Determining conditions for communication technology use induced work-life conflict and exhaustion: On the moderating roles of task interdependency and preferred
boundary management style.
Mikael Swiebel (6051308) Master’s Thesis
Corporate Communication
Graduate School of Communication University of Amsterdam
dr. J.M. Slevin 30-06-2017
Abstract
Communication technology use has an ambiguous relationship with work-life conflict and exhaustion. On the one hand, communication technology use can decrease work-life conflict and subsequently exhaustion by offering the opportunity to work wherever, whenever. On the other hand, communication technology use can increase work-life conflict and subsequently exhaustion by allowing demands from the ‘work’ domain to spill over to the ‘life’ domain. To successfully combat work-life conflict and exhaustion, organizations need to understand under which conditions and for which employees communication technology use can cause work-life conflict and exhaustion. Therefore, two moderators that might foster a positive relation between communication technology use and work-life conflict and subsequently exhaustion, are investigated: task interdependency and preferred boundary management style. In this study of 203 Dutch workers, a cross-sectional online survey is used to demonstrate that high task interdependency and a preference for segmenting the ‘work’ and the ‘life’ domain are conditions under which communication technology use can lead to work-life conflict and exhaustion. Organizations can use this information to protect their workforce from exhaustion and burnouts, by targeting vulnerable employees with work-life conflict and exhaustion counter measures.
Introduction
Burnouts cost the Dutch Economy 1.8 billion euros per year (TNS, 2015). Reducing the number of burnouts should thus be an important concern for many organizations. A burnout is an elongated response to emotional stressors on the job (Maslach, Schaufeli & Leiter, 2001). The core burnout dimension and the primary predictor of experiencing burnouts is exhaustion. A plethora of stimuli can cause exhaustion. Out of all these possible stimuli, this research paper focusses on work-life conflict (WLC), which has been linked to exhaustion in both cross-sectional (Bakker, Demerouti & Euwema, 2005) and longitudinal research (Demerouti, Bakker & Bulters, 2004).
The influx of internet based communication technologies on the work floor over the last three decades has led to the erosion of spatial and temporal communication barriers (Nansen, Arnold, Gibbs & Davis, 2010). Previous research shows that, due to this erosion, communication technology use (CTU) can both alleviate (Gajendran & Harrison, 2007) and intensify (Derks, van Duin, Tims & Bakker, 2014; Van Zoonen, Verhoeven & Vliegenthart, 2016) WLC and subsequently exhaustion. Therefore, the research problem that this study will address is that CTU allegedly leads to WLC and exhaustion in some situations, but not in
others. A social constructionist view of technology, which holds that the outcomes of adapting a new technology are in part dependent on the conditions under which the technology is used (Orlikowski, 2000), will be used to investigate this discrepancy. The aim of this study is thus to determine conditions under which CTU leads to WLC and subsequently exhaustion. By locating employees who are at risk for CTU induced WLC and exhaustion, organizations become enabled to administer WLC and exhaustion counter measures.
CTU allows employees to work without being bound to a specific time or place, offering employees the autonomy to regulate and synchronize demands from either the ‘work’ or the ‘life’ domain, possibly resulting in reduced WLC (Nansen et al., 2010; Yun, Kettinger & Lee, 2012). At the same time CTU entails a constant connectivity to both domains (Edwards & Rothbard, 2000; Nippert-Eng, 1996), making it easier for demands from one domain to spill over to the other domain, possibly leading to increased WLC (Standen, Daniels & Lamont, 1999). Qualitative research has shown that task interdependency, which means that employees have to interact, coordinate and cooperate to successfully complete their tasks (Wageman & Bakker, 1997; Stewart & Barrick, 2000), increases the amount of demands crossing the barrier from the ‘work’ to the ‘life’ domain (Mazmanian, Orlikowski & Yates, 2013). This potentially increases WLC and undermines the newly gained autonomy to satisfactorily manage demands from both domains. Because employees often feel obligated to answer to boundary crossing demands, they no longer freely choose when and where to work (Mazmanian et al., 2013). Even though the relationship of CTU, task interdependency and WLC has not been confirmed in quantitative research, high task interdependency seems to be a condition which causes the relation between CTU, and WLC and exhaustion, to become stronger.
