reduced, with use. Our findings demonstrate that the
relationship between the availability of flexible leave and organizational attachment is
complex, but is enhanced through use of flexible leave. We contribute toward HRM
scholarship about the relationship between employees' experience of HR practices and
their corresponding impact on employees' subsequent behavior.

KEYWORDS

flexible leave, perceived flexibility stigma, perceived organizational support, signaling theory,
social exchange theory

1 | I N T RO DU CT I O N the lack of clarity about the relationship between flexible working

arrangements and employee outcomes may be due to three limita-
Flexible working arrangements are considered to be a positive tions of extant studies.
development, leading to their proliferation, associated growing media First, scholars have employed “flexible working arrangements” as
attention, and a burgeoning literature on their effectiveness (Kossek, an umbrella term, assuming the effects of all flexible working policies
Sweet, & Pitt-Catsouphes, 2006). Empirical evidence of their effec- will be uniformly positive across a range of, proximal to distal,
tiveness, however, has been less than clear; meta-analyses of the employee outcomes (Beauregard & Henry, 2009; Perry-Smith & Blum,
growing literature include Byron's (2005) study showing a positive 2000). Second, scholars have tended to assume a direct relationship
impact of flexible working arrangements and Mesmer-Magnus and between policies for flexible working and employee outcome, with
Viswesvaran's (2006) study showing insignificant results. In reviewing effects that are indirect in nature remaining under-researched
the evidence, Shockley and Allen (2007) conclude that, even within (T. D. Allen, 2001; Beauregard & Henry, 2009; Butts, Casper, & Yang,
meta analytic research, there is a great deal of equivocaility associated 2013). To date, absent is any consideration of the question as to how
with the effectiveness of flexible working arrangements. Furthermore, flexible leave policies indirectly shape employees' experiences before
in their narrative review of the literature, Kelly and co-authors (Kelly leading to employee outcomes (Kelly et al., 2008). Third, scholars have
et al., 2008, p. 306) suggest that despite the rapid adoption of flexible paid little attention to whether or not the relation to the relationship
working arrangements, “managers in employing organizations simply between flexible working and employee outcomes is shaped by the
do not know whether and which organizational initiatives … are likely difference between the availability and use of such arrangements
to impact employees and the organization as a whole.” We suggest (Beauregard & Henry, 2009; Kelly et al., 2008). Differentiating

Hum Resour Manage. 2019;1–17. wileyonlinelibrary.com/journal/hrm © 2019 Wiley Periodicals, Inc. 1

2 WHITE ET AL.

between the availability and use of flexible leave enables us to two mediators, we contribute to extant scholarship by exploring how
address the question as to when, that is, under what conditions does flexible leave shapes employee outcomes.
flexible leave matter? In this article, we focus on one specific form of Third, we explore the differential effect of the availability and use
flexible working arrangement, flexible leave, to explain how the avail- of the policy (T. D. Allen, Johnson, Kiburz, & Shockley, 2013; Butts
ability and use of flexible leave shape employees' organizational et al., 2013; Kelly & Moen, 2007; Kossek, 2005; Kossek, Barber, &
attachment. Winters, 1999). We do so by developing a moderated mediation
We address the three limitations of extant studies, and contribute model to examine the difference between availability and use of the
to extant scholarship, as follows. First, we focus on a specific flexible policy on the relationship between our mediators (POS and perceived
working arrangement, flexible leave, because it is a flexible working flexibility stigma) and employee outcomes. A moderated mediation
arrangement that is available to all employees, regardless of family approach enables us to contribute to extant scholarship by exploring
responsibilities and circumstances (Smithson & Stokoe, 2005). Flexible the question as to when does flexible leave have an effect on
leave, however, remains the least understood of flexible working employee outcomes, through our two mediators POS and perceived
arrangements in terms of its effects on employee outcomes (Baltes, flexibility stigma. Simply stated, the moderated mediation model
Briggs, Huff, Wright, & Neuman, 1999; Kelly et al., 2008). In addition, enables us to simultaneously address the how and when questions
we note that extant studies have employed a wide range of outcomes relating to the relationship between flexible leave and employee
relating to employees' attitudes, ranging from measures of employees' outcomes.
perceptions of work–life balance well-being, and job satisfaction, to Theoretically, bringing the foci of our study together, we contrib-
organizational attachment encompassing both employees' organiza- ute toward HRM scholarship that has become increasingly concerned
tional commitment and turnover intentions. In this study, we focus on with the relationship between employees' experience of HR practices
the latter, employees' organizational attachment, because it is a general and their corresponding impact on employees' subsequent behavior
construct that encompasses the socio-emotional outcomes relating to (Alfes, Bailey, Shantz, & Soane, 2012; Baluch, 2017; Bos-Nehles &
employer–employee relationship (Lee & Mitchell, 1994), comprising Veenendaal, 2017; Bowen & Ostroff, 2004; Jiang, Hu, Liu, & Lepak,
both attitudinal (e.g., organizational commitment) and behavioral 2017; Purcell & Hutchinson, 2007; Purcell & Kinnie, 2006). Specifically
(e.g., turnover intention) components (Labianca & Brass, 2006). focusing upon flexible leave, we highlight the relationship between the
Second, to address the how question we develop a mediation availability of flexible leave and organizational attachment is complex,
model to link the availability of flexible leave to employee outcomes but is enhanced through actual use of flexible leave.
through their effect employees' attitudes toward their employer. Empirically, we examine the introduction of flexible leave at the Brit-
Drawing on social exchange theory (SET) (see Cropanzano & Mitchell, ish Broadcasting Corporation (BBC) in the UK. At the time of our study,
2005 for a review) and signaling theory (Spence, 1973), we explore the BBC was going through a period of cutbacks, which exemplifies a
the process of social exchange between the employer and the context within which recent calls for research to investigate HR practices
employee, which are characterized by interdependency and contin- in organizations undergoing austerity, can be addressed (Ruth-Eikhof &
gency, involving material as well as non-material exchange (Blau, Warhurst, 2013). We employed a longitudinal quasi-experimental
1964). Cropanzano and Mitchell (2005) suggest that employers pro- approach, with all participants surveyed when flexible leave policy was
vide the first “signal” of a desire for a closer working relationship with first introduced, and then 12 months later, which enabled us to delineate
their employees, through the introduction of a new HRM practice between the availability and use of the policy. To test our model, we
such as flexible leave. Employers hope that their actions will be recip- drew on a recent innovation in mediation modeling approach based on a
rocated in a positive manner by employees. It is through such signal- path analysis, which is derived from developments in structural equation
response incidents that employers intend to develop high quality modeling (Kline, 2015), and the simultaneous testing of multiple indirect
employer–employee relationships. effects (Preacher & Hayes, 2008).
Due to the equivocal findings of extant studies we include two
mediators to account for the potentially positive and negative impact
of flexible leave, perceived organizational support (POS) and per- 2 | M O D E L D E V E LO P M E N T
ceived flexibility stigma, respectively. We employ POS because it is a
key concept in SET, which acts as a mediator in linking HRM practices Flexible working spans a considerable range of organizational prac-
to employee outcomes (Eisenberger, Huntington, Hutchison & Sowa, tices, including flexible work hours, working from home, sharing a full-
1986). Our inclusion of perceived flexibility stigma stems from the time job between two employees (job sharing), family friendly leave
recent emergence of a literature about the potential dark side of poli- programs (e.g., parental leave, adoption leave, and compassionate
cies promoting flexible working (see Perrigino, Dunford, & Wilson, leave), on-site childcare, and financial and/or informational assistance
2018, for a narrative review). In their review, Perrigino et al. (2018) with childcare and eldercare services (Shockley & Allen, 2010). To
highlight that it is important to incorporate the dark side into studies date, there is a lack of studies that examine the effects of specific flex-
of the effects flexible working because the dark side is pervasive ible working policies on specific employee outcomes (Moen, Kelly, &
across different types of policies, and it enables a positioning of atti- Chermack, 2009). Each form of flexible working, however, should be
tudes across a broad spectrum from positive to negative. In employing treated as a distinct policy (Kossek & Nichol, 1992).
WHITE ET AL. 3

