The association of work-related extended availability with recuperation,well-being,life domain balance and work: A meta-analysis

WREA in the context of boundary theory

The lives of individuals take place within several domains. To get a grasp of what WREA means, how it can be conceptually distinguished from institutionalized forms of flexible working arrangements and in which way WREA may impact employee outcomes, we first delve into boundary theory. Boundary theory (Nippert-Eng, 1996) proposes that different life domains are separated by temporal (e.g., working hours), spatial (e.g., place of work) and psychological (e.g., social roles) boundaries (Hall & Richter, 1988). Their respective strength is determined by their flexibility and permeability (Clark, 2000). When both flexibility and permeability are high, boundaries between domains are weak. Although flexibility and permeability are related, they are conceptually distinct (Matthews & Barnes- Farrell, 2010). Flexibility refers to the degree that physical time and location markers (i.e., working hours and place of work) can be adapted (Hall & Richter, 1988). Permeability, by contrast, relates to being located in one domain, but being psychologically or behaviorally involved in another (Ashforth et al., 2000).

Employee-oriented forms of flexible working arrangements such as contractually regulated telecommuting are manifestations of the flexibility aspect of boundaries (Clark, 2002). When telecommuting, for instance, the boundaries between work and private life may expand and contract (e.g., longer or shorter work hours), but they can be upheld. Contractually regulated telecommuting may mean that the spatial and possibly the temporal boundaries between work and private life have shifted away from what the majority of workers may consider the default, but it does not necessarily imply a permeation of these boundaries. Conversely, accepting work-related contacts and/or performing other work-related tasks during leisure time (i.e., WREA), by definition constitutes a permeation of the boundaries between work and private life, because leisure time is part of the private life domain. Individuals performing work-related tasks during leisure time are located in the private life domain, but they are at the same time behaviorally involved in the work domain. Although telecommuting may contribute to increasing boundary permeability and WREA is restricted when spatial and temporal boundaries are inflexible, it is conceptually useful to distinguish between flexible working arrangements and WREA. This becomes even more apparent when considering the outcomes of employee-oriented forms of flexibility and WREA. While there is meta-analytic evidence for a positive association of telecommuting with employee well-being and life-domain balance (Gajendran & Harrison, 2007), the results of the scoping review (Pangert et al., 2016) suggest that WREA is mostly associated with adverse employee outcomes.

The constructs under the umbrella term of WREA can be subdivided into availability behaviors (i.e., actually performing work-related tasks during leisure time) and perceived availability demands (i.e., feeling that one should perform work-related tasks during leisure time). While it can be argued that the autonomy associated with availability behaviors could under certain circumstances constitute a resource, which may buffer possible negative effects of performing work-related tasks during leisure time (Ďuranová & Ohly, 2016), perceived availability demands imply a restriction of autonomy and should therefore amplify negative effects of performing work-related tasks during leisure time. However, in fact we expect a considerable overlap between availability behaviors and availability demands, because when availability behaviors are the target construct, oftentimes individuals will perform availability behaviors because they feel they have to. Then again, when availability demands are measured, it is likely that individuals who perceive demands to perform work-related tasks during leisure time will eventually act upon these perceived demands. Thus both availability behaviors and demands entail a higher degree of permeability of the home boundary, and we argue their associations with employee criteria will be similar. This view is supported by the relatively high intercorrelations between availability behaviors and demands when both were measured in the same study (e.g., Barber & Santuzzi, 2015; Mellner, 2016; Piszczek, 2017; Reinke & Chamorro-Premuzic, 2014). In the light of the arguments presented above, we therefore expect that the adverse relationships between availability demands and employee outcomes may in some cases be somewhat stronger compared to the relationships between availability behaviors and employee outcomes, but not as much as theoretical deliberations would suggest.

Boundary theory can help explaining how WREA affects employee's lives: Weak boundaries between life domains are associated with blurring between roles and cross-role interruptions (Ashforth et al., 2000). This can facilitate spillover from work into the private domain, create conflict between work and nonwork roles, inhibit recovery processes, and—eventually—lead to decreases in long-term well-being and health (e.g., Butts et al., 2015; Ďuranová & Ohly, 2016; Thörel et al., 2020b). A key assumption in this regard is that work-related cues such as role-specific persons or objects when encountered in the non-work domain trigger psychological transitions from the currently enacted role (e.g., husband) to the work role and that these transitions are associated with the emergence of work-related thoughts and emotions (Ashforth et al., 2000). Moreover, boundary theory proposes that role transitions tend to occur faster and more automatic, the more often a role transition has been repeated (Ashforth et al., 2000), so that for individuals who perceive strong availability demands or who often perform availability behaviors external triggers may not even be necessary for a role transition to occur. For instance, individuals who often perform availability behaviors may develop behavioral patterns such as routinely checking work e-mail during leisure time (Menz et al., 2016). WREA may, thus, directly impact employees’ short-term well-being (e.g., affect) and impede recuperation, which in turn may lead to decreases in long-term well-being.

WREA in the context of the effort-recovery and stressor-detachment model

The effort-recovery model (ERM; Meijman & Mulder, 1998) and the stressor-detachment model (SDM; Sonnentag & Fritz, 2015) both emphasize the importance of recovery from work. The basic assumption is that work effort is always associated with a certain degree of psychophysiological activation, which is characterized by the secretion of hormones and changes in physiological parameters (Sonnentag & Fritz, 2006). According to the ERM this state of activation is temporary and not harmful per se. Under normal circumstances, the psycho-biological system will return to baseline levels once the exposure to work demands ceases (Meijman & Mulder, 1998). However, WREA implies that employees are confronted with work matters in their private lives, and psychological transitions to the work role may occur, which curtails windows for uninterrupted recovery. When there is no or insufficient recovery, the ERM predicts that the physiological activation may become chronic. In this case, the resulting load effects may, at least in the long run, cause adverse structural changes in physiological and psychological functions and will ultimately lead to health problems (Meijman & Mulder, 1998). The SDM identifies mental distancing from work (i.e., psychological detachment) as central in the recovery process and considers it a mediating mechanism in the relationship between job stressors and strain-related outcomes (Sonnentag & Fritz, 2015). The role of WREA can be conceptualized as both a job stressor itself (e.g., when individuals feel pressured to be available) and as a mechanism that can mentally activate or sustain existing job stressors (e.g., a work call may trigger thoughts about unfinished work; Ayres et al., 2021). We believe that being confronted with work matters during leisure time impedes mental distancing from work. Following the SDM, WREA should therefore contribute in sustaining physiological arousal or in reactivating arousal after a phase of incomplete recovery. The recovery literature has found that failure to detach from work is related to impaired sleep, stress, emotional exhaustion and burnout (Sonnentag & Fritz, 2015; Wendsche & Lohmann-Haislah, 2017). Furthermore, Sonnentag (2011) suggested a provisional path in the SDM of how work stressors such as WREA might impinge on productivity and other work-related criteria: The effect of these stressors on work-related outcomes could be mediated through both psychological detachment and strain-related outcomes (Sonnentag, 2011). Accordingly, WREA might adversely impact work-related criteria due to undermining employee recovery, well-being and life-domain balance.

WREA in the context of the job demands-resource-model

The Job Demands-Resource Model (JD-R; Bakker & Demerouti, 2007) categorizes work properties into job demands (i.e., properties that require effort and involve costs for employees) and job resources (i.e., properties that alleviate stress and may foster work performance). We mainly consider WREA as a job demand as it blurs boundaries between work and private life, is likely to interfere with recovery and may have negative effects on well-being (Ďuranová & Ohly, 2016; Pangert et al., 2016). However, we acknowledge that WREA may act as a resource in some instances. For example, taking care of work-related matters during leisure time can help preventing work overload in certain situations (Ďuranová & Ohly, 2016) and doing so could be considered a deliberate strategy to stay on top of things (Menz et al., 2016). Moreover, access to work-related information during leisure time may prove helpful for solving work problems and increasing efficiency (e.g., Day et al., 2010). Looking at WREA this way, it is possible that WREA may have desirable effects on work-related outcomes. For instance, when taking care of work-related matters during leisure time helps being successful on the job, it is plausible that WREA be positively related to work-related attitudes such as work satisfaction, commitment or work engagement and be negatively related to absenteeism and turnover. We think that WREA may at the same time be detrimental to some aspects of employee's lives while being beneficial in other regards, especially concerning work-related criteria. In the following, we will successively introduce the criteria and present the hypotheses.