External factors such as the behavior of coworkers and the amount of interdependence with these coworkers seem to be important factors for the amount of experienced WLC, but internal factors likely play a role too. Boundary management theory states that employees differ in how they preferably manage the boundary between the ‘work’ and the ‘life’ domain (Ashforth, Kreiner & Fugate, 2000). This implies that an employee has some control over the amount of boundary crossing demands he/she receives. Since WLC occurs when demands from one domain make it difficult or impossible to meet demands in the other domain (Greenhaus & Beutell, 1985), the preferred boundary management style likely influences the amount of experienced WLC. In turn, this boundary management style seems to play a role in the effect strength of WLC on exhaustion. Person-job fit theory states that exhaustion can be the result of a misfit between an employee’s desired situation and his actual situation
(Kristof-Brown, Zimmerman & Johnson, 2005). When the actual situation is experiencing WLC, the misfit with the desired situation is likely greater for employees who have the desire to separate the ‘life’ and ‘work’ domain than for employees who have the desire to integrate these domains. Hence the effect strength of WLC could be dependent on the way an employee preferably manages the boundary between the ‘work’ and ‘life’ domain.
In line with the social constructionist view of technology, two variables that possibly affect the relationship of CTU, WLC and exhaustion have been determined. Therefore, the following research question is posed:
RQ: What are the effects of task interdependency and preferred boundary management style on the relationship between communication technology use, work-life conflict and exhaustion?
The primary scientific contribution this research paper is trying to make is a better understanding of the specific conditions under which CTU will lead to WLC and exhaustion. By combining boundary management theory (Ashforth et al., 2000), person-job fit theory (Kristof-Brown et al., 2005) and task interdependency with previous research highlighting the relationship between CTU, WLC and exhaustion (Van Zoonen et al., 2016; Derks & Bakker, 2014), a better understanding of the relationship of CTU, WLC and exhaustion will be developed. To my knowledge, task interdependency, person-job fit theory and boundary management theory have not yet been tested together in a quantitative study in this context. Organizations will benefit from understanding these conditions to the relationship between CTU, WLC and exhaustion in two ways. First, there are ways to reduce WLC and exhaustion, but to successfully administer these counter measures, employees at risk should first be located. This research paper will help organizations understand for which employees CTU could lead to WLC and exhaustion. Thereby giving organizations the opportunity to protect these employees from exhaustion and an eventual burnout by for example reducing the workload, restricting CTU or training them in time-management or stress reduction. Second, this knowledge will help organizations in developing CTU policies, for example limiting CTU in non-office hours, and shaping CTU culture in a way that prevents exhaustion, for example by creating a culture that respects an employee’s personal choice in how to shape the boundary between the ‘life’ and ‘work’ domain.
Theoretical background CTU and boundary management
Due to the erosion of temporal and spatial communication barriers, CTU are reputed to blur the boundaries between the ‘work’ and ‘life’ domains (Tomlinson, 2007). Boundary management theory states that employees create these ‘life’ and ‘work’ domains to organize their environment by cognitively and behaviorally separating social, temporal, and physical arenas (Rothbard & Ramarajan, 2009). These ‘mental fences’ around the different domains ‘work’ and ‘life’ are necessary to smoothen role transitions and to avoid activation of incongruent and contrasting aspects of an employee’s identity (Ashforth et al., 2000; Ollier-Malaterre, Rothbard & Berg, 2013; Rothbard & Ramarajan, 2009). While this erosion of spatial and temporal communication barriers does not diminish the importance of time and space (Harvey, 1990), these communication technologies do offer the opportunity to work regardless of time and space (e.g. Schieman & Young, 2013; Wajcman, 2008) and thus influence boundary management in two ways. They increase the permeability of the boundary between the ‘life’ and the ‘work’ domain, potentially leading to more work–life conflict (Standen et al., 1999; Haddon & Silverstone, 2000; Valcour & Hunter, 2005), and they increase the flexibility of the boundary between the ‘life’ and the ‘work’ domain, possibly resulting in reduced WLC (Kirchmeyer, 1995; Raghuram & Wiesenfeld, 2004; Yun et al., 2012).
Boundary management, WLC and exhaustion
Both the notion that boundary flexibility can help an employee to reduce WLC and the notion that boundary permeability can increase WLC are based on resource theory (e.g. Edwards & Rothbard, 2000; Small & Riley, 1990). WLC is a form of inter-role conflict in which the demands from the two different domains ‘life’ and ‘work’ are incompatible. The ‘life’ and the ‘work’ domain compete for the same finite resources, such as time, attention, and energy. Thus, WLC is thought to occur when demands from one domain drain the resources needed to meet the demands of the other domain (Greenhaus & Beuttell, 1985; Grandey & Cropanzano, 1999). Both boundary flexibility and boundary permeability affect the distribution of these resources (Nansen, et al., 2010).
Boundary flexibility refers to an employee’s degree of autonomy to decide when and where work is completed (Ashfort et al., 2000). High boundary flexibility enables employees
to determine the best way to distribute the finite resources previously mentioned (Allen, Johnson, Kiburz & Shockley, 2013). For example, regarding the resource time, this flexibility allows employees to adjust their schedule so that they can manage care activities or reduce time used for commuting. An example regarding the resource energy, is that this flexibility allows employees to create an environment that fosters productivity. This way work demands can be met, while spending less energy than in a less productive environment. The increased control over resources due to boundary flexibility allows employees to better manage the resources needed to meet demands from the ‘work’ and ‘life’ domain, resulting in less WLC. Hence, boundary flexibility has been linked to reduced WLC on numerous occasions (Bulger, Matthews & Hoffman, 2007, Matthews Barnes-Farrell & Bulger, 2010).