To aid with definitional clarity, scholars differentiate between To examine the nature of social exchange, involving the non-
flexibility associated with place of work (referred to as flexplace) and material benefit that stems from the availability of flexible leave, we
flexibility associated with scheduling of work (referred as flextime) draw on the related concept of signaling theory. Signaling theory was
(T. D. Allen et al., 2013). The distinction is important because flextime developed to examine how employers assess the unobservable attri-
policies are mainly used to manage various non-work and personal butes of potential employees (see Spence, 1973). Following Spence,
concerns, whereas flexplace practices are useful only to those with scholars have broadened the focus of signaling theory to examine
current or future dependents (Casper & Harris, 2008). We focus on employees' interpretations of the actions of an organization as signals
flextime, because unlike some forms of flexible working, it is a policy of unobservable intentions and motives (Goldberg & Allen, 2008).
that is available to all employees regardless of family responsibilities Applied to flexible leave, we suggest that availability of the policy pro-
and circumstances (Smithson & Stokoe, 2005). Specifically, we focus vides the employer's first “signal” of a desire for a closer working rela-
on the flextime policy of flexible leave, following calls to focus on the tionship with their employees, which, independent of use, the
effectiveness of specific policies (Gajendran & Harrison, 2007). Flexi- employer hopes that the non-material benefit will be reciprocated in a
ble leave constitutes taking leave when it suits the employee, thereby positive manner by employees (Cropanzano & Mitchell, 2005). Grover
allowing the employee greater control and autonomy over their and Crooker (1995, p. 274) suggest that family-supportive policies
work time. “symbolize corporate concern, and such perceived care for employees
To explore the impact of the availability and use of flexible leave may be construed positively by employees regardless of whether they
on employee outcomes, we draw on SET and signaling theory to personally benefit.” If successful, the non-material benefit will help
develop a moderated mediation model as outlined in the following. establish high-quality exchange relationships that create obligations
Specifially, we address the issue as to how POS and perceived flexibil- for employees to reciprocate in positive, beneficial ways (Eisenberger,
ity stigma mediate the relationship between the availability of flexible Huntington, Hutchison, & Sowa, 1986; Shore & Wayne, 1993).
leave and employee outcomes. Rather than having a direct effect on employees' perceptions of
the employer–employee relationship, we suggest that the effect of
the availability of flexible leave will be indirect in nature. We do so
2.1 | The mediators of flexible leave and employee because it is naïve to assume that the effect of a flexible working pol-
outcomes icy will have a direct effect on employee behavior, particularly given
that employee responses to HR practices are at the heart of all HRM-
SET is a key conceptual framework for explaining the link between performance models (Alfes et al., 2012; Baluch, 2017; Bos-Nehles &
employees' experience of, and attitudes toward, their employer Veenendaal, 2017; Bowen & Ostroff, 2004; Jiang et al., 2017; Pur-
(Cropanzano & Mitchell, 2005). At its core, SET focuses on the series cell & Hutchinson, 2007; Purcell & Kinnie, 2006). Given the equivocal
of interactions that create obligations between two or more parties findings of extant studies (see Byron, 2005; Kelly et al., 2008; Kossek,
(Emerson, 1976), with the interactions being interdependent and Lautsch, & Eaton, 2006; Mesmer-Magnus & Viswesvaran, 2006), we
contingent on the actions of others (Blau, 1964). Over time, include two mediators to account for the potentially positive and neg-
the interdependent interactions may lead to the development of ative impact of flexible leave, POS and perceived flexibility stigma,
high-quality reciprocal relationships, defined by trust, loyalty and respectively.
mutual commitment (Cropanzano & Mitchell, 2005). Reciprocal Following the lead of Casper and Harris (2008), we employ POS
interdependence brings into focus the contingent nature of interac- because it is a key concept in SET, which acts as a mediator in linking
tions and the exchange underpinning them. Reciprocal exchange does HRM practices to employee outcomes (Eisenberger et al., 2002). POS
not include explicit bargaining (Molm, 2000, 2003), rather one party is defined as employees' perception of the degree to which “an orga-
provides a benefit, and if the other reciprocates, new rounds of nization values their contribution and cares about their well-being”
exchange follow (Cropanzano & Mitchell, 2005). (Eisenberger et al., 1986, p. 501). We argue that the introduction of
An unresolved issue within SET, however, surrounds the causal flexible leave sends a signal to employees that they are valued, which
order of exchanges between parties and their impact on the nature of will enhance the employees' POS; in turn, POS will enhance
relationships (Cropanzano & Mitchell, 2005). In the case of flexible employees' organizational attachment as they reciprocate the non-
leave, the causal order of exchanges is shaped by the distinction material benefit gained from their employer. Evidence to support the
between the use and availability of the policy, which mirrors SET's dis- hypothesis that POS mediates the relationship between the availabil-
tinction between material and non-material exchange (see Homans, ity of flexible leave and organizational attachment is present in extant
1958). Use of flexible leave provides a material benefit for the literature. For example, there is strong empirical support for the rela-
employee in being able to take leave at their discretion. In contrast, tionship between a range of family supportive practices and POS
the availability of flexible leave, independent of use, provides non- (Kurtessis et al., 2017). In addition, there is strong evidence that POS
material benefit to the employee through the psychological and/or shapes the attitudes of employees toward their employer (Rhoades &
social gain that comes from feeling that their employer is supportive Eisenberger, 2002).
of them. In Homans's (1958), p. 606) terms, non-material benefit Hence, we suggest that the availability of flexible leave will send
derives from “the symbols of approval or prestige.” a signal from the employer to the employee that will increase
4 WHITE ET AL.

employees' POS, which will in turn increase their organizational and/or use of the policy. Paradoxically, the act of making the policy of
attachment as measured by organizational commitment and turnover flexible leave available will affect the social exchange relationship
intention. The above-mentioned argument leads us to hypothesize between the employer and employee, such that employees' perceive
H1a and H1b. there being an elevated risk of flexibility stigma. In turn, an elevated
level of perceived flexibility stigma will lead to a reduction in
H1a: POS mediates the positive relationship between employees' organizational attachment, through a reduction in their
the availability of flexible leave and organizational organizational commitment and intentions to remain with their
commitment, with higher POS being associated with employer (Cech & Blair-Loy, 2014).
higher organizational commitment. Hence, we suggest perceived flexibility stigma will mediate the
relationship between the availability of flexible leave and employee's
H1b: POS mediates the negative relationship between organizational attachment. Specifically, the availability of flexible leave
the availability of flexible leave and turnover intention, will elevate employees' perceived flexibility stigma, and the fear that
with higher POS being associated with lower turnover they will send a negative signal to their employer, which in turn will
intention. reduce employees' organizational attachment as measured by organi-
zational commitment and turnover intention. Based on the above-
Although the organizational intention of introducing a policy of flexi- mentioned argumentation, we suggest H2a and H2b.
ble working is to send a signal to the employee that they are valued,
the uptake of flexible working practices has been relatively slow H2a: Perceived flexibility stigma mediates the negative
(Tipping, Chanfreau, Perry, & Tait, 2012). We suggest that slow relationship between the availability of flexible leave
uptake of flexible working practices may be due in part to their poten- and organizational commitment, with higher perceived
tial dark side; that is, fear of employees that use of such practices dis- flexibility stigma being associated with lower organiza-
adavantages them. To date, consideration of this dark side has been tional commitment.
absent from studies linking flexible working to employees' organiza-
tional attachment (Perrigino et al., 2018). To explore how the dark side H2b: Perceived flexibility stigma mediates the positive
of a policy of flexible leave may shape employees' organizational relationship between the availability of flexible leave
attachment, we again explore the indirect effect of employees' and turnover intention, with higher perceived flexibility
responses (see Purcell & Hutchinson, 2007). stigma being associated with higher turnover intention.
We model the indirect effect of employees' response to the dark
side of flexible working by focusing on the dominant reason
employees give for not adopting flexible working arrangements. Spe- 2.2 | Availability and use of flexible working:
cifically, employees' fear that adopting flexible working will send their Moderating effects
employer a signal that their personal life responsibilities diminish their
commitment to the organization, and that managers will penalize them We now shift our attention to exploring the issue as to when flexi-
accordingly (Blair-Loy & Wharton, 2002; Blair-Loy, Wharton, & ble leave has an effect on employee organizational attachment for
Goodstein, 2011; McCloskey & Igbaria, 2003; Williams, Blair-Loy, & both of our mediators (POS and perceived flexibility stigma). Spe-
Berdahl, 2013). Such a fear is termed flexibility stigma, which is cifically, we develop a moderated mediation model, differentiating
defined as the discrimination toward workers using various types of between the availability and use of the policy, and in doing so,
flexible working arrangements for family responsibilities and care pur- simultaneously address the how and when questions relating to the
poses (Williams et al., 2013). Employees' fear of flexibility stigma is relationship between flexible leave and employee outcomes. We
very real, with over a third of all workers in the UK work–life balance do so because we cannot assume that the introduction of flexible
survey of 2011 agreeing with the statement that those who work leave will have a uniform effect on employees' organizational
flexibly create more work for others (Chung, 2018). attachment. Rather, our interest lies in exploring how employees'
To date, scholarship on flexibility stigma has focused on examin- responses to flexible leave may change when they use the policy
ing the impact on employees of requesting, and/or taking up different (Grover & Crooker, 1995; Kelly et al., 2008; Thompson, Payne, &
forms of flexible working, in terms of career penalties (see Berdahl & Taylor, 2015).
Moon, 2013; Blair-Loy, 2003; Epstein, Seron, Oglensky, & Saute, We begin by examining the mediator of POS. In our mediation
1999). Following the lead of Cech and Blair-Loy (2014), our interest hypotheses H1a and H1b, we argue that the availability of flexible
lies in examining how employees' perceptions of flexibility stigma are leave provides a non-material benefit, which operates in the form of a
influenced by the introduction of a policy of flexible leave, indepen- signal from the employer to the employee that they are valued,
dent of an employees' use of the policy. We suggest that the intro- enhancing their POS and organizational attachment in turn. We sug-
duction of a policy of flexible leave will raise the visibility of flexible gest that use of the policy will strengthen the mediating effect of
working in an organization, and with it employees' fear of the signals POS, upon the relationship between the availability of flexible leave
they may send to their employer through their attitudes toward and organizational attachment, for two main reasons.
WHITE ET AL. 5