WREA and recuperation

Recovery from work is a process of psychophysiological unwinding that is vital in maintaining long-term health (Geurts & Sonnentag, 2006). Psychological detachment is considered a core feature of recovery (Sonnentag & Fritz, 2015) that goes beyond the mere physical absence from work and implies a mental disengagement from work (Sonnentag & Fritz, 2007). Detachment is positively associated with desirable employee outcomes and negatively related to adverse outcomes (Wendsche & Lohmann-Haislah, 2017).

Consistent with the tenets of boundary theory, we assume that accepting work-related calls or taking care of work-related tasks during leisure time triggers role transitions and brings work-related thoughts into the private domain. Employees who are used to taking care of work-related tasks during leisure time are likely to switch into the work role quicker, and transitions may occur even at times when employees are not taking care of work-related tasks. Therefore, we assume that there is a negative relationship between WREA and detachment (H1a).

Another recovery experience is relaxation which is characterized by low activation and positive affect (Sonnentag & Fritz, 2007). Relaxation is related to decreases in physiological stress indicators and health complaints such as somatic or psychological symptoms (Van der Klink et al., 2001). The experience of relaxation is often associated with leisure activities or may result from the usage of certain techniques designed to achieve a state of relaxation (Sonnentag & Fritz, 2007). It is likely that performing work-related tasks during leisure time impedes relaxation. WREA implies that individuals are longer exposed to work demands, which keeps up psychophysiological activation. Moreover, WREA may also mean that employees spend less time for carrying out leisure activities needed to initiate relaxation. In an on-call study, Bamberg et al. (2012) found that social and low effort activities were reduced when employees were on call. Also, role transitions from the private to the work domain may largely undo any prior relaxation by eliciting endocrine responses characterized by high physiological activation (Sonnentag & Fritz, 2006). Thus, we hypothesize that there is a negative association between WREA and relaxation (H1b). Due to the assumed relationship between WREA and both detachment and relaxation, we also posit that WREA is negatively associated with employee's overall state of recovery (H1c).

Sleep is a basic homeostatic process with restorative effects on the body and the mind (Saper et al., 2005). In the occupational context, sleep is commonly conceptualized as either sleep quality or sleep quantity (Litwiller et al., 2017). Sleep quality is operationalized in terms of sleep problems, such as difficulties falling or staying asleep, and a lack of restorative effects (e.g. Bowen et al., 2018), whereas sleep quantity is measured as the amount of sleep (e.g., Lanaj et al., 2014). Sleep quality and sleep quantity are distinct constructs that are only weakly correlated (Litwiller et al., 2017). Based on the tenets of the SDM, we argue that when not detaching from work during leisure time, job stressors remain mentally present and elicit physiological responses resulting in impaired sleep (Sonnentag & Fritz, 2015). Furthermore, work-related tasks during leisure time are often carried out via ICT. Certain behaviors associated with information and communication technology use in the evening such as postponing sleep and leaving the smartphone switched on lead to higher levels of physiological arousal which may reduce the quantity and the quality of sleep (Thomeé, 2018). Furthermore, there is evidence that the blue light emitted from screens of smartphones and tablets curbs the production of sleep hormone (e.g., Wood et al., 2013). Therefore, we predict that WREA is negatively associated with sleep quality and quantity (H1d and H1e).

WREA and well-being

In the following, we will classify affective responses and depletion as indicators of short-term, psychological distress and perceived stress as indicators of mid-term, and emotional exhaustion and burnout as indicators of long-term well-being. As affective responses are concerned, work-related events generate affective experiences, which can determine both attitudes and behaviors in the work domain (Weiss & Cropanzano, 1996). It is known that affective responses such as moods and emotions can spill over from work into private life (Eby et al., 2010). WREA increases the permeability between life domains and role transitions bring along work-related thoughts and emotions (Ashforth et al., 2000) making spillover more likely, because individuals are confronted with work matters while they are in the private domain (Butts et al., 2015). Affect is a concept that includes both emotions and moods (Niven, 2013), and affect is divided into positive and negative affect (Göritz, 2007). Positive affect refers to a state of energy, concentration and pleasurable engagement, whereas negative affect includes a number of aversive psychological states such as anger, fear and nervousness (Watson et al., 1988). Psychological distress is a construct related to negative affect, but it can last longer. It can be defined as a set of adverse psychological symptoms (e.g., depressed mood) that are usually still within the bounds of normal mood fluctuations, but may also indicate the onset of a mental illness (American Psychological Association, 2015). There are at least two mechanisms that connect WREA to affective responses. First, getting in contact with work-related matters during leisure time can be considered an affective event in and of itself and therefore create a direct affective response (Weiss & Cropanzano, 1996). This initial response could consist of positive (e.g., happiness) or negative (e.g., anger) emotions (Butts et al., 2015), which may depend on attitudes, beliefs or needs an individual has in a given situation. Second, WREA weakens the boundaries between life domains, such that negative as well as positive emotions originating in the work domain may spill over and impact an individual's affect. It is plausible that emotional reactions and spillover may have a more than momentary impact on employees and cause psychological distress in the midterm, for example, when work contact triggers work-related fears. Therefore, we predict that WREA is positively related to both negative (H2a) and positive affect (H2b) as well as to psychological distress (H2c).

Moreover, performing work brings about self-control demands such as dealing with conflicting impulses, overcoming inner resistance or withstanding distraction (Schmidt & Neubach, 2010). Exerting self-control is thought to draw from a limited regulatory resource that can be depleted (Muraven & Baumeister, 2000; Singh & Göritz, 2019). Depletion is characterized by low willpower and cognitive exhaustion (Gombert et al., 2018a) and is related to decrements in later task performance (Diestel & Schmidt, 2012). We propose that WREA is positively associated with depletion (H2d) as self-control demands related to work are extended into leisure time, and the need to flexibly adapt to different roles may further deplete resources.

We believe that WREA may have further negative consequences for well-being. Perceived stress is defined in terms of the degree individuals appraise their lives as stressful (Cohen et al., 1983). Burnout is an adverse psychological state characterized by emotional exhaustion, cynicism and perceived lack of professional efficacy (Maslach & Jackson, 1981). Emotional exhaustion is often considered the core burnout dimension and the one that has been investigated most (Reichl et al., 2014): It can be conceptualized as a state of depleted energy assumed to be induced by excessive and long-term psychological and emotional demands on the job (Jackson et al., 1987).

As mentioned above, WREA can be conceptualized as both a job stressor itself and as a mechanism that can mentally activate existing job stressors (e.g., thoughts about unfinished work tasks). We argued that being confronted with work matters during leisure time inhibits detachment. Following the SDM, this should sustain physiological arousal or even activate it after a phase of recovery. The recovery literature suggests that failure to detach from work is related to stress, burnout and emotional exhaustion (Sonnentag & Fritz, 2015; Wendsche & Lohmann-Haislah, 2017). Therefore, we assume that WREA is positively associated with stress (H2e), emotional exhaustion (H2f) and burnout (H2g).

WREA and private life

From the perspective of boundary theory, desirable as well as undesirable effects of WREA on private life are conceivable. We will first consider possible undesirable effects. Ashforth et al. (2000) point out that flexible and permeable boundaries may kindle role conflict. Work-to-Family Conflict (WFC) refers to a form of interrole conflict in which role demands from the work domain complicate the fulfillment of role expectations in the family domain (Greenhaus & Beutell, 1985). There are three forms of WFC: Time-based WFC (TB-WFC) denotes a negative influence of activities performed in the work role on the time that is available for family activities, strain-based WFC (SB-WFC) refers to strain caused by work demands reducing participation in the family role and behavior-based WFC (BB-WFC) designates the impact of certain behavioral patterns inherent in the work role on the family role (Greenhaus & Beutell, 1985).