Boundary permeability refers to the degree in which it is possible for demands from one domain to spill over to the other domain (Ashfort et al., 2000). High boundary permeability reduces autonomy over the distribution of the finite resources such as time, attention, and energy (Nansen et al., 2010). For example, time that is exclusively available for the ‘life’ domain in low permeability settings because of the inability to connect to the ‘work’ domain, becomes available for the ‘work’ domain in high permeability settings. As a result, work demands are more likely to occupy time that previously was solely available to the ‘life’ domain. An example regarding the resource energy is that the increased amount of interruptions and information overload that permeability facilitates, lowers productivity (Fonner & Roloff, 2012; Bucher, Fieseler & Suphan, 2013). This way more energy needs to be spent to meet demands. The decreased control over resources that boundary permeability causes, obstructs the management of resources needed to meet demands from the ‘work’ and ‘life’ domain. Hence, permeable boundaries can mean an increase in WLC (Kossek, Lautsch, & Eaton, 2006; Lapierre & Allen, 2006).
The effects of WLC on exhaustion have been well documented. There is a plethora of research linking WLC to detrimental effects on employee exhaustion: Smartphone induced WLC has been related to emotional exhaustion (Yun et al., 2012; Derks & Bakker, 2014) and conflict between different domains brings tension, which is likely to drain emotional energy (Golden, 2012). Moreover, both cross-sectional (Bakker et al., 2005) and longitudinal (Demerouti et al., 2004) studies show a significant positive relationship between WLC and exhaustion.
Communication technologies thus, by the virtue of increased boundary permeability and increased boundary flexibility, have the potential to both increase and decrease WLC and subsequently exhaustion. A social constructionist view of technology holds that, while the
materiality of a technology determines key constraints and affordances, the outcomes of adapting a technology will reflect social processes in the immediate context of the technology’s use (Leonardi & Barley, 2008). While this view recognizes that technology use is always situated and emergent, it does not imply that technology use and its outcomes are always unique. How a technology is used, and thus its outcomes, are in part dependent on the conditions under which such a technology is used (Orlikowski, 2000).
CTU, boundary management and task interdependency
Of particular interest in this light is the ‘autonomy paradox’ (Mazmanian et al., 2013), which can be interpreted as that increased autonomy due to more boundary flexibility, is undermined by decreased autonomy due to more boundary permeability in high task interdependency settings. In the current study, task interdependency is defined as: the degree to which an employee relies on others to accomplish one’s work and how others rely on the employee to accomplish their work (Wong, DeSanctis & Staudenmayer, 2007). The autonomy that is offered by communication technologies due to increased boundary flexibility is at odds with high task interdependency, because task interdependency requires high levels of interaction and coordination of employee’s tasks in timing and sequence (Wageman, 1995; Stewart & Barrick, 2000). High task interdependency inherently means that employees need to coordinate their efforts and cooperate to fulfill their tasks successfully (Guzzo & Shea, 1992; Wageman & Bakker, 1997). Due to communication technologies, the interactions, coordination and cooperation tied to task interdependency are no longer bound to time and space. The 24-hour connectedness that communication technologies entail (Bucher et al., 2013) allows employees to place demands on colleagues at any time of the day, even while out of office. Therefore, to uphold to the need of interaction, coordination and cooperation in high task interdependency environments, more demands from the ‘work’ domain, will spill over to the ‘life’ domain. Employees often feel they must respond to these demands, due to increased perceptions and expectations of availability (Mazmanian et al., 2013) and the inherent trait of interdependency that not adhering to such a demand makes a colleague’s work impossible (Wageman & Bakker, 1997). So, while CTU embodies the ability for employees to control when to work and where to work, in high task interdependency environments, demands from coworkers take away this newly gained control, by forcing an employee to cross boundaries from the ‘life’ to the ‘work’ domain. In the words of Mazmanian et al. (2013, pp. 2): “by allowing employees to work anywhere/anytime, we observe them becoming caught in a collective spiral of escalating engagement where they end
up working everywhere/all the time”. Task interdependency seems to be a condition that shapes CTU and the social processes around CTU in such a way that the effects of boundary flexibility are diminished, while the effects of boundary permeability are intensified. Hence, the following hypothesis is posed (see figure 1):
H1: There is a positive indirect relationship between communication technology use and exhaustion through work-life conflict, when task interdependency is high.