First, in relation to the signal sent from the employer to the et al., 2013). Employees' use of flexible leave, therefore, will enhance
employee through the availability of flexible leave, users of flexible an employee's social exchange relationship with their employer,
leave will view the policy as being fairer (Parker & Allen, 2001) and through dampening the mediating effect of perceived flexibility stigma
more targeted toward them (Lambert, 2000), as compared with non- on the relationship between the availability of flexible leave and orga-
users. Users of the policy, therefore, will gain a higher level of non- nizational attachment.
material benefit from the availability of the flexible leave as compared Based on the above-mentioned argument, we hypothesize the
with non-users. Second, users of flexible leave will gain an additional effect of the mediator of perceived flexibility stigma on the relation-
material benefit from use of the policy, as compared with those that ship between the availability of flexible leave and organizational
do not use the policy. Employing a logic of self-interest (see Lind & attachment, will be weaker for users as compared with non-users.
Tyler, 1988), we suggest that the effect will be pronounced as users Hence, H4a and H4b are as follows.
of flexible leave will value the policy more highly than non-users.
Taken together, we argue that employees' use of the flexible leave will H4a: Use of flexible leave moderates the indirect rela-
provide them with additional non-material and material benefits, tionship between the availability of flexible leave and
which will enhance the social exchange relationship with their organizational commitment in such a way that the
employer, such that the relationship between POS and organizational effect perceived flexibility stigma on organizational
attachment will be stronger for users as compared with non-users. commitment is weaker for users as compared with
Based on the above-mentioned argument, we hypothesize that non-users of flexible leave.
the effect of the mediator of POS on the relationship between the
availability of flexible leave and organizational attachment will be H4b: Use of flexible leave moderates the indirect rela-
stronger for users as compared with non-users. Hence, H3a and H3b tionship between the availability of flexible leave and
are as follows: turnover intention in such a way that the effect per-
ceived flexibility stigma on turnover intention is wea-
H3a: Use of flexible leave moderates the indirect rela- ker for users as compared with non-users of flexible
tionship between the availability of flexible leave and leave.
organizational commitment in such a way that the
effect of POS on organizational commitment is stron- A summary of our theoretical model is presented in Figure 1.
ger for users as compared with non-users of flexible
leave.
3 | DA T A A N D M E T H O D S
H3b: Use of flexible leave moderates the indirect rela-
tionship between the availability of flexible leave and Our study took place at the BBC during a period of cutbacks, when
turnover intention in such a way that the effect of POS the organization was subject to significant budget cuts in the run up
on turnover intention is stronger for users as compared to the renewal of its charter in 2016–2017. A pressing need to
with non-users of flexible leave. respond to performance measures imposed by BBC stakeholders and
consumer groups had led to the introduction of a range of new mea-
We now turn to our mediator of perceived flexibility stigma. In our sures: a more flexible workforce; reducing the number of senior man-
mediation hypotheses H2a and H2b, we argue that the availability of agers and flattening the structure; modernizing terms and conditions
flexible leave will elevate employees' perceived flexibility stigma, via for staff; increasing out-of-London production; and reducing the
the fear that they will send a negative signal to their employer, which BBC's property estate. Job cuts were announced in October 2011,
in turn will reduce employees' organizational attachment. We suggest which led to the Unions claiming staff morale was at an all-time low.
that use of flexible leave will dampen the mediating effect of per- At the same time, the BBC remained an employer enacting good HR
ceived flexibility stigma, on the relationship between the availability practices as they compete for, and seeked to retain, talent.
of flexible leave on organizational attachment, as follows. In an attempt to balance the competing tensions, of the need for
Employees' use of flexible leave will provide the employer an austerity and to attract and retain talent, management (including the
opportunity to demonstrate their positive intention toward, and com- HR Director) and trade union representatives sought to introduce
mitment to, the introduction of the policy. As employees use the pol- flextime as a means of improving job outcomes for staff. In this study,
icy, they will gain confidence about their employer's positive we focus on the introduction of a flexible leave policy, which allowed
intentions for, and commitment to flexible leave, and the security of employees to choose and distribute their leave entitlement, with its
knowledge that they are not being judged or penalized for doing aim being to give employees maximum latitude in scheduling their
so. Accordingly, employees' use of flexible leave will reduce their per- leave. The introduction of flexible leave policy was clearly communi-
ceived flexibility stigma, as their fear of being negatively judged or cated to all employees participating in the initiative, with managers
penalized by their employers is diminished (Blair-Loy et al., 2011; also receiving guidance on how to handle a flexible leave request with
Blair-Loy & Wharton, 2002; McCloskey & Igbaria, 2003; Williams a clear steer to approve requests whenever possible.
6 WHITE ET AL.

F I G U R E 1 A moderated mediation model

Perceived
of the influence of flexible leave on
Organizational
Support organizational attachment
H1 H1

H3
Organizational attachment:
Availability of Use of Flexible (a) Organizational commitment
Flexible Leave Leave
(b) Turnover Intention
H4

H2
H2
Perceived
Flexibility Stigma

Hypothesized effect
Non-Hypothesized effect

3.1 | Data the research team collected monthly leave records. Each month the
researchers would ask each participant in the experimental group how
We adopted a quasi-experimental design for the study (Shadish, much leave they had taken that month. The data on leave was not
Cook, & Campbell, 2002), in order to compare an intervention group shared with the employer. Finally, in Stage 4 (1 month after the pilot),
(which were subjected to the introduction of the flexible leave we administered the post-test survey to all employees (in both sites)
policy—located in the Bristol office) with a near-equivalent control when the flexible leave policy was implemented and then 12 months
group (no introduction of the policy—located in the London office) at later. The use of repeated measurements in our design allowed each
two points in time with researchers controlling for pre-existing differ- individual to be used as his or her own control, which typically
ences, allowing estimates of the intervention's effectiveness, even increases the power and precision of statistical tests (Schmidt &
when intervention and control groups are non-equivalent (Shadish Hunter, 2004). The timing of the posttest measurement is an impor-
et al., 2002). As we can only assume our estimators are unbiased tant aspect of our research and 12 months provided sufficient dura-
when using a randomized research design, it was advantageous there tion for the effects to take place (Baltes et al., 1999).
was no interference between the intervention group and the control
group as they were based on different sites, some distance apart
(Bristol and London). Thus, we avoided contamination effects, where 3.2 | The sample
members of the intervention group may influence members of the
control group and vice versa (Cook & Campbell, 1976). The participants for the study were a heterogeneous sample rep-
The data collection involved four stages. Stage 1 (3 months prior resenting a wide cross-section of employees in terms of gender, age,
to the pilot) involved a “town hall” meeting with all staff at Bristol, organizational tenure, level of education, or their rank in the organi-
which was conducted to inform and discuss the introduction of the zation. Participants were employees from three sections of the BBC
flexible leave scheme to be piloted. Stage 2 (1 month prior to the in both London and Bristol (Natural History Unit, Features and Doc-
pilot) involved an online pre-test questionnaire, which was adminis- umentaries). Participation in the study was limited to those
tered to all staff at both sites 1 month before the 12-month pilot of employees who were on full-time contracts and fixed term contracts
the new flexible leave scheme. The online survey captured data relat- of more than a year. At Time 1, the survey was completed by 235 of
ing to our model variables (as detailed later in the article). Scale items 250 potential respondents, giving a response rate of 87%. There
were presented in randomized order within question blocks to reduce were 66.5% women and 33.5% men in the intervention group and
order effects and common method bias (Podsakoff, MacKenzie, 68.6% women and 31.4% men in the control group. The average age
Podsakoff, & Lee, 2003). The survey was linked to employee records of participants was 33.34 years, and their average tenure was
in order to collect data on demographic characteristics of gender, edu- 39.76 months. At Time 2, the same members were asked to com-
cation, tenure, job type (full-time or part-time), and age group. In plete the survey, including those who did not respond at Time 1 or
Stage 3, the pilot scheme was rolled out for 12 months and individuals had joined since Time 1. The survey was completed by 222 of
in the intervention group were asked whether they wished to opt in 250 potential respondents, giving a response rate of 85%. In terms
or opt out of the pilot. Here, the researchers had no control over of robustness check, Chi-square and t-tests confirmed there were no
assignment to intervention condition. Following the start of the pilot, differences between the sample of employees that were in the
WHITE ET AL. 7

intervention group and those who formed our control group in terms 1 (strongly disagree) to 7 (strongly agree). The Cronbach's alpha for
of their gender, age, organizational tenure, level of education, or perceived flexibility stigma was 0.81.
their rank in the organization. Chi-square and t-tests also confirmed Availability of flexible leave policy was assessed as an individual-
that there were no differences between our sample for the study item dichotomous measure, where 1 indicates the availability of the
and the total possible sample on any of the above-mentioned policy (intervention site) and 0 otherwise.
demographics. Use of the arrangement is measured as a binary variable and
The longitudinal sample, comprising those who responded at both reflects whether the members of the treatment group used the flexi-
Time 1 and Time 2, was 210, of which there were 125 team members ble leave policy for the 12 month period. A value of 1 indicates that
in the intervention group and 85 in the control group. Of the 210 Time the person had used the policy for the 12 months and 0 they had not
1 respondents, 204 potential respondents remained at Time 2, as six used the policy.
employees had either moved elsewhere within the organization, quit Finally, we also included control variables including tenure (num-
voluntarily, or taken maternity leave. The response rate for the Time ber of years with the BBC), gender (0 = female, 1 = male), and age,
1 respondents remaining at Time 2 was 97.1%. Finally, 60% of the which has been identified in extant research as a significant correlate
intervention group opted to use the arrangement. of the measures of interest in this study (C. Meyer, Mukerjee, &
To test for the effects of participant attrition, we followed the Sestero, 2001). Given the importance of gender to previous studies of
lead of Goodman and Blum (1996), conducting a logistical regression flexible working, although not the focus of our study, we comment on
in which the dependent variable is a dichotomous variable rep- any gender specific effects in our results section. In addition, we also
resenting those present at Time 1 and Time 2 (stayers) and those rep- examined other sample characteristics that may affect our predicted
resented at Time 1 but had left in Time 2 (leavers). All main study effects, employees who are married-cohabiting (1 = married-
variables were entered as independent variables. No significant effect cohabiting, 0 = otherwise) and/or who have dependents (1 = with
was found, indicating that participant attrition should not bias the dependents, 0 = otherwise).
results. All variables were mean centered to provide a meaningful zero
point for interpreting the results (as suggested for mediation stud-
ies by Judd, McClelland, & Culhane, 1995), whereby tests of the
3.3 | Measures interactions could be carried out, and also to reduce the likelihood
that multicollinearity could influence the results (Aiken, West, &
We employed five model variables to test our conditional mediated Reno, 1991).
model that examines the effect of job outcomes (organizational
commitment and turnover intention), arising from the impact of
flexible working arrangements on POS and perceived flexibility 3.4 | Statistical method
stigma. The measure for each model variable is detailed in the
following, for which we compute the Cronbach's alpha for all multi- Although our study did not conform to the strict criteria for a ran-
item scales. All Cronbach's alpha scores were all 0.80 or above domized control trial, our data set is characterized by observations
indicating a high degree of internal consistency in responses to the drawn from the same participants at two points in time and across
individual questions. two locations (i.e., we have longitudinal data for both a treatment
Organizational commitment was measured using N. J. Allen and group and a control group). Using the difference-in-difference (DID)
Meyer's (1996) abridged five-item version of J. P. Meyer and Allen's approach to longitudinal pre-test–post-test design, we were able to
(1984) affective commitment scale. The items reflect the respondent's consider differences between the mediators and dependent vari-
commitment based on emotional attachment the employee develops ables between Time 1 (pre-test) and Time 2 (post-test) (Shadish
with the employer, predominantly from positive work-based experi- et al., 2002). In doing so, we imposed two assumptions in order to
ences, and were measured using a five-point Likert scale. Cronbach's restrict the scope of possible bias (MacKinnon, 2008). The first
alpha for organizational commitment was 0.81. assumption is that the treatment variable (i.e., exposure to the avail-
Turnover intention was measured using the four-item scale ability of flexible leave) is time-invariant. We also included a time
employed by Colarelli (1984), involving a five-point Likert scale. The variable that indicated observations at Time 2 (controlling for
Cronbach's alpha for turnover intention was 0.89. Time 1), with the treatment effect being the product of both the
POS was measured using an eight-item scale with respondents treatment and time variable. In doing so we made our second
asked to indicate the extent to which they disagreed or agreed with assumption, the common trend assumption, that the confounders
the statements, ranging from 1 (strongly disagree) to 5 (strongly agree) varying across the intervention and control groups are time-invari-
(Eisenberger et al., 1986). The Cronbach's alpha for POS measure ant, and time-varying confounders are group-invariant.
was 0.81. The empirical testing of multiple moderation and mediation
1
Perceived flexibility stigma was measured using the four-items models is methodologically complex, and so we draw on recent devel-
suggested by Blair-Loy (2003) with respondents asked to indicate opments in structural equation modeling (Kline, 2015) and the simul-
the degree to which they disagreed or agreed ranging from taneous testing of multiple indirect effects (Preacher & Hayes, 2008).
8 WHITE ET AL.