We expect that WREA is positively related to overall WFC (H3a) and all three forms of WFC (TB: H3b, SB: H3c and BB: H3d). Taking care of work-related matters during leisure time means that activities associated with the work role consume time that is no longer available for fulfilling tasks and obligations in the private domain. Likewise, WREA may also facilitate the transmission of work strain into private life when there is no buffer between roles. Negative emotions and states caused by work demands may pass the boundary between work and private life more easily if the work domain permeates the private domain. When this work strain is transmitted into private life, the experiences made while in the work role may complicate taking care of duties associated with the family role. Furthermore, WREA also makes it more likely that certain behavioral patterns associated with the work-role may be transferred into the private domain. For instance, when handling a work call while spending time with the family, for a manager it may be difficult to switch cognitive gears after the call and let go of a “cognitive, directive management style” and employ a “more emotional, cooperative one” (Ashforth et al., 2000). Finally, as a result of the extensive WFC we predict, we expect WREA to be negatively associated with self-rated work-life balance (H3e).

When individuals are not fully capable of fulfilling expectations associated with their roles in private life, this may lead to difficulties in functioning in these roles. Family role performance is defined as the fulfilment of duties and expectations associated with roles within the family domain (Chen et al., 2014). Due to the extensive WFC we expect, we also predict WREA to be negatively associated with family role performance (H3f). Furthermore, WREA may enforce interpersonal conflicts in spousal relationships, because even externally imposed penetrations of work into private life (e.g., answering calls) are often interpreted as a matter of choice by the spouse (Hall & Richter, 1988). Due to the expected negative impact of WREA on role performance in private life and relationship conflicts, we predict that WREA will be related to lesser satisfaction with private life (H3f).

As to potentially desirable effects of WREA on private life, boundary theory leaves open the possibility that with a higher degree of permeability, ideas, insights, skills or positive emotions originating in the work domain pass the boundary into private life (Clark, 2000). In the work-family literature, this concept is referred to as positive spillover (Edwards & Rothbard, 2000) or enrichment (Greenhaus & Powell, 2006). Greenhaus and Powell (2006) differentiated between an affective and an instrumental path of how work can positively impact private life. We suggest that WREA facilitates both the transfer of positive emotional states as well as beneficial skills and behaviors into the private domain and that WREA will be positively associated with positive overall (H3h) as well as with positive affective (H3i) and instrumental (H3j) spillover.

WREA and work-related criteria

As we have argued above, there are two possible mechanisms of how WREA may impact work-related criteria, such as job satisfaction, affective commitment, job involvement, work engagement, absence from work, or turnover intentions. On the one hand, the SDM suggests that the effect of WREA on work-related outcomes could be mediated through both psychological detachment and strain (Sonnentag, 2011). We thus would expect WREA to have an adverse effect on work-related criteria, due to its assumed negative effects on employee recuperation, well-being and private life. On the other hand, in line with the JD-R, availability behaviors may sometimes act as a resource that helps reaching work-related goals. WREA may thus be favorably related to desirable work-related attitudes and negatively to undesirable behaviors such as absence from work or turnover intentions. Due to mixed evidence and plausibilities, we will refrain from putting forward specific hypotheses on the relationship between WREA and work-related criteria and consider them on an exploratory basis.

Moderators

Boundary theory suggests that there are interindividual differences in the degree individuals prefer segmenting or integrating life domains and their inherent roles (Ashforth et al., 2000). Segmentation preferences describe the extent to which individuals desire to create and sustain boundaries between work and private life (Kreiner, 2006). Person-environment-fit theory (PE-fit theory; Edwards et al., 1998) suggests that negative outcomes, for instance stress, do not arise either from the individual or the environment separately, but from the (mis)fit with one another. It is likely that segmentation preferences moderate the relationship between WREA and some of the criteria, because for employees who prefer integrating work and private life, WREA might be a desirable state of affairs, whereas for people who prefer segmenting life domains, WREA should be adverse (Kreiner, 2006). We therefore assume that undesirable relationships between WREA and recuperation, well-being and private life should be stronger in samples with a higher preference for segmentation (H4).

Moreover, the relationship between WREA and employee outcomes may depend on boundary conditions that moderate the relationships (Pangert et al., 2016). On the one hand, WREA may have more severe consequences if employees have a higher workload (Pangert et al., 2016) as such individuals have a stronger need for recovery (Meijman & Mulder, 1998), and expanding work tasks into leisure time reduces the time available for recovery. On the other hand, these associations might also be weaker for individuals with high workloads as availability behaviors can be considered a way of dealing with high work demands and relieving stress (Menz et al., 2016). Gender may be another boundary condition: Due to the double burden of family and work, adverse associations may be stronger for women compared to men, as women continue doing more housework and care work (Sayer, 2016) and may therefore suffer more serious consequences by work creeping into private life. Cultural factors may also be of importance: Braukmann (2017a) found moderator effects of culture for the relationships between aspects of WREA and WFC, work-life-balance, work-family guilt and detachment. However, Braukmann’s (2017a) results did not all point into the same direction. For instance, the relationship between frequency of ICT use and WFC was more adverse in the Chinese subsample, whereas there was a stronger negative relationship for detachment in the German subsample.

Last, the associations of WREA and several criteria may not be the same for all age groups. On the one hand, older employees have better emotion regulation in regard to certain job demands such as workload or conflict between life domains (Mauno et al., 2013), which suggests stronger adverse relationships between WREA and criteria for younger employees. On the other hand, the effects may be less severe for younger employees as they are more accustomed to ICT, and WREA may be habitual for them.

To investigate these speculations, we will explore working hours (as an indicator of workload), gender, the origin of the sample (as an indicator of culture) and sample age as possible moderator variables. Due to mixed plausibilities, we will not put forward any specific moderator hypotheses for these variables, but perform these moderator analyses on an exploratory basis.

Inclusion and exclusion criteria

Studies had to meet the following inclusion criteria: a) quantitative data, b) a predictor variable constituting WREA behaviors or demands (see Table 1), c) a relevant criterion variable related to recovery, well-being (immediate, mid- or long-term), private life or work (see Table 2), and d) an association (e.g., correlation) between WREA and the criterion variable is reported, e) the sample consists of working individuals, and the study is published f) in a peer-reviewed scientific journal g) either in English or German. We excluded research dealing with the effects of teleworking, flextime and the like, because these studies capture employee-oriented forms of flexibility (Hornung et al., 2008) and not what we consider WREA in the above defined sense. Furthermore, we excluded studies dealing with the effects of on-call work, as on-call work constitutes a regulated form of availability.

OPEN IN VIEWER

Table 1.

Codings for the predictor constructs.

Synthetic construct	Operationalization
Availability behaviors	boundary crossing/ permeability/ boundarylessness boundary strength (reversed) connectivity behaviors family-to-work transitions integration technology assisted supplemental work/ bringing work home work contact work-related smartphone use during off-job time/ after hours/ in the evening work-related technology use during off-job time/ after hours/ in the evening work interrupting non-work behaviors work-related extended availability
Availability demands	24/7 availability extended availability for work org. expectations to respond to email outside of work response expectations workplace norms

OPEN IN VIEWER

Table 2.

Codings for the criterion constructs.

Domain	Construct	Operationalization
Recuperation	Relaxation	relaxation
	Psychological detachment	psychological detachment
	Recovery	recovery
	Sleep quality	sleep problems sleep quality (poor)
	Sleep quantity	sleep quantity
Well-being	Negative affect	negative affect (negative) reactions to role-boundary interruptions anger/guilt
	Positive affect	positive affect happiness
	Depletion	cognitive/ ego depletion
	Psychological distress	psychological distress depression/ anxiety stress (when focus is on psychopathological symptoms) strain
	Stress	stress (if focus is not on strain symptoms)perceived stress
	Burnout	general burnout
	Emotional exhaustion	emotional exhaustion
	Physical exhaustion	physical exhaustion/ fatigue
Private life	Work-to-family conflict	general work-to-family/life/home conflictnegative work-to-home/family spillover
	Time-based WFC	time-based work-to-family conflict
	Strain-based WFC	strain-based work-to-family conflict
	Behavior-based WFC	behavior-based WFC
	Work-life-balance	work-life-balancesatisfaction with work-life-balance
	Family-role performance	family-role performance(good) family functioningdisruptive to family role (reversed)
	Life satisfaction	life/ family/ home/ marital satisfactionquality of social life
	Positive spillover	positive work-family/home/life spilloverwork-to-home enrichment
	Positive affective spillover	positive affective work-family/home/life spillover
	Positive instrumental spillover	positive instrumental work-family/home/life spillover
Work	Absence from work	sickness absenceabsenteeism
	Affective commitment	affective commitment
	Job involvement	job involvement
	Job satisfaction	job satisfaction
	Turnover intentions	turnover intentions
	Work engagement	work engagement

Search strategies

We used the multi-page search string developed by Pangert et al. (2016) for identifying relevant studies and conducted a systematic and stepwise parallel literature search in PsychInfo, PSYNDEX, and Medline via EBSCOhost. The original search was conducted on March 13^th 2018 and updated several times with the last update on September 10^th, 2021. Furthermore, we used the ancestry approach by searching the reference lists of the articles identified for inclusion in the first phase of the literature search. Finally, we also employed the descendancy approach by identifying research that had cited studies already included.