CTU, WLC and Preferred boundary management style
Previously, an organizational condition that affects the social processes revolving around CTU has been described. While social processes can be shaped by organizational conditions, they are too always involving employees who engage in these social processes. Therefore, it makes sense to look at an employee’s characteristics involved with boundary management and CTU. Preferred boundary management style (PBMS) is such a personal characteristic. Employees differ in how the boundary between the ‘work’ and the ‘life’ domain is preferably managed. When these different boundary management styles are placed on a continuum, these styles range from complete integration to complete segmentation of the different domains an employee has construed (Ashforth et al., 2000).
Segmentors will actively try to create spatial and temporal barriers, reserving specific areas and times for their ‘work’ and ‘life’. They will try to actively separate these domains based on these boundaries (Ashforth et al., 2000). For example, a segmentor will reserve the office space and office hours for the ‘work’ domain and reserve the home and non-office hours for the ‘life’ domain, resulting in the choice to limit answering work e-mails while at home. That is, segmentors will try to manage their CTU in a way that the boundary between the ‘work’ and ‘life’ domain is respected (Ashfort et al., 2000). By reserving time and space for specific roles, they reduce demands from one domain influencing the other. Less permeable boundaries, something that segmentors constantly try to achieve, result in less WLC (Clarke, 2002; Park & Jex 2011). Indeed, segmentors have been shown to experience less WLC than integrators (Olson-Buchanan & Boswell, 2006). In contrast, integrators will try to lessen the barriers between the ‘life’ and the ‘work’ domain (Ashfort et al., 2000). An integrator will mix the ‘life’ and ‘work’ domain by for example taking personal calls during office hours or answering work e-mails in their time off. By integrating these different roles, demands from one domain have a higher chance of affecting the other domain. Correspondingly, integration of the two different domains has been associated with more
WLC (Olson-Buchanan & Boswell, 2006; Park & Jex 2011). And so, the following hypotheses are posed (see figure 1):
H2b: A segmentation preference leads to a weaker relationship between work related CTU and work-life conflict.
H2a: An integration preference leads to a stronger relationship between work related CTU and work-life conflict.
WLC, PBMS and exhaustion
While integrators seem to experience a stronger relation between CTU and WLC, the effect of WLC on exhaustion seems to be reversed; the relationship between WLC and exhaustion is likely stronger for segmentors than for integrators. Person-job fit theory suggests that exhaustion can result from a discrepancy between an employee’s needs, desires and preference and his job and organizational environment (Kristof-Brown et al., 2005). Segmentors, who prefer to keep the ‘life’ and ‘work’ domain separated thus face such a discrepancy when ‘work’ demands spill over to the ‘life’ domain, as due to WLC. Whereas for integrators, alternating between demands from the two domains is the preferred situation, corresponding with a fit between the employee and the organization when demands from the ‘work’ domain spill over to the ‘life’ domain. Boundary management theory further substantiates the idea that segmentors experience more exhaustion than integrators as a result of WLC. Segmentation makes switching between domains more difficult, because the psychological gap between domains is larger, and less frequent (Ashforth et al., 2000). This reduced frequency of switching between the two different domains ensures that segmentors are less practiced in swiftly and effectively handling the change of domains. This bigger psychological gap and less practice in managing the two different domains at the same time, will possibly strengthen the effect of WLC on exhaustion for segmentors. Based on person-job fit theory and boundary management theory, the following hypotheses are posed (see figure 1):
H3a: A segmentation preference leads to a stronger relationship between work-life conflict and exhaustion.
H3b: An integration preference leads to a weaker relationship between work-life conflict and exhaustion.
Figure 1 A conceptual model of the moderated indirect relationship of CTU and exhaustion.
Method Sample and procedure
In this study, 203 Dutch employees participated in a voluntary cross-sectional online survey. The survey was held on the online platform Qualtrics. The participants were recruited via Facebook. This survey was shared in the researcher’s personal network. The requirements for participating were a workweek of at least 25 hours in an organization with at least 24 employees. Since the survey was administered in English, a sufficient command of the English language was a hidden requirement. Participants were asked to fill in the survey over a period of 12 days.
Measures
Communication technology use. CTU was measured with the following question: On a scale of 1 to 7 (1 = almost never, 7 = very often) to what extent do you use the following technologies in a work-related context (i.e. either at work or for work)? In line with Park, Fritz and Jex (2011), respondents were asked to rate how often they use a smartphone, a laptop, e-mail or social media in this context. The hardware was selected because, at its core, the mobility offered by these two devices is what enables CTU to affect boundary flexibility and boundary permeability. E-mail was picked because of its common use in organizational settings. Last, social media was picked because of its growing organizational use.