We employ a path-analytic framework (Preacher & Selig, 2012), mediation effects non-parametrically, through bootstrapping, and
expressing the relationships between our constructs via regression then using the estimations to generate confidence intervals for the
equations, and incorporate moderation by extending these equations moderated mediation effects (Muller et al., 2005; Preacher & Hayes,
with the moderator variable and its product with the mediator vari- 2008). Our approach is considered to be most effective for small
ables (Edwards & Lambert, 2007). The path-analytic approach has a samples and least vulnerable to Type I errors (Preacher & Hayes,
number of advantages. First, mediation can be performed through a 2008). The resultant confidence interval, when not containing the
single test of the indirect effects (Hayes, 2013; Preacher & Hayes, value of zero, demonstrates there is a significant difference in the
2004; Preacher & Selig, 2012), which reduces likelihood of making an change of coefficients for the test of mediation. The results of our
inferential error (Hayes, 2013). Second, a path-analytic framework non-parametric bootstrapping approach are presented in Table 4
facilitates the testing of multiple mediators, with the ability to com- (Panels A and B).
pare the relative magnitude of indirect effects through the mediators,
and assess whether the mediators are independent of one another
(MacKinnon, 2000; MacKinnon, 2008; Preacher & Hayes, 2008). 4 | EMPIRICAL FINDINGS
Third, a path-analytic approach enables us to model the indirect effect
as a function of another variable, allowing researchers to test how Table 1 presents the correlations among all the study variables for the
indirect effects depend on other variables to explore moderated medi- samples within time period 1 and time period 2. The correlations show
ation (Preacher, Rucker, & Hayes, 2007). Our approach involves a the model variables are not confounded by differences in levels of
focus on the estimation of interactions between the moderator and married-cohabiting, tenure and age, as judged by the absence of sta-
the pathways that defined the indirect effect (Edwards & Lambert, tistically significant correlations with the model variables. In addition,
2007; Muller, Judd, & Yzerbyt, 2005; Preacher et al., 2007), which we found that the variables for married/cohabiting were correlated
emphasizes the estimation of conditional indirect effects (i.e., the with the variable for dependents. Consequently, married-cohabiting,
value of indirect effects conditioned on values of the moderator). The tenure, and age are excluded from further analysis to reduce the num-
results for our path model of mediation and moderation is presented ber of parameters estimated, and because analysis that includes
in Table 3. unnecessary control variables may yield biased parameter estimates
For further tests of the indirect effects of our mediators and (Becker, 2005). We retained gender and dependents because they
moderators, rather than employing the Sobel test (Sobel, 1982), were significantly correlated with our model variables, and previous
which is commonly used to further establish the mediation (and studies of flexible working arrangements have identified them as key
moderated mediation) effects (as suggested by Baron & Kenny, predictors of job outcomes (Beauregard & Henry, 2009). Finally, we
1986), we adopt a non-parametric approach (Efron & Tibshirani, can see from Table 1 there is substantial correlation between the
1993; MacKinnon, Lockwood, & Williams, 2004). We do so because same variable measured at pre-test time and post-test time, indicating
the Sobel test requires all coefficients to be normally distributed, that our relationships among variables are stable over time.
which is usually not true for smaller samples (MacKinnon, Lockwood, Table 2 shows that all model variables were similar across the
Hoffman, West, & Sheets, 2002). Specifically, we employed a non- intervention and control groups for time period 1, with no significant
parametric bootstrapping approach, which has different assumptions differences in group means. In contrast, in period 2, we found signifi-
relating to normal distribution, symmetries, and large sample sizes cant differences across the means for the intervention group and the
(Bollen & Stine, 1990; Shrout & Bolger, 2002). Our approach control group for all model variables (all at p < .5). When turning to
involved estimating the sampling distribution of the moderated the comparison of means across time, as expected, we found no

TABLE 1 Correlations between main study variables for periods 1 and 2

1 2 3 4 5 6 7 8 9
Organizational commitment (1) .76** .36** 0.23* .31** −.07 −.14 −.05 −.01 .08
Turnover intention (2) .30** −.73* .12* −.47** −.10 −.05 .01 −.04 .01
Perceived flexibility stigma (3) −.10 .18* −.63* .03 −.16* −.04 −.07 −.03 .01
perceived organizational support (POS) (4) .24** −.48* .06 .75* .06 −.24* .16 .04 .04
Gender(5) −.05 −.08 −.00 .13 - −.10 .09 .02 .05
Dependents (6) −.13 −.12 −.07 −.19* −.10 - .42* −.34* .03
Married-cohabiting (7) −.05 .02 −.09 .11 .09 .42* - −.03 .04
Age (8) −.01 −.06 −.06 .06 .02 −.34* −.03 - .16*
Tenure (9) .06 −.04 .01 .03 .05 .03 .04 .16* -

Note: Correlations below the diagonal are for period 1 (N = 235); correlations above the diagonal for period 2 (N = 222); diagonals represent correlations
between period 1 and period 2; correlations significant at *p < .05 and **p < .01.
WHITE ET AL. 9

significant differences for all model variables for the control group. In model 3, availability has a positive and significant relationship with
contrast, we found significant differences for organizational commit- organizational commitment (coefficient = 0.17, p < .05) indicating par-
ment, POS and perceived flexibility stigma (all at p < .05), but no sig- tial mediation, but in model 4, the availability is not significantly asso-
nificant differences for turnover intention. ciated with turnover intention indicating full mediation. For models
Our anlaysis progressed in a series of stages. First, tested our 3 and 4, we also note that our treatment effects were statistically sig-
mediation and moderated mediation models employing ordinary nificant, with the intervention group is seeing a reduction in organiza-
least squares regression analysis, incorporating DID as our data tional commitment (model 3: −0.12, p < .05) and turnover intention
are quasi-experimental in nature. Table 3 outlines the different (model 4: −0.16, p < .05) as compared with the control group.
elements of our path model. First, we examined the regression Third, we examined the moderated mediation effects (models
coefficients for the effects of the availability of the flexible leave 5 and 6), the path models for which are presented in Figures 2 and
policy on our mediators, POS and perceived flexibility stigma 3. The inclusion of the moderator of use of flexible leave led to a
(models 1 and 2). The results indicate that the availability of flexi- significant increase in the predictive power of our models (model
ble leave is associated with the mediators, having a positive and 5 R2 = 0.15, p < .01; model 6 R2 = 0.26, p < .01). The results for
significant effect on POS (coefficient = 0.38, p < .05) and a posi- POS indicate that the mediator remained positive and significantly
tive and significant effect on perceived flexibility stigma (coeffi- related to organizational commitment (coefficient = 0.34, p < .01)
cient = 0.2, p < .05). and negatively and significantly related to turnover intention (coef-
Second, we examined the paths for our mediators on our out- ficient = −0.35, p < .01). In addition, the coefficient for the cross
come variables of organizational commitment (model 3) and turnover product (POS × use) was positive and significant for organizational
intention (model 4). Model 3 indicates that POS has a positive and sig- commitment (coefficient = 0.11, p < .05) and negative and signifi-
nificant (coefficient = 0.31, p < .01), and perceived flexibility stigma a cant for turnover intention (coefficient = −0.15, p < .05). The
negative and significant (coefficient = −0.06, p < .05), relationship results for perceived flexibility stigma indicate that the mediator
with organizational commitment. Model 4 indicates that POS has a remained negatively and significantly related to organizational
negative and significant (coefficient = −0.35, p < .01), and perceived commitment (coefficient = −0.04, p < .05) and positively and signif-
flexibility stigma a positive and significant (coefficient = 0.02, p < .01), icantly related to turnover intention (coefficient = 0.02, p < .01). In
relationship with turnover intention. Turning to the non-hypothesized addition, the coefficient for the cross product (perceived flexibility
paths, indicating the direct effect of availability of flexible leave, in stigma x use) was not significant for organizational commitment
but was negative and significant for turnover intention (coeffi-
cient = −0.16, p < .05). We also note from Table 3, our estimates of
T A B L E 2 Comparison of means between intervention and control the treatment effect were significant (model 5: −0.12, p < .05;
groups for periods 1 and 2 model 6: −0.16, p < .05).
Intervention Control group Fourth, we formally tested our model of mediation using the
group mean (SD) mean (SD) bootstrapped method as an alternative to the Sobel Test, and
Organizational commitment (1) 3.91 4.02 employing confidence limits as a test for mediation rather than signifi-
(0.91) (0.75) cance tests (Preacher & Selig, 2012; Shrout & Bolger, 2002), the