In addition, we contacted 176 researchers in the field and asked them to send relevant unpublished research (dissertations, unpublished manuscripts, conference papers etc.), we conducted a systematic database search for relevant dissertations and an unsystematic manual search via Google scholar. We did not include these studies in our main analyses, because they do not fit the inclusion criteria. However, we conducted additional analyses, showing that expanding the data base did not substantially change the results (see section on publication bias). Detailed results for the analyses including unpublished research can be requested from the first author of this research.

Screening and eligibility

For the screening, we looked at titles and abstracts, and the studies that were not discarded after this first step were assessed by reading the full text. The inclusion process is displayed in Figure 1. Eligibility was coded independently by two raters. For the original search (80% of studies), the first author and a research assistant double-coded all studies. The agreement in the eligibility ratings was 91% and Cohen's kappa κ = .80. The updates (20% of studies) were coded by the first and the second author. The agreement was 93%, and Cohen's kappa κ = .86. All disagreements were resolved through discussion.

OPEN IN VIEWER

Figure 1.

Flow diagram on literature search.

Study coding

We coded measures indicating that employees perform work-related tasks during leisure time or felt demands to do so as indicators of WREA. In all cases, we carefully reviewed the scale items to determine whether the measure constituted an adequate representation of WREA's conceptual core. When reviewing the scale items, we found that the measures can be grouped into a) measures capturing availability behaviors such as taking work-related calls, reading work e-mails or performing other work-related tasks during leisure time and b) measures capturing perceived demands to do so (see Table 1). We conducted separate analyses for both groups. Also, we found measures that could not be properly categorized as emotional exhaustion as they assessed physical forms of exhaustion. As we did not hypothesize on physical exhaustion, but did not want to discard the construct, we conducted an additional exploratory analysis for physical exhaustion. We coded the mean age, the percentage of women, work hours, study origin, and the mean segmentation preference scores as possible moderators. Segmentation preferences were measured with a 5-point Likert scale in some studies and with a 7-point Likert scale in others. To achieve a common metric, we divided the sample means by the number of response options.

Statistical methods

All analyses were conducted with the metafor-package (Viechtbauer, 2010) for R (R Core Team, 2017). As effect size measure, we used the between-person correlation coefficient. We aggregated effect sizes if there were at least two studies reporting a particular criterion. If no correlations were reported, we contacted the authors and asked for a correlation matrix. In four cases, we received no answer, but used the reported beta coefficients applying the formula by Peterson and Brown (2005) to estimate correlations. Four studies reported correlations separately for men and women. Here, we converted correlations into Fisher's z, computed the average and converted the average back (Silver & Dunlap, 1987). In two cases, we computed Hedges’ g on the basis of the means, SDs and group sizes provided in the study and converted Hedges’ g into a correlation coefficient (Field & Gillet, 2010). For depletion as a criterion we did not use between-person, but within-person correlations, because only two of four studies reported between-person correlations, but all studies reported within-person correlations.

If several articles used the same sample, we included the correlation between WREA and a criterion variable only once to preserve the independence of the effect sizes (Johnson & Eagly, 2014). If multiple variables could be considered operationalizations of WREA, we used the measure that represented our construct best, the exception being that we extracted two correlations if both a measure for availability behaviors and perceived availability demands were included, as behaviors and demands were analyzed separately. Because of the expected heterogeneity due to differences in the operationalization of predictor and criterion variables, we decided to conduct a random-effects meta-analysis to account for the between-study variance (DerSimonian & Laird, 1986), and we utilized the Hunter and Schmidt estimator from the metafor-package. We conducted both a meta-analysis on the raw correlation coefficients with corrections for sampling error ( r ¯ = ( Σ n i r i / Σ n i ) ) and an artifact-corrected version in which we take both sampling and measurement error ( r c = ( r x y / r x x r y y ) ) into account (Hunter & Schmidt, 2004). We used the Cronbach's alphas provided in the respective studies to correct for measurement error in both the predictor and the criterion variable, and if they were not provided in the studies, we fell back upon values from the original validation studies. In case of one-item measures, we assumed perfect reliability as a conservative estimate. For all analyses, we looked at possible outliers, which we identified as studentized residuals with z-values greater than 3 and conducted analyses with and without outliers. As long as there was no relevant difference between the analysis with and without outliers, we only report the former, and when there were differences, we report both. When conducting moderator analyses, we adjusted for the mostly small number of studies in the analyses (Knapp & Hartung, 2003). The Knapp-Hartung adjustment uses the quantiles of the t- instead of the standard normal distribution and helps to account for the uncertainty in the estimate of the variance (Viechtbauer, 2010), which constitutes a more conservative test for moderation.

We calculated confidence (CI) and prediction intervals (PI) for the artifact corrected effect sizes (IntHout et al., 2016). As a measure of heterogeneity, we calculated I², which indicates the proportion of observed variation in study estimates due to heterogeneity (Higgins & Thompson, 2002). Values between I² = 25 and 50 indicate low levels of heterogeneity, values between I² = 50 and 75 constitute moderate and values greater I² = 75 high levels of heterogeneity (Shadish & Haddock, 2009). Moderator analyses were performed if at least k = 10 studies were available for a given analysis. We performed subgroup analyses for categorical and mixed-effects meta-regression for continuous moderators. To assess possible publication bias, we looked at funnel plots and conducted Egger's regression test for funnel plot asymmetry (Egger et al., 1997).

Description of studies

We included 113 articles with 121 studies that relied on k = 117 independent samples. The total sample size was N = 131,872. Sample sizes ranged from N = 39 to N = 34,399. Our total sample (49% women) consisted of employees from different countries, industries and occupations. The mean age was 39.6 years. ¹ All articles were published in peer-reviewed journals between 2002 and 2021. For readability, we only report effect sizes and confidence intervals in the text. More detailed information can be found in Tables 3–5.

OPEN IN VIEWER

Table 3.

Meta-analytic correlations between availability behaviors and criterion variables.