Work-life conflict. WLC was measured using five items from Netemayer, Boles and McMurrian (1999). This scale is well validated and closely corresponds with the WLC definition of Greenhaus and Beuttell (1985), which is used in this study. An example question is: ‘The demands of my work interfere with my home and family life’. These items were rated on 7-point Likert scales (1 = completely disagree, 7 = completely agree). A
principal component analysis determined one component with an eigenvalue of more than 1 (2.58). This component explained 51.71% of the variance. Factor loadings ranged from .55 To .83. Cronbach’s alpha for this scale was .75.
Exhaustion. Exhaustion was measured using a five-item scale of the Maslach Burnout Inventory (Maslach et al., 2001). The Maslach Burnout Inventory is widely recognized as the leading measure for exhaustion and burnouts and has been well validated over its 16 years of use. Example questions are: ‘I feel mentally drained by my work’ and ‘At the end of the day I feel empty’. Items were rated on 7-point Likert scales (1 = completely disagree, 7 = completely agree). A principal component analysis determined one factor meeting Kaiser’s criteria (3.40). Factor loadings ranged from .76 to .87. This component explained 68% of the variance in exhaystion. A credibility test revealed a Cronbach’s alpha of .88.
Preferred boundary management style. Segmentation and integration are mutually exclusive. When one scores high on segmentation, he/she necessarily scores low on integration. Kreiner’s (2006) scale to measure preference for segmenting work from life was used. An example question is: ‘I prefer to keep work life at work’. Participants were asked to agree or disagree on 7-point Likert scales (1 = completely disagree, 7 = completely agree). In line with Park and Jex (2011) this scale was adjusted to include preference for segmenting life from work. An example question used is: ‘I prefer to keep my family/personal life detached from my work life’. These measurements were then combined into one 8-item scale. The questions in this scale closely match the definition for PBMS coined in boundary management theory (Ashfort et al., 2000), which is used in this study. A principal component analysis revealed one factor with an eigen value of more than one (2.70). Factor loadings ranged from .74 to .87. This component explained 67.47% of the variance in PBMS. Cronbach’s alpha for these items was .83. Scoring high on PBMS corresponds with a high segmentation preference. Scoring low on PBMS corresponds with a low segmentation preference, i.e. a high integration preference.
Task interdependency. Task interdependency was measured using a seven-item scale adapted from Van der Vegt, Emans and Van de Vliert (2000). In this scale task interdependency is treated as a characteristic of a job in contrast to a characteristic of a team. The definition used in the current study is too centered around task interdependency on the personal level. Example questions are: ‘I depend on my colleagues for the completion of my work’ and ‘My colleagues depend on me for the completion of their work’. Items were rated on 7-point Likert scales (1 = completely disagree, 7 = completely agree). A principal component analysis determined one factor meeting Kaiser’s criteria (3.12). Factor loadings
ranged from .56 to .75. This component explained 44.54% of the variance. A credibility test revealed a Cronbach’s alpha of 0.79.
Control variables. Sex, age, number of kids, marital state, management position, workhours per week, number of years active in current organization, size of the organization, work sector, job insecurity, job involvement and social norms for boundary management were measured as control variables.
Job insecurity. Job insecurity has been shown to influence WLC, because employees with high job insecurity have difficulties saying no to boundary crossing demands (Boswell, Olsen-buchanon & Harris, 2014), and is thus included as a control variable. This construct was measured using a 6-item scale from Boswell et al. (2014). Participants were asked the following question: ‘How likely is it that each of these events might actually occur to you in your current job? (1 = very unlikely, 7 = very likely)’. An example event is: ‘Lose your job and be laid off permanently’. A principal component analysis determined one factor meeting Kaiser’s criteria (3.09). Factor loadings ranged from .65 to .82. This component explained 51.48% of the variance. A credibility test revealed a Cronbach’s alpha of 0.81.
Job involvement. Job involvement influences WLC, because highly involved individuals are more prone to accepting boundary crossing demands (Boswell & Olsen-buchanon, 2007) and is therefore included as a control variable. This construct was measured using a 10-item scale from Kanungo (1982). An example question is: ‘The most important things that happen to me involve my present job’. Items were rated on 7-point Likert scales (1 = completely disagree, 7 = completely agree). Two items were dropped due to unsatisfactory factor loadings and a possible Cronbach’s alpha increase, after which a principal component analysis determined one component with an eigenvalue of more than 1 (4.12). This component explained 41.16% of the variance. Factor loadings ranged from .60 to .77. Cronbach’s alpha for this scale was .83.