Organizational commitment (2) 4.08a 4.00b results for which are presented in Table 4 (Panels A and B). The results
in Table 4 (Panel A) show, based on 5,000 bootstrap samples, the 95%
(0.91) (0.76)
confidence interval range for POS and organizational commitment
Turnover intention (1) 2.62 2.63
(coefficient = 0.31, 95% CI: 0.11 to 0.61) was positive and different
(0.95) (0.95)
from zero and POS and turnover intention (coefficient = −0.35, 95%
Turnover intention (2) 2.55 2.62b
CI: −0.27 to −0.03) was negative and different from zero, supporting
(0.91) (0.93)
H1a and H1b. The results for perceived flexibility stigma were similar,
perceived organizational 3.77 3.75 with zero lying outside the 95% confidence interval for perceived flex-
support (POS) (1)
ibility stigma and organizational commitment (coefficient = −0.03,
(1.36) (1.45)
95% CI: −0.11 to 0.06) and outside the 95% confidence interval for
POS (2) 3.99a 3.78b
perceived flexibility stigma and turnover intention (coefficient = 0.13,
(1.31) (1.43) 95% CI: 0.03 to 0.26). The results lead us to support both H2a
Perceived flexibility stigma (1) 4.09 4.03 and H2b.
(1.30) (1.32) We note that the availability of flexible leave had a direct effect
Perceived flexibility stigma (2) 3.81a 4.02b on organizational commitment and no direct effect on turnover inten-
(1.38) (1.31) tion, but when we compared the indirect effects with the direct effect
a of the availability of flexible leave on our outcome variables we found
Significant difference across time within the group, p < .05.
b
Significant difference between intervention group and control group full mediation of turnover intention. In addition, we note that POS
within time point, p < .05. and perceived flexibility stigma partially mediated the effects of the
10 WHITE ET AL.

TABLE 3 Regression results for testing mediation and moderated mediation

Organizational Organizational attachment

Mediating variables attachment (mediation) (moderated mediation)

perceived
organizational Perceived Organizational Turnover Organizational Turnover i
support (POS) flexibility stigma commitment intention commitment ntention
Coefficient (SE) Coefficient (SE) Coefficient (SE) Coefficient (SE) Coefficient (SE) Coefficient (SE)
(Model 1) (Model 2) (Model 3) (Model 4) (Model 5) (Model 6)
Availability of flexible leave 0.38* 0.20* 0.17* −0.10 0.20* −0.10
(0.16) (0.06) (0.07) (0.14) (0.08) (0.14)
POS 0.31** −0.35** 0.34** −0.35**
(0.05) (0.05) (0.07) (0.05)
Perceived flexibility stigma −0.06* 0.02** −0.04* 0.02**
(0,03) (0.01) (0.02) (0.01)
Use 0.10* −0.20*
(0.04) (0.09)
POS x use 0.11* −0.15*
(0.05) (0.07)
Perceived flexibility stigma x use −0.07 −0.16*
(0.11) (0.07)
Treatment effects −0.16 −0.16 −0.12* −0.16* −0.12* −0.16*
(0.11) (0.10) (0.05) (0.08) (0.05) (0.08)
Gender −0.21* 0.14 −0.06 −0.13 −0.04 −0.15
(0.10) (0.10) (0.10) 0.10 (0.11) 0.10
Dependents −0.12 0.45* −0.28 −0.07 −0.27 0.08
(0.10) (0.11) (0.14) (0.10) (0.14) (0.09)
R-squared 0.03** 0.04** 0.14** 0.04** 0.15** 0.26**

Note: N = 204; SEs in parentheses.

*p < .05; **p < .01; ***p < .001.

F I G U R E 2 A path model of the influence

Perceived
Organizational
of flexible leave on organizational
Support .34** commitment
.38*

.11*

Availability of Use of Flexible

Organizational Commitment
Flexible Leave Leave

-.07

.20*
-.04*
Perceived
Flexibility Stigma

-.17*
Note: *p<.05, **p<.01
Hypothesized effect
Non-Hypothesized effect

availability of flexible leave on organizational commitment. Overall, Fifth, the bootstrapping results for the moderated mediation
the bootstrapping results corroborate the results of the regression models (models 9 and 10) are presented in Table 4 (Panels A and B).
analysis for mediation effects. As our moderating variable is categorical, we examine its effect at two
WHITE ET AL. 11

F I G U R E 3 A path model of the influence

Perceived
of flexible leave on turnover intention Organizational
Support -.35**
.38*

-.15*

Availability of Use of Flexible

Turnover Intention
Flexible Leave Leave

-.16*

.20*
.02**
Perceived
Flexibility Stigma

-.10
Note: *p<.05, **p<.01
Hypothesized effect
Non-Hypothesized effect

TABLE 4 Analysis of direct and indirect effects

Panel A: Direct and indirect effects (no moderator)

Organizational commitment (Model 7) Turnover intention (Model 8)

Coefficient Bootstrap (95% CI) Coefficient (SD) Bootstrap (95% CI)
(SD)
Direct effect: Availability of flexible leave 0.17 (0.07) 0.01 to 0.34 −0.10 (0.14) −0.10 to 0.04
Indirect effect: perceived organizational support (POS) 0.31 (0.13) 0.11 to 0.61 −0.35 (0.11) −0.27 to −0.03
Indirect effect: Perceived flexibility stigma −0.03 (0.01) −0.11 to −0.06 0.13 (0.05) 0.03 to 0.26

Panel B: Direct and conditional indirect effects (with moderator)

Organizational commitment (Model 9) Turnover intention (Model 10)

Coefficient (SD) Bootstrap (95% CI) Coefficient (SD) Bootstrap (95% CI)
Indirect effect: POS 0.07 (0.02) −0.04 to 0.05 −0.02 (0.25) −0.01 to 0.10
Moderator: No use of flexible leave
Indirect effect: POS 0.20 (.10) 0.09 to 0.56 −0.17 (0.08) −0.35 to −0.03
Moderator: Use of flexible leave
Indirect effect: Perceived flexibility stigma 0.01 (0.02) −0.04 to 0.05 0.01 (0.15) −0.05 to 0.10
Moderator: No use of flexible leave
Indirect effect: Perceived flexibility stigma −0.02 (0.02) −0.06 to 0.01 −0.11 (.05) −0.25 to −0.18
Moderator: Use of flexible leave

Note: LLCI is lower level confidence interval; ULCI is upper limit confidence interval. The number of bootstrap samples for bias corrected bootstrap
confidence intervals: 5000.

values (non-use = 0; use = 1). For POS (see Table 4, Panel A) the effect of POS on the relationship between flexible leave and organiza-
results indicate that non-use of the flexible leave policy did not tional commitment and turnover intention, respectively. Figures 4 and
strengthen the effect of POS on organizational commitment or turn- 5 plot the moderating effects of use on our two mediators based on
over intention, with zero lying inside respective 95% confidence inter- our estimates of the path models presented in Figure 2.
vals. In contrast, the relationship between POS and organizational Finally, the bootstrapping results for perceived flexibility stigma
commitment (coefficient = 0.20, 95% CI: 0.09 to 0.56) and turnover presented in Table 4 (Panel B) indicate that non-use of flexible leave
intention (coefficient = −0.17, 95% CI: −0.35 to −0.03) were policy did not moderate the mediator of perceived flexibility stigma
strengthened by the use of flexible leave, as zero was not contained on organizational commitment nor the effect of perceived flexibility
within either of the 95% confidence intervals. The results support stigma on turnover intention, with zero lying inside of the respective
H3a and H3b that use of flexible leave moderates the mediating 95% confidence intervals. In addition, use of the flexible leave policy
12 WHITE ET AL.

F I G U R E 4 The moderating effect of use on the relationship F I G U R E 6 The moderating effect of use on the relationship
between perceived organizational support and organizational between perceived flexibility stigma and organizational commitment
commitment

Finally, we controlled for the effect of employee gender in our

models. While this was not the focus of our study, we found that
while gender had an effect on POS (coefficient = −0.21, p < .05), it
had no effect on our overall moderated mediation models.

5 | DI SCU SSION AND CO NCLUSIO NS

We make a contribution to HRM literature, through our empirical

example of flexible leave policy, to confirm that employees' experi-
ence of HR practices has significant impact upon their subsequent
behavior, in our empirical case, their organizational commitment and
turnover intention (Alfes et al., 2012; Baluch, 2017; Bos-Nehles &
Veenendaal, 2017; Bowen & Ostroff, 2004; Jiang et al., 2017; Pur-
cell & Hutchinson, 2007; Purcell & Kinnie, 2006). Specifically focusing
upon flexible leave, we highlight the relationship between the avail-
ability of flexible leave and organizational attachment is complex, but
a positive perception of employees about flexible leave that gives rise
to their increased organizational attachment is enhanced through
F I G U R E 5 The moderating effect of use on the relationship actual use of flexible leave.
between perceived organization support and turnover intention The motivation for focusing our study on flexible leave stems
from the equivocality of evidence, even with meta-analytic research,
associated with the effectiveness of flexible working arrangements
did not significantly moderate the mediator of perceived flexibility (Shockley & Allen, 2007); that is, it is a particularly interesting empiri-
stigma on organizational commitment, but did significantly moderate cal example of our broader research concern. To tackle the equivocal
the mediator of perceived flexibility stigma on turnover intention nature of findings to date, our study addressed three main concerns in
(coefficient = −0.11, 95% CI: −0.25 to −0.18), as zero was not con- extant literature. First, rather than treating flexible working arrange-
tained within the 95% confidence interval. Hence, our results do not ments as an umbrella concept, we focused specifically on the case of
support H4a but support H4b. Figures 6 and 7 plot the moderating flexible leave (Beauregard & Henry, 2009; Perry-Smith & Blum, 2000).
effects of use on our two mediators based on our estimates of the We did so because flexible leave is available to all employees, regard-
path models presented in Figure 3. less of family responsibilities and circumstances (Smithson & Stokoe,
WHITE ET AL. 13