Criterion	k	N	r	rc	95% CIc		95% PIc		I2	Hypothesis
					LLc	ULc	LLc	ULc
Psychological detachment	31	20,299	−.31	−.34	−.43	−.24	−.56	−.11	85.85	H1a (-), supported
Relaxation	4	9,571	−.15	−.16	−.27	−.06	−.32	−.00	70.93	H1b (-), supported
Recovery	4	10,023	−.16	−.18	−.35	−.01	−.45	.10	93.34	H1c (-), supported
Sleep quality	12	65,512	−.10	−.10	−.15	−.05	−.20	−.01	82.86	H1d ( + ), supported
Sleep quantity	3	851	−.15	−.18	−.26	−.10	−.26	−.10	0.00	H1e (-), supported
Negative affect	7	3,671	.03	.03	−.04	.11	−.10	.17	52.75	H2a ( + ), rejected
Positive affect	5	2,428	.05	.06	.02	.10	.02	.10	0.00	H2b ( + ), supported
Psychological distress	12	68,832	.08	.08	.00	.16	−.07	.23	93.65	H2c ( + ), supported
Depletion	4	355	.09	.09	−.02	.20	−.02	.20	0.00	H2d ( + ), rejected
Stress	8	26,326	.05	.05	.00	.11	−.02	.13	37.98	H2e ( + ), supported
Burnout	7	1,930	.14	.14	.01	.27	−.20	.51	84.05	H2f ( + ), supported
Emotional exhaustion	26	25,531	.09	.12	.00	.24	−.09	.33	76.47	H2g ( + ), supported
Physical exhaustion	4	57,511	.12	.12	.01	.24	−.08	.32	98.49	exploratory analysis
WFC	56	45,262	.28	.34	.25	.43	.12	.56	86.41	H3a ( + ), supported
TB-WFC	9	2,424	.25	.28	.19	.37	.05	.51	81.57	H3b ( + ), supported
SB-WFC	9	2,424	.23	.25	.14	.37	−.05	.55	88.70	H3c ( + ), supported
BB-WFC	2	706	.04	.05	−.06	.16	−.11	.21	48.47	H3d ( + ), rejected
Work-Life Balance	7	19,582	−.13	−.16	−.32	−.01	−.39	.06	80.96	H3e (-), supported
Family role performance	4	1,843	−.22	−.24	−.46	−.02	−.62	.14	88.86	H3f (-), supported
Life satisfaction	9	3,576	.03	.04	−.02	.09	−.10	.17	52.68	H3g (-), rejected
Positive spillover	10	3,374	.08	.09	.04	.14	.00	.18	26.87	H3h ( + ), supported
Positive affective spillover	6	3,577	.10	.11	.06	.17	.03	.20	30.12	H3i ( + ), supported
Positive instrumental spillover	3	1,418	.00	.01	−.08	.09	−.14	.15	51.56	H3j ( + ), rejected
Absence from work	2	24,063	.05	.05	.04	.07	.04	.07	0.00	exploratory analysis
Affective commitment	5	1,968	.13	.14	−.08	.36	−.37	.66	94.18	exploratory analysis
Job involvement	4	1,169	.30	.38	.25	.50	.14	.62	71.18	exploratory analysis
Job satisfaction	8	2,300	.14	.17	.06	.28	−.11	.45	78.40	exploratory analysis
Turnover intentions	8	3,982	.05	.05	.00	.10	−.03	.14	28.15	exploratory analysis
Work engagement	8	1,909	.21	.24	.11	.37	−.11	.59	84.20	exploratory analysis

Note: k = number of studies; N = number of participants; r = effect size only corrected for sampling error; r_c = effect size corrected for sampling error and attenuation; 95% CI = 95% confidence interval of r_c with LL = lower limit of the CI and UL = upper limit of the CI; 95% PI = 95% prediction interval of r_c with LL = lower limit of the CI and UL = upper limit of the CI; I² = percentage of variation due to heterogeneity; Hypothesis ( + ) = positive relationship expected and (-) = negative relationship expected.

OPEN IN VIEWER

Table 4.

Meta-analytic correlations between availability demands and criterion variables.

Criterion	k	N	r	rc	95% CIc		95% PIc		I2	Hypothesis
					LLc	ULc	LLc	ULc
Psychological detachment	8	5,653	−.33	−.36	−.43	−.30	−.49	−.24	58.47	H1a (-), supported
Sleep quality	2	849	−.30	−.34	−.41	−.27	−.41	−.27	0.00	H1d ( + ), supported
Burnout	4	933	.09	.10	.03	.18	.03	.18	0.00	H2f ( + ), supported
Emotional exhaustion	10	20,463	.13	.15	.10	.19	.08	.21	21.34	H2g ( + ), supported
WFC	8	19,281	.34	.37	.30	.45	.26	.48	47.81	H3a ( + ), supported
Positive spillover	2	782	.10	.11	.03	.19	.03	.19	0.00	H3h ( + ), supported
Turnover intentions	2	839	.06	.07	−.01	.15	−.01	.15	0.00	exploratory analysis

Note: k = number of studies; N = number of participants; r = effect size only corrected for sampling error; r_c = effect size corrected for sampling error and attenuation; 95% CI = 95% confidence interval of r_c with LL = lower limit of the CI and UL = upper limit of the CI; 95% PI = 95% prediction interval of r_c with LL = lower limit of the CI and UL = upper limit of the CI; I² = percentage of variation due to heterogeneity; Hypothesis ( + ) = positive relationship expected and (-) = negative relationship expected.

OPEN IN VIEWER

Table 5.

Moderator analyses with at least k = 10 studies.

Outcome	Moderator		k	ES	95% CI		F	R2
					LL	UL
Detachment	Mean age	b	25	0.00	−0.02	0.02	0.05	0.81
	Women (%)	b	30	0.00	−0.00	0.01	0.14	0.93
	Work hours	b	12	0.01	−0.03	0.06	0.47	6.86
	Study origin		31				0.22	4.16
	North America	r	9	−.39	−.52	−.26
	Europe	r	20	−.34	−.47	−.20
	Asia	r	2	−.37	−.55	−.19
Sleep quality	Women (%)	b	11	−0.01	−0.02	0.01	1.49	60.87
	Study origin		12				3.07	30.93
	North America	r	4	−.14	−.25	−.02
	Europe	r	7	−.10	−.18	−.01
	Others	r	1	−.34	−.55	−.13
Distress	Mean age	b	10	0.04	−0.04	0.11	1.25	31.51
	Women (%)	b	12	−0.01	−0.03	0.02	0.46	28.74
	Study origin		12				1.63	38.62
	North America	r	8	.18	.04	.31
	Europe	r	3	.07	−.09	.23
	Others	r	1	.32	.02	.62
Exhaustion	Mean age	b	17	−0.00	−0.02	0.02	0.04	1.46
	Women (%)	b	26	−0.00	−0.01	0.00	0.07	0.48
	Study origin		26				0.80	8.60
	North America	r	7	.06	−.08	.20
	Europe	r	13	.12	−.12	.36
	Asia	r	6	.19	.03	.35
WFC	Mean age	b	40	−0.01	−0.03	0.01	0.42	9.03
	Women (%)	b	56	−0.00	−0.00	0.00	0.00	0.00
	Mean work hours	b	24	0.00	−0.03	0.04	0.02	0.36
	Segmentation pref.		14	1.10	0.36	1.76	10.78*	61.06
	Study origin		56				0.31	5.12
	North America	r	36	.35	.27	.44
	Europe	r	14	.32	.09	.55
	Asia	r	5	.34	.20	.48
	Other	r	1	.48	.20	.77
Pos. spillover	Mean Age	b	10	−0.00	−0.02	.01	0.62	27.46
	Women (%)	b	10	−0.00	−0.01	0.01	0.15	7.50
	Study origin		10				1.62	60.28
	North America	r	6	.10	.04	.17
	Europe	r	4	.03	−.09	.15

Note: k = number of studies; ES = effect size; 95% CI = 95% confidence interval of the ES; LL = lower limit of the CI; UL = upper limit of the CI; F = result of the moderator test; R² = amount of heterogeneity accounted for; Segmentation pref. = segmentation preferences;.

*p ≤ .05.

Is there cause for concern, and can WREA sometimes be beneficial for employees?

Do availability demands and availability behaviors relate differently to employee criteria?

In line with our argumentation, there was no clear picture pertaining to systematic differences in effect sizes between availability demands and availability behaviors. The only exception was for sleep quality, such that with availability demands the negative correlation with sleep quality was considerably larger (−.34, 95% CI [−.41, −.27]) than with WREA behaviors (−.10, 95% CI [−.15, −.05]).

Are there conditions that buffer or amplify the associations of WREA and employee criteria?

Publication bias

Our additional analyses including unpublished studies confirmed our results with one exception (the supplementary material describing these analyses can be requested from the first author): The relationship between availability behaviors and stress was no longer significant. Egger's regression test was significant 4 out of 31 times, but in each case, looking at the funnel plot suggested that the significant result was not due to publication bias, but caused by outliers. Detailed results of the analyses can be obtained from the first author upon request.

Is there cause for concern, and can WREA sometimes be beneficial for employees?

Taken together our results show that, as expected, WREA is adversely associated with employee recuperation, their long-term psychological well-being as well as their private life. However, despite these adverse correlations, we did not find WREA to be related to negative affect. On the contrary, employees seem to experience positive affect when they engage in work-related activities in their leisure time and even carry over this positive affect into their private lives. Furthermore, our exploratory analyses support the notion that to some extent, WREA may be perceived a resource when it comes to work-related criteria, as more WREA was associated with more job involvement and work engagement as well as with higher job satisfaction. Nevertheless, WREA also appeared to be positively related to both absence from work and turnover intentions, albeit these effect sizes were very small.