Boundary management norm. The last control variable included in this study measures the social norms for boundary management. A segmentation norm lowers the amount of WLC experienced by employees (Park et al., 2011). Boundary management norm (BMN) was measured using a 4-item scale from Park et al. (2011). Participants were asked to which extend they agreed with statements such as: ‘The people I work with forget about work when they’re at home.’ Items were rated on 7-point Likert scales (1 = completely disagree, 7 = completely agree). The principal component analysis yielded on factor with an eigenvalue of more than one (3.06) which explained 76.55% of the variance. Factor loadings ranged from .86 to .92. Cronbach’s Alpha for this scale is .90.
Results Descriptives
The respondents mean age was 32.16 (SD = 10.27), 53% was female and 66.7% had a degree in higher education. The respondents, of which 15.9% were married, had on average 0.37 children that were dependent on them. Respondents worked 40.74 hours a week on average (SD = 10.70), had been active for their current employer for 5.46 years on average (SD = 7.46) and 18.4% of the respondents held a managerial position. Table 1 shows an overview of the means and standard deviations of the demographic variables that had atleast one significant correlation with either technology use, WLC, task interdependency, PBMS or exhaustion. Hence the variables in table 1 are included in all the analyses as control variables. Table 2 shows an overview of the means, standard deviations, Cronbach’s alpha and correlation of technology use, WLC, task interdependency, PBMS, exhaustion, and the control variables: job involvement, BMN and job insecurity. Correlation for demographics have been determined but are not included in this table because the size of the table became unfeasible. Correlations between constructs range from -.40 to .60. When correlation between different technologies are excluded, correlations range from -.40 to .44. Which demonstrates the distinctiveness of the measured constructs. BMN and involvement correlate with WLC and uncertainty correlates with exhaustion. BMN, involvement and uncertainty are thus included in all analyses as control variables.
Table 1
Means and standard deviations of demographics
Variable M SD Gender .47 .50 Marital state .16 .37 Age 32.16 10.27 Children .37 .73 Managerial position .18 .37
Workhours per week 40.74 10.70
Table 2
Correlations and Descriptive statistics.
Variable M SD 1 2 3 4 5 6 7 8 9 10 11
1 Smartphone use 4.87 2.24
2 Laptop use 3.83 2.13 .47*
3 E-mail use 4.45 2.25 .60* .55*
4 Social media use 2.88 2.44 .58* .46* .40*
5 WLC 3.69 1.36 .29* .23* .44* .14* .75 6 Task interdependency 3.02 1.12 -.24* .03 -.14* -.07 -.13 .79 7 PBMS 5.21 1.27 -.14* -.22* -.28* -.10 -.41* .10 .83 8 Exhaustion 3.66 1.36 .04 .07 .10 .17* .19* .03 .22* .88 9 BMN 3.77 1.30 .03 -.02 -.09 .01 -.40* .15* .12 -.14* .90 10 Involvement 3.50 1.05 .38* .22* .29* .33* .28* -.27* -.27* .17* .05 .84 11 Uncertainty 2.03 1.03 .05 .08 .04 -.04 .09 -.08 -.01 .20* -.00 .20* .81
Hypothesis testing
To test the proposed hypotheses, the process macro for SPSS was used (Hayes, 2013). The 95% confidence interval was bootstrapped 5000 times for all of the following analyses. All independent variables were mean centered to make interpretation easier (Hayes, 2013). Since the process macro uses OLS regression, the data had to fit all the assumptions for OLS, such as but not limited to linearity, homoscedascity and non-collinearity. All assumptions were met to a satisfactory degree. In all analyses the following control variables were used: gender, marital state, age, children, managerial position, actual workhours per week, work years at the current employer, BMN, involvement and uncertainty. Size of the organization and work sector were measured, but not included as control variables because of a lack of significant correlations with both the used dependent and independent variables. Moreover, when PBMS and task interdependency were not included as a moderator they were used as control variable because of their correlation with technology use. Of these control variables one showed a particularly strong effect. Depending on which independent variable (smartphone, laptop, e-mail or social media use) was used, the effect strength of BMN on WLC ranged from -.40 to -.50. All these effect were significant at p = .000.
When explaining moderators, they are refered to as being high, average and low; high meaning mean + 1 SD, average being the mean and low being mean – 1 SD. The process macro does not provide beta values. Beta values can be achieved by standardising variables before analysing them with process. However, since some covariates are dummy variables, this is not desirable for this study. Moreover, all research variables, excluding covariates, are measured on the same scale, reducing the necessity of beta values. When the process macro determined significant effects that were not hypothesized, these effects are mentioned in text, but are not depicted in the models. This way the models remain clear and comprehensible.