effects. We begin by addressing the how question, through discussing

the importance of our mediators, POS and perceived flexibility stigma,
which were both found to be significant in shaping the relationship
between the availability of flexible leave and employees' organiza-
tional commitment.
For POS, we argue that the availability of flexible leave consti-
tutes a non-material social exchange, in the form of a signal from the
employer to the employee, that will shape employees' perceptions of
the employer's unobservable intentions and motives (Goldberg &
Allen, 2008). Independent of use, the availability of flexible leave
sends a signal of their commitment to, and concern for, employees'
(Grover & Crooker, 1995; Suazo, Martínez, & Sandoval, 2009). The
signal will enhance employees' POS, which will in turn translate
through into their enhanced organizational commitment. The availabil-
ity of flexible leave, therefore, should be viewed as an antecedent to
the development of high-quality social exchange relationships, which
lead employees to feel a greater sense of obligation to reciprocate
(Eisenberger et al., 1986; Shore & Wayne, 1993).
F I G U R E 7 The moderating effect of use on the relationship The inclusion of the mediator of perceived flexibility stigma
between perceived flexibility stigma and turnover intention (Cech & Blair-Loy, 2014) enabled us to embrace the potential dark
side of flexible working (for a review see Perrigino et al., 2018). Per-
ceived flexibility stigma is an important concept that has yet to be
2005), yet it remains the least understood of flexible working arrange- incorporated into scholarship on flexible working and employee out-
ments (Baltes et al., 1999; Kelly et al., 2008). Second, rather than comes, with the notable exception of Cech and Blair-Loy (2014) who
assuming that the effects of the availability of flexible leave are direct correlated perceived flexibility stigma with employee characteristics
in nature, we examined the question as to how flexible leave policies and outcomes. Our findings indicate that perceived flexibility stigma is
indirectly shape employees' experiences before leading to employee elevated by the availability of flexible leave, which in turn reduces
outcomes (Kelly et al., 2008). Drawing on SET (see Cropanzano & employee's organizational attachment. Hence, perceived flexibility
Mitchell, 2005 for a review) and signaling theory (Spence, 1973), we stigma is real and has a detrimental effect on the social exchange rela-
developed a model of the indirect effects of flexible leave in terms of tionship between an employer and employee.
POS and perceived flexibility stigma (T. D. Allen, 2001; Beauregard & To address the when question we employed our moderated medi-
Henry, 2009; Butts et al., 2013). Third, we developed a moderated ation models to delineate the effect of availability and use of flexible
mediation model in order to explain the differential effect of the avail- leave on organizational attachment (Kossek, Lautsch, & Eaton, 2006),
ability and use of the policy of flexible leave (T. D. Allen et al., 2013; and highlight the importance of our work with reference to Butts and
Butts et al., 2013; Kelly & Moen, 2007; Kossek, 2005; Kossek et al., co-authors (Butts et al., 2013, p. 2) who suggest that: “To our knowl-
1999). Differentiating between availability and use of flexible leave edge, no research has examined the processes through which policy
enables us to address the question as to when, that is, under what availability and use relate to employee work attitudes.” In all four
conditions does flexible leave matter. models we found the relationship between the mediator and the out-
Employing a quasi-experimental research design, and a moderated come variable was strengthened, but the effect was greatest for POS,
mediation model, we were able to address our three concerns men- as compared with perceived flexibility stigma.
tioned earlier. Specifically, we were able to incorporate multiple indi- For POS, our results indicate that the additional non-material and
rect effects, and delineate between availability and use of the policy material benefit users gain from the availability of the flexible leave,
over time, which is novel in the wider HRM literature (Harley, Allen, & as compared with non-users, will enhance their social exchange rela-
Sargent, 2007; Macky & Boxall, 2007, 2008; Mohr & Zoghi, 2008; tionship with their employer, such that the relationship between POS
Takeuchi, Chen, & Lepak, 2009). Due to recent methodological devel- and organizational attachment will be stronger for users as compared
opments in structural equation modeling (Kline, 2015), we were able with non-users. The moderator of use is statistically significant for
to simultaneously test multiple indirect effects (Preacher & Hayes, both organizational commitment and turnover intention.
2008), via a path model allowing for mediation and moderated media- For perceived flexibility stigma our findings were more mixed. We
tion models. We encourage other scholars of HRM to follow our suggest that use of flexible leave will enable employees may gain
methodological lead. more confidence of the employer's commitment to flexible leave, and
In summary, our findings demonstrate that the availability of flexi- so be more secure in the knowledge that they are not being judged or
ble leave enhances employees' organizational attachment, but that penalized by their employer. Accordingly, use of flexible leave will
the relationship is complex, being shaped by indirect and moderating reduce employees' fear of sending a negative signal to their employer,
14 WHITE ET AL.

and thus their perceived flexibility stigma, and so the relationship Finally, our results show the importance of differentiating
between availability of flexible leave and organizational attachment, between the availability and use of flexible leave, both in enhancing
will be weaker for users as compared with non-users. The mixed the mediating role of POS, and diminishing mediating role of per-
results stem from the fact that the effect of the moderator of use was ceived flexibility stigma, on employee outcomes. We suggest, there-
only found to be significant for turnover intention but not organiza- fore, that the continued and sustained availability of flexible leave is
tional commitment. important in fostering a supportive organizational climate, and is
We suggest, therefore, that studies of flexible working arrange- necessary for organizations to fully realize the benefits of flexible
ments need to include use, as well as availability, due to its discretion- working (Beauregard & Henry, 2009).
ary nature (Shockley & Allen, 2010; Veiga, Baldridge, & Eddleston,
2004). Furthermore, the effects of perceptions on employee out-
comes may be enhanced through use, which suggests that it is impor- 5.2 | Limitations and future research
tance to increase participation in such schemes if their true benefits
are to be felt by all employees. As with all studies our study does have limitations, which we hope
In summary, our moderated mediation model demonstrates the can provide directions for future research. First, our sample of respon-
importance of focusing on the material and non-material social dents was composed entirely of BBC employees, therefore, it is a mat-
exchanges, between employers and employees, to understand how ter of debate as to whether the findings obtained can be generalized
the availability of flexible leave shapes employees' organizational to other populations. The BBC is a public organization, incorporated
attachment. Rather than having a direct effect on employee out- by Royal Charter. We suggest future research replicate our study in
comes, we find that the indirect effects of our model dominate private organizations and across industrial sectors to see whether our
employee outcomes, and attest to the complex relationships that exist results hold beyond the BBC.
between the introduction of flexible leave and employee outcomes. Second, our study is based on questionnaire data only; therefore,
we could not confirm that common method bias is not a problem in
our data (Podsakoff et al., 2003). To mitigate against the problem of
5.1 | Managerial implications common methods bias, as outlined in Section 3, we did the following:
we selected different types of scales and formats for our constructs,
Our study shows the importance of understanding how flexible work- and used reverse-scored items where possible (Podsakoff & Organ,
ing arrangements are perceived by employees, if they are to have a 1986). In addition, we conducted a Harman's one-factor test. The
positive effect on outcomes. In essence, we suggest that flexible Harman test involved entering all of our variables into an exploratory
working arrangements may be best viewed as a necessary, but not factor analysis to determine the number of factors that account for
sufficient, condition for enhancing employee outcomes. The danger the variance in the variables, and then the variables were loaded on
for employers being that employees view such arrangements as one factor to examine the fit of the confirmatory factor analysis
tokenistic. Our mediation model demonstrates the importance of model. We found five factors emerge from exploratory factor analysis,
employers sending their employees a signal they are serious about and the one-factor confirmatory factor analysis model did not fit the
flexible working arrangements as a means of enhancing employees' data well, confirming that common method bias is unlikely to be a
POS, if they are to enhance organizational attachment. At the same problem with our data.
time, however, employees may have concerns about the availability of Our study addresses calls for more longitudinal approaches to
a policy flexible leave raising perceived flexibility stigma, and indirectly study flexible working arrangements (Grover & Crooker, 1995; Kelly
reducing organizational attachment. et al., 2008). For future research, while the estimation of longitudinal
We suggest employers can further enhance POS through the mediation models is complex, it is also rewarding, since only through
introduction of flexible leave by better signaling that flexible working such an analysis can we understand the process whereby flexible
are part of a raft of initiatives encompassed within broader HR strate- working arrangement effects are produced (MacKinnon, 2008).
gies, such as those for greater work–life balance (Lewis, Gambles, & Without knowledge of this process, generalizing flexible working
Rapoport, 2007) and/or a more diverse workforce (Atkinson & policy or intervention effects may be difficult (Butts et al., 2013;
Sandiford, 2016). Such strategies, and the place of flexible leave MacKinnon, 2008). However, the predominance of mediation
within them, should be clearly communicated as a signal of the value hypotheses for flexible working are still tested with cross-sectional
they place on employees. In addition, for organizations to fully harness data (Casper, Eby, Bordeaux, Lockwood, & Lambert, 2007). Further-
the positive effects of flexible leave they need to ensure that they more, consensus on the best approach to longitudinal mediation
address employees' concerns about perceived flexibility stigma. To modeling methods is largely lacking. In our study, we were limited in
minimize the perceived flexibility stigma employers need to demon- that our design only collected two waves of data. Some scholars
strate that employees will not suffer career penalties for using flexible advocate that three waves or more of data are necessary to truly
leave. We suggest such an approach will require senior managers test mediation (Collins, Graham, & Flaherty, 1998). There is now
ensure all line managers implement policy consistently throughout the a growth in interest in models through which three or more waves
organization (Galea, Houkes, & De Rijk, 2014). of data are collected. These include cross-lagged panel models
WHITE ET AL. 15