It needs to be pointed out, however, that research on WREA and work-related criteria lacks studies that allow for causal interpretations. It may well be that work-related variables such as job involvement, work engagement and job satisfaction are not only or not at all favorable outcomes of WREA but attitudes that promote employees’ willingness to engage in work-related activities during leisure time (Pangert et al., 2016). Still, in the sense of the JD-R, at least job satisfaction is plausible to (also) result from WREA, as WREA may reduce the stress employees would experience during their official working hours if they did not attend to work-related tasks during their leisure time. By contrast, job involvement and work engagement may be antecedents of WREA rather than consequences. On balance, given the many adverse associations between WREA and employee criteria of recuperation, well-being and private life, it looks like WREA to some extent benefits the work domain but may come at the price of neglecting other life domains and needs. Again, our analyzing cross-sectional data in this meta-analysis does not allow for causal conclusions, but there is preliminary evidence from diary studies (Derks et al., 2014; Eichberger et al., 2021) as well as longitudinal research (Dettmers, 2017; Kinnunen et al., 2017; Thörel et al., 2021) that support the assumption that WREA causes difficulties detaching from work psychologically. Another mechanism that may be at play in the relationship between WREA and well-being in the long run is the positive affect we found availability behaviors to be associated with, perhaps resulting from being active or from the satisfaction of getting a job done. Assuming that WREA indeed causes positive affect to some extent, this seemingly positive consequence may interfere with employees’ ability to recognize harmful behavior in time to prevent adverse mid- and long-term consequences such as impaired recuperation, psychological well-being, and functioning in private life. If employees are not alerted by experiencing negative affect (e.g., Smith & Lazarus, 1990) during or shortly after carrying out availability behaviors but instead experience positive affect, their engaging in work-related activities during leisure time may be reinforced, putting them at risk of developing conditions such as emotional exhaustion or even full-fledged burnout in the long run as well as neglecting responsibilities in their private lives up to a point that they experience a sustained conflict between the private domain and the work domain.

The effect sizes pertaining to the hypotheses ranged from r = .05 (WREA and positive affect) to r = −.34 (WREA and detachment) and r = .34 (WREA and WFC). These effect sizes indicate that for employee criteria such as detachment and WFC, WREA may have a powerful effect even in the short run (Funder & Ozer, 2020). In contrast, for criteria such as positive affect, which might reinforce WREA, the association may be less consequential at the level of single events (Funder & Ozer, 2020). However, taking into account that an employee may experience a slight increase in positive affect at some of the times when engaging in work-related activities during leisure time, it could be potentially consequential for the total number of occasions in which an employee makes room for WREA over several years (Abelson, 1985; Funder & Ozer, 2020).

Do availability demands and availability behaviors relate differently to employee criteria?

In the meta-analysis at hand, the associations between most criteria and availability demands or availability behaviors respectively did not differ substantially in size nor in direction, supporting our reasoning that for methodological reasons, availability demands and behaviors exhibit similar correlational patterns. From Belkin et al. (2020), Dettmers et al. (2016a) and Mellner (2016), however, one can conclude that availability behaviors explain variance in impaired detachment over and beyond the variance explained by availability demands, suggesting that they may affect detachment from work via different mechanisms. Above, we have argued that WREA can be considered a job stressor in itself (e.g., because employees feel pressured to be available) and may also mentally activate or sustain other job stressors (e.g., when a work call triggers unpleasant thoughts about unfinished work tasks). We speculate that perceived availability demands may primarily affect detachment as a job stressor in its own right, while availability behaviors unfold their negative impact on detachment through momentary transitions from the roles exhibited in private life to the work role. In an on-call study by Dettmers et al. (2016b) employees had an elevated cortisol awakening response on days with availability requirements compared to days without. This lends support to our reasoning that availability demands may act as a stressor in its own right and affect detachment via a general state of activation, although availability requirements in the context of on-call work may be more consequential compared to perceived availability demands.

Although relationships between availability behaviors and demands with different criteria were for the most part similar, there was one exception. The negative association we found for availability demands and sleep quality was much stronger (r = −.34) than the negative association between availability behaviors and sleep quality (r = −.10). This difference may be due to the limited number of studies (n = 2) that could be extracted for the association between availability demands and sleep quality, but there may also be content-related explanations: As we have argued above in the context of job satisfaction, employees may feel that by working in their leisure time, they can reduce the stress they would otherwise experience at work the next day. In that vein, availability behaviors seem to some extent buffer against the negative effects impaired detachment has on sleep quality, perhaps due to a sense of relief that goes along with completing a task that would otherwise be on the already cramped to-do list tomorrow. Importantly, employees may feel autonomous when they choose to work during leisure time to achieve an outcome they consider beneficial for themselves (i.e., stress reduction on the next day) or if they work for their own enjoyment (i.e., intrinsic interest in the task). However, this feeling of autonomy is likely not to be roused if employees feel that they are expected to work or working is demanded by supervisors or colleagues or if employees feel that working during leisure time is necessary to keep up with their heavy workload (i.e., when WREA is realized with a prevention focus rather than a promotion focus). Finally, rumination impairs sleep (Clancy et al., 2020), and perceiving availability demands generates mental pressure, which resembles rumination more than the behavior that is acted out as part of availability behaviors. This may explain why availability demands are associated with impaired sleep quality to a higher degree than availability behaviors. This reasoning is in line with empirical research that found the relationship between availability behaviors and emotional exhaustion to be no longer significant when availability demands were controlled for (Dettmers et al., 2016a; Piszczek, 2017). Based on the results of this meta-analysis, on the cited empirical research and on our reasoning, and despite a substantial overlap between the two constructs, we conclude that availability demands and availability behaviors should not be viewed as interchangeable.

Are there conditions that buffer or amplify the associations of WREA and employee criteria?

Aligning with findings of primary research regarding other criterion variables associated with WREA (Derks et al., 2016; Thörel et al., 2020b; Xie et al., 2018), in samples with a stronger preference for segmenting work and private life, the association between availability behaviors and WFC was more pronounced than in samples with weaker segmentation preferences. Even though it was not possible to consider segmentation preferences as a potential moderator for all criteria examined in the present meta-analysis due to a lack of sufficient primary research, we deem it plausible that an employee's preference for integration or segmentation affects their affective responses and psychological distress as a consequence of WREA, as well. If employees who prefer separating work and private life need to engage in work-related activities during their time off work, they likely experience dissonance between their attitudes and behavior, which is followed by a negative affective response.

However, this meta-analysis can only be viewed as a starting point when it comes to gaining a comprehensive picture of boundary conditions in the context of WREA. In an area that is as complex as the work domain where jobs with the same description can vastly differ, it is almost certain that there are characteristics that have an influence on the associations between WREA and certain employee criteria. Moreover, and potentially interacting with these work conditions, there are interindividual characteristics, ranging from demographics to work-related attitudes, traits and personal circumstances, that in all likelihood affect how different individuals experience the same work situation. The fact that the strength of associations investigated in this meta-analysis was characterized by a high degree of heterogeneity supports the notion that there are circumstances that are relevant as boundary conditions. Yet, these constructs have rarely been in the focus of research, nor were they assessed in a consistent manner, if assessed at all. This meta-analysis was thus limited in several ways when it comes to examining potential moderators. First, only few of them were assessed in a sufficient number of primary studies for us to be able to conduct analyses. These variables were mostly demographic (i.e., age, gender, and study origin). While age is commonly assessed in some form (e.g., in age categories), it is often not assessed as a continuous variable, excluding the respective studies from our analyses for this variable. Second, even if a potential moderator variable was assessed in enough studies to meet the inclusion criterion, the power to detect effects was likely reduced due to the limited total number of samples for each moderator variable. Third, we were limited with regard to the operationalization of potential boundary conditions. For instance, the attempt of operationalizing workload in this meta-analysis was via the number of work hours per employee, which is a simplistic operationalization of workload (Jahns, 1973 as cited in Fan & Smith, 2017). We consider it likely that these methodological issues play an important part in the majority of our moderator analyses not reaching significance.