The conditional indirect effect of CTU on exhaustion dependent on task interdependency. To test hypothesis 1; which states that there is a significant indirect relation between CTU and exhaustion through WLC when task interdependency is high, two regression models were employed. The first model has WLC as dependent variable, CTU as independent variable and task interdependency as moderator. This model tests if CTU and task interdependency have explanatory power for WLC. The second model has exhaustion as a dependent variable and technology use as independent variable, task interdependency as moderator and WLC as mediator, and tests if these variables have explanatory power for exhaustion. Both models need to be significant for an indirect effect to be present. The first model, with WLC as the dependent variable, was significant for smarthpone use
(F(14,189)=9.48, p=.000, R2=.45) which explained 45% of the variance in WLC (see figure 2), Laptop use (F(14,189)=8.14, p=.000, R2=.41) which explained 41% of the variance in WLC (see figure 3), e-mail use (F(14,189)=10.17, p=.000, R2=.47) which explained 47% of the variance in WLC (see figure 4), and social media use (F(14,189)=4.29, p=.000, R2=.26) which explained 26% of the variance in WLC (see figure 5).
Smartphone use (See figure 2) had a positive effect on WLC (b=0.15, CI 95% [0.08, 0.23], t(189)=3.94, p=.000). Task interdependency showed no direct relationship with WLC. However, there was a significant interaction effect of smartphone use and task interdependency on WLC (b=0.06, CI 95% [0.00, 0.12], t(189)=2.1, p=.037). This implies that the effect of smarthpone use on WLC is conditional (See table 3). The effect of smartphone use on WLC when task interdependency was high (b=0.22, CI 95% [0.13, 0.32], t(189)=4.55, p=.000), was bigger than the effect of smartphone use on WLC when task interdependency was average (b=0.15, CI 95% [0.07, 0.22], t(189)=3.94, p=.000), which was bigger than the effect of smartphone use on WLC when task interdependency was low (b=0.08, CI 95% [-0.03, 0.18], t(189)=1.49, p=.139). This last effect was not significant, meaning that the effect of smartphone use on WLC is only present when task interdependency is atleast average.
Figure 2 Moderated relation of smartphone Figure 3 Moderated relation of laptop
use and WLC. use and WLC.
Note: N=203. *significant at p<.05. Note: N=203. *significant at p<.05.
Table 3
The conditional direct effect of smartphone use on WLC, dependent on task interdependency
Task interdependency b t p LLCI ULCI
low .08 1.49 .139 -0.03 0.18
high .22* 4.55 .000 0.13 0.32 Note: N=203. *significant at p<.05.
Laptop use drew a different picture (see figure 3). While there was a direct effect of laptop use on WLC (b=0.10, CI 95% [0.03, 0.18], t(189)=2.68, p=.008), no significant interaction effect of laptop use and task interdependency was present. E-mail use (see figure 4) showed a similar pattern as laptop use, with a significant direct effect on WLC(b=0.19, CI 95% [0.12, 0.27], t(189)=5.29, p=.000), but no significant interaction effect of e-mail use and task interdependency on WLC. So when an individual uses a laptop or e-mail more often, he/she is expected to experience more WLC, regardless of task interdependency.
Figure 4 Moderated relation of e-mail use Figure 5 Moderated relation of social media
and WLC. use and WLC.
Note: N=203. *significant at p<.05. Note: N=203. *significant at p<.05.
Table 4
The conditional direct effect of social media use on WLC, dependent on task interdependency
Task interdependency b t p LLCI ULCI
low -.02 -0.46 .649 -0.13 0.08
average .06 1.48 .140 -0.02 0.14
high .14* 2.55 .012 0.03 0.25
Note: N=203. *significant at p<.05.
Social media use (see figure 5) had no significant direct effect on WLC, however there was a significant interaction of social media use and task interdependency on WLC (b=0.07, CI 95% [0.01, 0.14], t(189)=2.19, p=.030), which again implies that the effect of social media on WLC is conditional (see table 4). The effect of social media use on WLC was only significant in high task interdependency settings (b=0.14, CI 95% [0.03, 0.25], t(189)=2.55,
p=.012). In average (b=0.06, CI 95% [-0.02, 0.14], t(189)=1.48, p=.140) and low (b=-0.02, CI 95% [-0.13, 0.08], t(189)=-0.46, p=.649) task interdependency settinges, no significant effect of social media use on WLC was present.
The second model tested for hypothesis 1 used exhaustion as the dependent variable, CTU as independent variable, WLC as a mediator and task interdependency as moderator. This model was significant for smarthpone use (F(13,190)=4.63, p=.000, R2=.27) and explains 27% of the variance in exhaustion (See figure 6). The effect of WLC on exhaustion was significant (b=0.32, CI 95% [0.16, 0.49], t(190)=3.84, p=.000) and while smartphone use did not have a direct effect on exhaustion, there was a conditional indirect effect of smartphone use on exhaustion. This effect was significant when task interdependency was high (b=0.07, CI 95% [0.03, 0.13]) or average (b=0.05, CI 95% [0.02, 0.10]), but not when task interdependency was low (b=0.03, CI 95% [-0.00, 0.07]). The process macro does not provide p values for indirect effects, but a 95% confidence interval that does not include zero, also implies significance at the 5% level. Thus, for smartphone use there was an indirect effect on exhaustion through WLC when task interdependency is high or average.