(MacKinnon, 2008), latent difference scores models (MacKinnon, Blair-Loy, M., Wharton, A. S., & Goodstein, J. (2011). Exploring the rela-
2008; McArdle, 2001), and latent growth curve models (Bollen & tionship between mission statements and work-life practices in organi-
zations. Organization Studies, 32(3), 427–450.
Curran, 2006). Longitudinal mediation modeling has much to offer
Blau, P. (1964). Exchange and power in social life. New York, NY: Wiley.
future research. Bollen, K. A., & Curran, P. J. (2006). Latent curve models: A structural equa-
tion perspective (Vol. 467). Hoboken, NJ: John Wiley & Sons.
ACKNOWLEDGMENTS Bollen, K. A., & Stine, R. (1990). Direct and indirect effects: Classical and
bootstrap estimates of variability. Sociological Methodology, 20,
The first author would like to thank the British Academy, United King-
115–140.
dom, for funding this study (Grant ref. SG120877).
Bos-Nehles, A., & Veenendaal, A. R. (2017). Percpetions of HR practices
and innovative work behaviour: The moderating effect of an innova-
ORCID tive climate. The International Journal of Human Resource Management,
Leroy White https://orcid.org/0000-0001-6827-1323 30(18), 2661–2683.
Bowen, D. E., & Ostroff, C. (2004). Understanding HRM-firm performance
Graeme Currie https://orcid.org/0000-0002-4825-9711
linkages: The role of the ‘strength’ of the HRM system. Academy of
Management Review, 29(2), 203–221.
ENDNOTE Butts, M. M., Casper, W. J., & Yang, T. S. (2013). How important are work–
1
The measure for perceived flexibility stigma relates to both formal and family support policies? A meta-analytic investigation of their effects
informal HR policies. In this study, our focus is on formal HR poli- on employee outcomes. Journal of Applied Psychology, 98(1), 1–25.
cies only. Byron, K. (2005). A meta-analytic review of work–family conflict and its
antecedents. Journal of Vocational Behavior, 67, 169–198.
Casper, W. J., Eby, L. T., Bordeaux, C., Lockwood, A., & Lambert, D. (2007).
RE FE R ENC E S A review of research methods in IO/OB work-family research. Journal
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and inter- of Applied Psychology, 92(1), 28–43.
preting interactions. Newbury Park, CA: SAGE. Casper, W. J., & Harris, C. M. (2008). Work-life benefits and organizational
Alfes, K., Bailey, C., Shantz, A., & Soane, E. (2012). The link between per- attachment: Self-interest utility and signaling theory models. Journal of
ceived human resource management practices, engagement and Vocational Behavior, 72(1), 95–109.
employee behaviour: A moderated mediation model. The International Cech, E. A., & Blair-Loy, M. (2014). Consequences of flexibility stigma
Journal of Human Resource Management, 24(2), 330–351. among academic scientists and engineers. Work and Occupations, 41
Allen, T. D. (2001). Family-supportive work environments: The role of (1), 86–110.
organizational perceptions. Journal of Vocational Behavior, 58, Chung, H. (2018). Women's work penalty’ in access to flexible working
414–435. arrangements across Europe. European Journal of Industrial Relations,
Allen, T. D., Johnson, R. C., Kiburz, K. M., & Shockley, K. M. (2013). Work– 52(2), 491–507.
family conflict and flexible work arrangements: Deconstructing flexibil- Collins, L. M., Graham, J. J., & Flaherty, B. P. (1998). An alternative frame-
ity. Personnel Psychology, 66(2), 345–376. work for defining mediation. Multivariate Behavioral Research, 33(2),
Allen, N. J., & Meyer, J. P. (1996). Affective, continuance, and normative 295–312.
commitment to the organization: An examination of construct validity. Cook, T. D., & Campbell, D. T. (1976). The design and conduct of quasi-
Journal of Vocational Behavior, 49(3), 252–276. experiments and true experiments in field settings. In M. D. Dunnette
Atkinson, C., & Sandiford, P. (2016). An exploration of older worker flexi- (Ed.), Handbook of industrial and organizational psychology
ble working arrangements in smaller firms. Human Resource Manage- (pp. 223–326). Rand McNally: Chicago.
ment Journal, 26(1), 12–28. Colarelli, S. M. (1984). Methods of communication and mediating pro-
Baltes, B. B., Briggs, T. E., Huff, J. W., Wright, J. A., & Neuman, G. A. cesses in realistic job previews. Journal of Applied Psychology, 69(4),
(1999). Flexible and compressed workweek schedules: A meta-analysis 633–642.
of their effects on work-related criteria. Journal of Applied Psychology, Cropanzano, R., & Mitchell, M. S. (2005). Social exchange theory: An inter-
84(4), 496–513. disciplinary review. Journal of Management, 31(6), 874–900.
Baluch, A. M. (2017). Employee perceptions of HRM and well-being in Edwards, J. R., & Lambert, L. S. (2007). Methods for integrating moderation
non-profit organizations: Unpacking the unintended. The International and mediation: A general analytical framework using moderated path
Journal of Human Resource Management, 28(14), 1912–1937. analysis. Psychological Methods, 12(1), 1–22.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable dis- Efron, B., & Tibshirani, R. J. (1993). An introduction to the bootstrap.
tinction in social psychological research: Conceptual, strategic, and sta- New York, NY: Chapman and Hall.
tistical considerations. Journal of Personality and Social Psychology, 51, Eisenberger, R., Huntington, R., Hutchison, S., & Sowa, D. (1986). Per-
1173–1182. ceived organizational support. Journal of Applied Psychology, 71(3),
Beauregard, T. A., & Henry, L. C. (2009). Making the link between work-life 500–507.
practices and organizational performance. Human Resource Manage- Emerson, R. M. (1976). Social exchange theory. Annual Review of Sociology,
ment Review, 19(1), 9–22. 2, 335–362.
Becker, T. E. (2005). Potential problems in the statistical control of vari- Epstein, C. F., Seron, C., Oglensky, B., & Saute, R. (1999). The part-time par-
ables in organizational research: A qualitative analysis with recommen- adox. New York, NY: Routledge.
dations. Organizational Research Methods, 8, 274–289. Gajendran, R. S., & Harrison, D. A. (2007). The good, the bad, and the
Berdahl, J. L., & Moon, S. H. (2013). Workplace mistreatment of middle unknown about telecommuting: Meta-analysis of psychological media-
class workers based on sex, parenthood, and caregiving. Journal of tors and individual consequences. Journal of Applied Psychology, 92(6),
Social Issues, 69(2), 341–366. 1524–1541.
Blair-Loy, M. (2003). Competing devotions: Career and family among women Galea, C., Houkes, I., & De Rijk, A. (2014). An insider's point of view: How
executives. Cambridge, MA: Harvard University Press. a system of flexible working hours helps employees to strike a proper
Blair-Loy, M., & Wharton, A. S. (2002). Employees' use of work-family poli- balance between work and personal life. The International Journal of
cies and the workplace social context. Social Forces, 80(3), 813–845. Human Resource Management, 25(8), 1090–1111.
16 WHITE ET AL.