Theoretical implications

The vast majority of our results is consistent with the postulates of the different theories we drew upon to derive the hypotheses, suggesting that the differential associations between WREA and recuperation, well-being, private life and work criteria cannot be explained by one single theory but by a combination of extant theories. As suggested by boundary theory (Ashforth et al., 2000), we found WREA to be adversely associated with indicators of recovery, particularly with psychological detachment from work, as well as with conflict between life domains. In line with the ERM and SDM, through the lens of which WREA impacts detachment and recovery directly but is only a distal influence on employees’ mid—and long-term psychological well-being, we found stronger relationships between WREA and recovery-related criteria and weaker associations between WREA and sleep as well as indicators of long-term psychological well-being. As being beyond the scope of this meta-analysis, we neither tested indirect effects of WREA on long-term well-being through detachment nor did we test associations between detachment and indicators of long-term well-being, but there is evidence for both the effects of detachment on well-being (e.g., a meta-analysis by Wendsche & Lohmann-Haislah, 2017) and indirect effects of WREA on indicators of long-term well-being via detachment (e.g., Belkin et al., 2020; Derks et al., 2014; Dettmers, 2017), so our results fit into the picture. Our finding that the strength of segmentation preferences determines the strength of the relationship between WREA and WFC is consistent with the tenets of PE-fit theory as well as with primary research (e.g., Butts et al., 2015; Derks et al., 2016). Moreover, while boundary theory predicts that individuals vary in their preference to segment or integrate their work and private roles (Ashforth et al., 2000), the moderator effect we found for segmentation preferences in this meta-analysis shows that these preferences actually make a difference when it comes to employee criteria.

Although apart from sleep quality, we did not find systematic differences in the strength of relationships between availability demands versus availability behaviors with any criterion, some primary studies suggest that availability behaviors and demands are indeed differentially related to certain employee criteria. Emotional exhaustion may primarily occur due to availability demands and not behaviors (Dettmers et al., 2016a; Piszczek, 2017), whereas detachment may be impacted by both via different mechanisms (Belkin et al., 2020; Dettmers et al., 2016a; Mellner, 2016). A general problem in the literature is that while there are two common measures for availability demands (Day et al., 2012; Dettmers et al., 2016a), availability behaviors are assessed with an array of different measures, which have often been newly constructed for the respective study (e.g., Fenner & Renn, 2010). Moreover, most studies include a measure for either availability demands or availability behaviors. Based on the high intercorrelations between availability demands and behaviors (e.g., Barber & Santuzzi, 2015; Mellner, 2016; Piszczek, 2017; Reinke & Chamorro-Premuzic, 2014) it can be assumed that when a study includes a measure for one of the two, this measure is always contaminated by the other construct. To draw firmer conclusions on the extent to which these two constructs can be differentiated with regard to their relationship with employee criteria, first, a measure for the assessment of availability behaviors is needed that demonstrates discriminant validity over existing measures assessing availability demands. Second, when studies on WREA are conducted, both constructs should be assessed to be able to control for the other one and thus determine the incremental variance that the respective other construct explains.

Directions for future research

As indicated above, a consistent theoretical framework to explain WREA does not exist yet. Future research would benefit from an integrative perspective to gain a more comprehensive understanding of WREA, its short and long-term employee consequences as well as of moderators affecting these relationships. Such a framework would facilitate our understanding of cross-connections between the phenomena considered in this meta-analysis and would benefit the identification of mechanisms through which WREA impacts employee outcomes.

In that vein, future studies need to examine how availability demands and availability behaviors relate to one another. To our knowledge, only two studies have explicitly looked at the relationship between availability demands and behaviors (Belkin et al., 2020; Piszczek, 2017), both finding that availability demands predict availability behaviors. However, both studies were cross-sectional, so even though we consider it reasonable to assume that availability demands precede availability behaviors, we do not know how they relate to one another over time and which other factors are at play in the decision to take care of work-related tasks during leisure time given that availability demands are perceived. Furthermore, as indicated above, future research should disentangle the effects of availability demands from the effects of availability behaviors.

While a few studies have focused on affective experiences as an outcome of availability behaviors (Eichberger et al., 2021; Reinke & Ohly, 2021), to date no study has addressed the role of affect in encouraging availability behaviors, that is, affect as an antecedent. Positive affect, especially as a result of WREA, may reinforce availability behaviors and the absence of negative affect, especially as a consequence of WREA, may do nothing to prevent an employee from engaging in work-related tasks during leisure time in the future. This proposed mechanism has yet to be addressed in empirical research. Research on WREA tends to use diary studies to achieve a longitudinal perspective. However, constructs are assessed retrospectively and are thus affected by recall bias and compromised ecological validity. As affective reactions are fleeting, we propose that research on affect as an antecedent, mediator and consequence of WREA would benefit from ecological momentary assessment as it allows for the assessment of feelings, experiences and behaviors in real time and in employees’ real world environments (Shiffman et al., 2008).

Relatedly, our meta-analysis lacks the foundation for concluding whether work-related attitudes are an antecedent and/or a consequence of WREA. Moreover, assuming that positive work-related attitudes are in fact a consequence of WREA, we do not know how sustainable these work-related attitudes are. More generally, we do not know the timeframe and the dynamics of how WREA affects employee outcomes. While some of the outcomes in this meta-analysis have been investigated in diary studies covering periods of three to ten days (e.g., Derks et al., 2016; Gombert et al., 2018a) very few studies have examined longer time frames using crossed-lagged panel designs. Two studies have covered periods of a few months (Dettmers, 2017; Thörel et al., 2021) and one study a period of one year (Kinnunen et al., 2017). Our causal understanding of how WREA relates to relevant employee criteria would benefit from empirical research that covers longer periods of time. With regard to potential long-term consequences of WREA, such as burnout and WFC, one probably needs a coverage of months and years.

In the present meta-analysis, WREA was related to both short- and long-term well-being. Although there are primary studies that suggest that detachment mediates the relationship between both WREA and sleep as well as long-term well-being (e.g., emotional exhaustion), there is no study tackling the process as a whole: Both detachment and sleep may mediate WREA's impact on long-term well-being (Li et al., 2020). Besides us believing that WREA impedes employees’ long-term well-being through short-term outcomes, long-term employee outcomes from one domain may affect long-term employee outcomes from another domain. For instance, if an employee's WREA causes WFC, this conflict may deepen this employee's exhaustion. Preliminary evidence supports this assumption as Dettmers (2017) found WFC to mediate the relationship between availability demands and emotional exhaustion. To pinpoint the interrelatedness of short- and long-term effects of WREA as well as of relevant phenomena across different domains, and to provide different leverage points for interventions, we suggest that researchers pay more attention to the role of recuperation as well as to criteria related to conflict between domains as mediators in the relationship between WREA and employees’ long-term well-being.

We encourage researchers to study potential moderator variables in future research. For instance, the degree to which both employees’ work and family role are central aspects of their self-concept (work and family centrality; e.g., Carr et al., 2008; Rosso et al., 2010) is a likely moderator. The more central an employee's family role is to their identity relative to their work role, the more dissonance between availability behaviors and values is to be expected and thus more adverse effects on well-being. Furthermore, future research may examine the extent of actual responsibilities employees have in their private life to moderate the effects of WREA on well-being. After all, engaging in work-related tasks during leisure time effectively takes away time from other activities. If an employee cannot fulfill the role expectations family members hold toward them, this is likely to affect them negatively both in the short- and long-term.

Employees may perceive their autonomy compromised when they feel obliged to engage in WREA to fulfill expectations of others. Aside from considering autonomy as a mediator between WREA and employee outcomes, an employee's perception of autonomy may also affect the strength of these relationships. Negative effects may be less pronounced if employees feel that their availability behavior was based on their own decision rather than resulting from (perceived) demands of their supervisors or colleagues. A construct related to autonomy is boundary control, that is, employees’ perception that in spite of boundary violations, they still have control over when and to what extent they allow spillover from work to their private life (e.g., deciding that they will not be taking calls during a family activity).

There are work characteristics that likely impact the degree to which WREA affects certain employee outcomes. Due to the methodological limitations in examining workload as a potential moderator in the meta-analysis at hand, we encourage future research to address the role of workload. Another potential work-related moderator is job type. If an employee's job, for instance, consists predominantly of cognitive work, external triggers are not necessary for role transitions to occur so that work-related thoughts will more likely encroach upon their private lives. Thus, detachment from work is likely going to be more difficult for this employee than it would be if she had to be in a particular environment (e.g., in front of a computer) to carry out work-related tasks.

Lastly, we wish to point out directions for future research from a methodological standpoint. While most of the studies considered in this meta-analysis made use of self-report measures with a handful of exceptions in which also spouses or supervisors were included (e.g. Carlson et al., 2018), we would better understand WREA if both other data sources (i.e., colleagues, supervisors, family members) and more objective measures were used. For instance, the perspective of spouses or family members is a valuable source to paint a more comprehensive picture about conflict between domains as perspectives may differ between family members. Moreover, employees may not always be aware of their own stress response and level of physiological activation when engaging in work-related tasks during leisure time. Assessing certain physiological parameters such as electrodermal activity (Posada-Quintero & Chon, 2020; Rockstroh et al., 2020a), hair cortisol (Fendel et al., 2020; Staufenbiel et al., 2013), heart parameters (Blum et al., 2019; Kim et al., 2018), keyboard typing (Freihaut & Göritz, 2021; Vizer et al., 2009), breathing patterns (Rockstroh et al., 2020b; Widjaja et al., 2013) or mouse usage (Freihaut et al., 2021; Zimmermann et al., 2003), may contribute to a more objective picture of the strain associated with WREA. We also believe it valuable if studies not only assess the consequences but also the actual extent of WREA more objectively than is possible with self-reports, as employees may not necessarily be aware of the total time they spend engaging in WREA.

Practical implications

Several practical implications can be derived from our results. Given the adverse associations between WREA and a range of employee criteria, it seems that many employees would benefit from a reduction of WREA. Such a reduction can be achieved by both organizational and individual measures (Pauls et al., 2019). Since WREA is often not explicitly demanded by employers but instead encouraged by organizational culture (i.e., “a pattern or system of beliefs, values, and behavioral norms that come to be taken for granted as basic assumptions and eventually drop out of awareness”; Schein & Schein, 2017, p. 6), first, we consider it important that employers make an effort to become aware of potential implicit influences in their organization that may make employees feel that WREA is demanded of them (e.g., by asking their employees directly or by conducting a survey among them). Such influences may be behaviors exhibited by managers, such as sending e-mails outside employees’ working hours. There may also be employees receiving calls from their colleagues during leisure time to discuss work-related matters, or employees having often experienced that important decisions are made while they are off (Kossek, 2016). Depending on the organizational influences discovered that cause employees to attend to work-related tasks during their leisure time, interventions need to be tailored accordingly. For instance, employers may change their own behaviors or communicate clearly what is expected of employees and what is not.

However, even when employers create an environment that gives employees the chance to create or maintain boundaries between their work and private lives, employees do not necessarily have to make use of it. Based on the finding that WREA is related to positive but not to negative affect, employees may have other reasons to engage in work-related matters than merely perceived external demands. This indicates that if employers take measures to strengthen boundaries between work and private life for their employees, it might not have the desired effect because employees may disregard or try to circumvent these measures. For instance, some German manufacturers have been reported to restrict their employees’ access to work e-mails to their official work hours (Strobel, 2013). If employees feel that working on emails during leisure time benefits them (e.g., by reducing the next day's workload; Menz et al., 2016), they could try circumventing this restriction by forwarding e-mails to their personal e-mail address before they leave their work place. Moreover, if employees feel motivated to work during their leisure time, a restriction may backfire because it compromises their sense of autonomy. Thus, employers should not only take (perceived) organizational demands for WREA into account but also their employees’ motivation to engage in work-related tasks while being off work. For instance, employers could make efforts to establish an organizational atmosphere in which employees feel comfortable to communicate when their workload exceeds their capacities or sensitize supervisors in their company to proactively and regularly query the employees about their current workload. If employees feel that they are able to complete their work during their official work hours, they will less likely feel inclined to take their work home with them (Thörel et al., 2020a).

Apart from these organizational strategies, we suggest that employees will also benefit from being taught individual strategies to deal with WREA such as creating healthy boundaries wherever their work circumstances allow for it (e.g., turning off notifications for e-mails or messages in case employees use the same mobile device for work-related and private activities) or techniques that aim at stress reduction when a certain extent of WREA cannot be avoided. Again, whether it is employers who address their employees with interventions that aim at teaching them individual strategies or health campaigns, it needs to be kept in mind that employees may have a personal motivation for WREA and may not experience negative affect when engaging in work-related activities outside their official work hours. Thus, for employees to actually apply individual strategies, they may first need to be convinced of why they should refrain from more-than-once-in-a-while engaging in work-related activities during their time off work. This could be achieved by educating employees about possible long-term consequences of WREA.

Based on our result that segmentation preferences moderated the relationship between WREA and WFC we suggest that employers take into account that employees differ in their preference to integrate or segment their work and private life. From the perspective of PE-fit theory this could mean that employers take measures to hire employees whose preferences align with the degree of integration a particular position demands. One way of achieving this is to assess the segmentation preferences of potential employees during the application process and making the results of that assessment part of the decision process of whom to hire. Another way would be to give employees the chance for self-selection, for instance by providing them with a realistic job preview that clearly states potential demands for WREA. Apart from aiming at selecting employees that match the availability demands of a given position, employers may also have certain degrees of freedom to allow for employees and their families to be compensated for instances in which WREA cannot be avoided. For example, if it is necessary that an employee finishes a task after work, employers could establish options for employees to get compensated for this boundary violation by allowing them to leave early or start work later on another day within this week (Pauls et al., 2019). While employees are potentially rewarded with the satisfaction of their employer or colleagues when they engage in WREA, their families do not directly benefit from such rewards, but bear the brunt of the negative consequences of when their partner or parent engages in WREA (i.e., their failure to fulfill certain role expectations and responsibilities). Given that WFC is the result of a mid- or long-term asymmetry between the fulfilment of work- and family-related roles, granting employees the flexibility to be compensated with time off work when needed could be a buffer against the negative effects of WREA on WFC. This reasoning is supported by meta-analytic results on the negative relationship between different forms of workplace flexibility and WFC, although the effect sizes were small (Allen et al., 2013).

Another recommendation that has sometimes been made (e.g., Pauls et al., 2019) is to substitute WREA by a formal arrangement and compensate employees financially for their additional time in case they are needed to work during leisure time. However, evidence from on-call studies (e.g., Bamberg et al., 2012; Dettmers et al., 2016b; Lee et al., 2020) suggests that although a financial compensation for work during leisure time may be better in terms of fairness and appreciation, the effects on employee health do not seem to be less severe. On-call work is also associated with problems detaching from work (Dettmers et al., 2016b), reduced well-being (Bamberg et al., 2012; Lee et al., 2020) and sleep problems (Lee et al., 2020).

Limitations

There are limitations to this meta-analysis, of which the reader should be aware. Although the theoretical and implicit assumption is that WREA is the independent variable and influences the criterion variables, the analyzed correlations in this meta-analysis are cross-sectional, so causal relationships cannot be ascertained. With several variables looked at in this meta-analysis it seems likely that they affect WREA or that there are reciprocal relationships. Another issue is that with some exceptions, the number of studies per criterion is not large, hence more evidence is needed to draw firmer conclusions. Relatedly, moderator analyses were restricted in that many potentially relevant moderators have been assessed in only a small number of studies if at all, with the makeshift solution of employing proxies for constructs of interest (e.g., age as a proxy for job experience). Thus, on the one hand, the null findings concerning most moderators do not preclude the constructs of interest from being relevant. On the other hand, it must be acknowledged that with 20 tests, the one significant moderator effect we found is within the margin of error. However, since there was a strong theoretical basis for segmentation preferences as a moderator and there are also primary studies that identified segmentation preferences as a moderator (e.g., Derks et al., 2016), we believe it may be a true positive finding. Moreover, WREA being an umbrella term that covers a range of constructs, although we conducted separate analyses for availability behaviors and perceived availability demands, the constructs assembled under each of the labels availability behaviors and availability demands are not homogeneous, which is also reflected by the statistical heterogeneity we encountered. Finally, although we attenuated for measurement error as much as possible, in studies where constructs were measured with a single item we assumed perfect reliability as a conservative approach. These one-item measures primarily occurred in studies with many participants, which thus have most weight in the overall result. Therefore, some associations between WREA and criteria may be higher than reflected in this meta-analysis.