Goldberg, C. B., & Allen, D. G. (2008). Black and white and read all over: Lewis, S., Gambles, R., & Rapoport, R. (2007). The constraints of a ‘work–
Race differences in reactions to recruitment web sites. Human life balance’ approach: An international perspective. The International
Resource Management, 47(2), 217–236. Journal of Human Resource Management, 18(3), 360–373.
Goodman, J. S., & Blum, T. C. (1996). Assessing the non-random sampling Lind, E. A., & Tyler, T. R. (1988). The social psychology of procedural justice.
effects of subject attrition in longitudinal research. Journal of Manage- New York, NY: Plenum.
ment, 22(4), 627–652. MacKinnon, D. P. (2000). Contrasts in multiple mediator models. In
Grover, S. L., & Crooker, K. J. (1995). Who appreciates family-responsive J. Rose, L. Chassin, C. C. Presson, & S. J. Sherman (Eds.), Multivariate
human resource policies: The impact of family-friendly policies on the applications in substance use research (pp. 141–160). Mahwah, NJ:
organizational attachment of parents and non-parents. Personnel Psy- Erlbaum.
chology, 48(2), 271–288. MacKinnon, D. P. (2008). Multivariate applications series. Introduction to
Harley, B., Allen, B. C., & Sargent, L. D. (2007). High performance work statistical mediation analysis. New York, NY: Erlbaum and Taylor
systems and employee experience of work in the service sector: The Francis Group.
case of aged care. British Journal of Industrial Relations, 45(3), 607–633. MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional limits for the indirect effect: Distribution of the product and
process analysis. New York, NY: Guilford Press. resampling methods. Multivariate Behavioral Research, 39(1),
Homans, G. C. (1958). Social behavior as exchange. American Journal of 99–128.
Sociology, 63(6), 597–606. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., &
Jiang, K., Hu, J., Liu, S., & Lepak, D. P. (2017). Understanding employees' Sheets, V. (2002). A comparison of methods to test mediation and
percpetions of human resource practices: Effects of demographic dis- other intervening variable effects. Psychological Methods, 7(1),
similarity in managers and coworkers. Human Resource Management, 83–104.
56(1), 69–91. Macky, K., & Boxall, P. (2007). The relationship between ‘high performance
Judd, C. M., McClelland, G. H., & Culhane, S. E. (1995). Data analysis: Con- work practices’ and employee attitudes: An investigation of additive
tinuing issues in the everyday analysis of psychological data. Annual and interaction effects. International Journal of Human Resource Man-
Review of Psychology, 46(1), 433–465. agement, 18, 537–567.
Kelly, E. L., & Moen, P. (2007). Rethinking the clockwork of work: Why Macky, K., & Boxall, P. (2008). High-involvement work processes, work
schedule control may pay off at work and at home. Advances in Devel- intensification and employee well-being: A study of New Zealand
oping Human Resources, 9, 487–506. worker experiences. Asia Pacific Journal of Human Resources, 46(1),
Kelly, E., Kossek, E., Hammer, L., Durham, M., Bray, J., Chermack, K., … 38–55.
Kaskubar, D. (2008). Getting there from here: Research on the effects McArdle, J. J. (2001). A latent difference score approach to longitudinal
of work-family initiatives on work-family conflict and business out- dynamic structural analysis. In R. Cudeck, S. du Toit, & D. Sorbom
comes. The Academy of Management Annals, 2, 305–349. (Eds.), Structural equation modeling: Present and future (pp. 342–380).
Kline, R. B. (2015). Principles and practice of structural equation modeling. Lincolnwood, IL: Scientific Software International.
New York, NY: Guilford Publications. McCloskey, D. W., & Igbaria, M. (2003). Does “out of sight” mean “out of
Kossek, E. E. (2005). Workplace policies and practices to support mind”? An empirical investigation of the career advancement pros-
work and families. In S. Bianchi, L. Casper, & R. King (Eds.), Work, pects of telecommuters. Information Resources Management Journal, 16
family, health, and well-being (pp. 97–116). Hillsdale, NJ: Erlbaum (2), 19–34.
Press. Mesmer-Magnus, J. R., & Viswesvaran, C. (2006). How family-friendly
Kossek, E. E., Barber, A. E., & Winters, D. (1999). Using flexible schedules work environments affect work/family conflict: A meta-analytic exam-
in the managerial world: The power of peers. Human Resource Manage- ination. Journal of Labor Research, 27(4), 555–574.
ment, 38, 33–46. Meyer, J. P., & Allen, N. J. (1984). Testing the" side-bet theory" of organi-
Kossek, E. E., Lautsch, B. A., & Eaton, S. C. (2006). Telecommuting, control, zational commitment: Some methodological considerations. Journal of
and boundary management: Correlates of policy use and practice, job Applied Psychology, 69(3), 372–378.
control, and work–family effectiveness. Journal of Vocational Behavior, Meyer, C., Mukerjee, S., & Sestero, A. (2001). Work-family benefits: Which
68(2), 347–367. ones maximise profits? Journal of Managerial Issues, 13, 28–44.
Kossek, E. E., & Nichol, V. (1992). The effects of on-site child care on Moen, P., Kelly, E., & Chermack, K. (2009). Learning from a natural experi-
employee attitudes and performance. Personnel Psychology, 45(3), ment: Studying a corporate work-time policy initiative. In
485–509. A. C. Crouter & A. Booth (Eds.), Work-life policies that make a real
Kossek, E. E., Sweet, S., & Pitt-Catsouphes, M. (2006). The insights gained difference for individuals, families, and organizations (pp. 97–131).
from integrating disciplines. In M. Pitt-Catsouphes, E. E. Kossek, & Washington, DC: Urban Institute.
S. Sweet (Eds.), The handbook of work–family: Multi-disciplinary perspectives, Mohr, R. D., & Zoghi, C. (2008). High-involvement work design and job
methods, and approaches (pp. 67–72). Mahwah, NJ: Lawrence Erlbaum. satisfaction. ILR Review, 61(3), 275–296.
Kurtessis, J. N., Eisenberger, R., Ford, M. T., Buffardi, L. C., Molm, L. D. (2000). Theories of social exchange and exchange networks.
Stewart, K. A., & Adis, C. S. (2017). Perceived organizational support: In G. Ritzer & B. Smart (Eds.), Handbook of social theory (pp. 260–272).
A meta-analytic evaluation of organizational support theory. Journal of Thousand Oaks, CA: SAGE.
Management, 43(6), 1854–1884. Molm, L. D. (2003). Theoretical comparisons of forms of exchange. Socio-
Labianca, G., & Brass, D. J. (2006). Exploring the social ledger: Negative logical Theory, 21, 1–17.
relationships and negative asymmetry in social networks in organiza- Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is medi-
tions. Academy of Management Review, 31(3), 596–614. ated and mediation is moderated. Journal of Personality and Social Psy-
Lambert, S. J. (2000). Added benefits: The link between work-life benefits chology, 89(6), 852–863.
and organizational citizenship behavior. Academy of Management Jour- Parker, L., & Allen, T. D. (2001). Work/family benefits: Variables related to
nal, 43(5), 801–815. employees' fairness perceptions. Journal of Vocational Behavior, 58(3),
Lee, T. W., & Mitchell, T. R. (1994). Organizational attachment: Atti- 453–468.
tudes and actions. In J. Greenberg (Ed.), Organizational behavior: Perrigino, M. B., Dunford, B. B., & Wilson, K. S. (2018). Work–family back-
The state of the science (pp. 83–108). Hillsdale, NJ: Lawrence lash: The “dark side” of work–life balance (WLB) policies. Academy of
Erlbaum Associates. Management Annals, 12(2), 600–630.
WHITE ET AL. 17

Perry-Smith, J. E., & Blum, T. C. (2000). Work-family human resource bun- Takeuchi, R., Chen, G., & Lepak, D. P. (2009). Through the looking glass of
dles and perceived organizational performance. Academy of Manage- a social system: Cross-level effects of high-performance work systems
ment Journal, 43(6), 1107–1117. on employees' attitudes. Personnel Psychology, 62(1), 1–29.
Podsakoff, P. M., & Organ, D. W. (1986). Self reports in organizational Thompson, R. J., Payne, S. C., & Taylor, A. B. (2015). Applicant attraction
research: Problems and prospects. Journal of Management, 12, 531–544. to flexible work arrangements: Separating the influence of flextime
Podsakoff, P. M., MacKenzie, S. B., Podsakoff, N. P., & Lee, J.-Y. (2003). and flexplace. Journal of Occupational and Organizational Psychology,
Common method biases in behavioral research: A critical review of the 88(4), 726–749.
literature and recommended remedies. Journal of Applied Psychology, Tipping, S., Chanfreau, J., Perry, J., & Tait, C. (2012). The fourth
88(5), 879–903. work-life balance employee survey. Employment Relations Research
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for esti- Series 122. London, England: Department for Business and
mating indirect effects in simple mediation models. Behavior Research Innovation.
Methods, Instruments, & Computers, 36(4), 717–731. Veiga, J. F., Baldridge, D. C., & Eddleston, K. A. (2004). Toward under-
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strate- standing employee reluctance to participate in family-friendly pro-
gies for assessing and comparing indirect effects in multiple-mediator grams. Human Resource Management Review, 14(3), 337–351.
models. Behavior Research Methods, 40, 879–891. Williams, J. C., Blair-Loy, M., & Berdahl, J. L. (2013). Cultural schemas,
Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated social class, and the flexibility stigma. Journal of Social Issues, 69(2),
mediation hypotheses: Theory, methods, and prescription. Multivariate 209–234.
Behavioral Research, 42, 185–227.
Preacher, K. J., & Selig, J. P. (2012). Advantages of Monte Carlo confidence inter-
vals for indirect effects. Communication Methods and Measures, 6(2), 77–98. AUTHOR BIOGRAPHIES
Purcell, J., & Hutchinson, S. (2007). Front-line managers as agents in the
HRM-performance causal chain: Theory, analysis and evidence. Human
Resource Management Journal, 17(1), 3–20. Leroy White is a Professor in Operational Research at Warwick
Purcell, J., & Kinnie, N. (2006). HRM and business performance. In Business School. His research interests are in social networks in
P. Boxall, J. Purcell, & P. Wright (Eds.), The Oxford handbook of human organizations, group behavior and decision-making, leadership,
resource management. Oxford, England: Oxford University Press.
and partnership working spanning a range of private, public, and
Rhoades, L., & Eisenberger, R. (2002). Perceived organizational support: A
review of the literature. Journal of Applied Psychology, 87(4), 698–714. not-for-profit organizations, including Barclays Bank and the Brit-
Ruth-Eikhof, D., & Warhurst, C. (2013). The promised land? Why social ish Broadcasting Corporation.
inequalities are systemic in the creative industries. Employee Relations,
35(5), 495–508. Andy Lockett is a Professor of Strategy and Entrepreneurship and
Schmidt, F. L., & Hunter, J. (2004). General mental ability in the world of the Dean at Warwick Business School. His research interests span
work: Occupational attainment and job performance. Journal of Person- the interface of strategy and entrepreneurship focusing on new
ality and Social Psychology, 86(1), 162–173.
and established ventures in both the private and public sectors.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and
quasi-experimental designs. Boston, MA: Houghton Mifflin Co.
Graeme Currie is a Professor of Public Management at Warwick
Shockley, K. M., & Allen, T. D. (2007). When flexibility helps: Another look
Business School. He studies human resource management, leader-
at the availability of flexible work arrangements and work–family con-
flict. Journal of Vocational Behavior, 71(3), 479–493. ship, innovation, and strategy in public services organizations. He
Shockley, K. M., & Allen, T. D. (2010). Investigating the missing link in flexi- currently leads two large-scale programs of research focused
ble work arrangement utilization: An individual difference perspective. upon the implementation of evidence-based innovation in health
Journal of Vocational Behavior, 76, 131–142.
and social care that are funded by the National Institute for
Shore, L. M., & Wayne, S. J. (1993). Commitment and employee behavior:
Comparison of affective commitment and continuance commitment Health Research and the Economic and Social Research Council in
with perceived organizational support. Journal of Applied Psychology, the UK. Besides publishing in Human Resource Management,
78(5), 774–780. Graeme's work has featured in the Academy of Management Jour-
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and non-
nal, The Leadership Quarterly, and Organization Studies. In a former
experimental studies: New procedures and recommendations. Psycho-
logical Methods, 7(4), 422–445. life, prior to becoming an Academic, Graeme worked in organiza-
Smithson, J., & Stokoe, E. H. (2005). Discourses of work–life balance: tion development for the NHS and a car manufacturer.
Negotiating ‘genderblind’ terms in organizations. Gender, Work and
Organization, 12(2), 147–168.
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in
structural equation models. Sociological Methodology, 13, 290–312.
Spence, M. (1973). Job market signaling. Quarterly Journal of Economics, How to cite this article: White L, Lockett A, Currie G. How
87, 355–374. does the availability and use of flexible leave influence the
Suazo, M. M., Martínez, P. G., & Sandoval, R. (2009). Creating psychologi-
employer–employee relationship? Hum Resour Manage. 2019;
cal and legal contracts through human resource practices: A signaling
theory perspective. Human Resource Management Review, 19(2), 1–17. https://doi.org/10.1002/hrm.22004
154–166.