Figure 6 Moderated indirect relation of smartphone use and exhaustion Note: N=203. *significant at p<.05.
The model including Laptop use (F(13,190)=4.48, p=.000, R2=.28) was significant and explained 28% of the variance in exhaustion (See figure 7). WLC had a direct effect on exhaustion (b=0.28, CI 95% [0.12, 0.44], t(190)=3.43, p=.001). In contrast to smarthpone use, there was a normal indirect effect of laptop use one exhaustion through WLC, regardless of task interdependency (b=0.03, CI 95% [0.01, 0.07]).
Figure 7 Moderated indirect relation of laptop use and exhaustion Note: N=203. *significant at p<.05.
The model for e-mail use (F(13,190)=4.55, p=.000, R2=.29) (see figure 8), which explained 29% in the variance of exhaustion, showed similarities with the model for laptop use. WLC had a direct effect on exhaustion (b=0.22, CI 95% [0.05, 0.39], t(190)=2.54, p=.012), while e-mail use had an indirect effect on exhaustion (b=0.04, CI 95% [0.01, 0.09]). So, as for laptop use, the indirect effect of e-mail use on exhaustion is not dependend on task interdependency.
Figure 8 Moderated indirect relation of e-mail use and exhaustion Note: N=203. *significant at p<.05.
The model with social media use as the independent variable (F(13,190)=4.97, p=.000, R2=.27) explained 27% of the variance in exhaustion (see figure 9). A direct effect of WLC (b=0.30, CI 95% [0.16, 0.44], t(190)=4.19, p=.000) and a conditional indirect effect of social media use on exhaustion were present. The conditional indirect effect was significant for high task interderpendency (b=0.04, CI 95% [0.01, 0.09]), but not for average (b=0.02, CI 95% [-0.00, 0.05]) or low (b=-0.01, CI 95% [-0.04, 0.03]) task interdependency.
Figure 9 Moderated indirect relation of social media use and exhaustion Note: N=203. *significant at p<.05.
Since smarthpone use and social media use showed a significant indirect effect on exhaustion when task interdependency is high and laptop use and e-mail use showed a significant indirect effect on exhaustion regardless of task interdependency, hypothesis 1, which states that there is a significant indirect relation between CTU and exhaustion through WLC when task interdependency is high, is accepted.
The conditional effect of CTU on WLC dependent on PBMS. To test hypothesis 2; the effect of CTU on WLC is stronger for integrators than for segmentors, a regression model with CTU as independent variable, PBMS as moderator and WLC as dependent variable was used. This model was significant for smartphone use(F(14,189)=9.24, p=.000, R2=.44) (see figure 10). 44% of the variance in WLC can be explained by these variables. Smarthpone use had a positive effect on WLC (b=0.17, CI 95% [0.10, 0.24], t(189)=4.08, p=.000), while PBMS had a negative direct effect on WLC (b=-0.30, CI 95% [-0.42, -0.17], t(189)=-5.65, p=.000). The interaction of smartphone use and PBMS on WLC was not significant. This means that while segmentors are expected to experience less WLC, the strength of the effect of smartphone use on WLC is equal for segmentors and integrators.
Figure 10 Moderated relation of Figure 11 Moderated relation of laptop smartphone use and WLC use and WLC
Note: N=203. *significant at p<.05. Note: N=203. *significant at p<.05.
The model for laptop use showed similarities with the model for smartphone use and was significant (F(14,189)=4.62, p=.000, R2=.27). The model explained 27% of the variance in WLC (see figure 11). Laptop use had a positive effect on WLC (b=0.11, CI 95% [0.02, 0.19], t(189)=2.43, p=.016), while PBMS had a negative direct effect on WLC (b=-0.35, CI 95% [-0.50, -0.21], t(189)=-4.77, p=.000). The interaction of laptop use and PBMS on WLC was not significant. This means that while segmentors are expected to experience less WLC, the strength of the effect of laptop use on WLC is equal for segmentors and integrators.
The model was significant for e-mail use (F(14,189)=6.45, p=.000, R2=.36) and explained 36% of the variance in WLC (see figure 12). E-mail use had a positive effect on WLC (b=0.20, CI 95% [0.12, 0.28], t(189)=5.08, p=.000), while PBMS had a negative direct effect on WLC (b=-0.33, CI 95% [-0.47, -0.19], t(189)=-4.59, p=.000). The interaction of e-mail use and PBMS on WLC was not significant. So as for smartphone use and laptop use, segmentors are expected to experience less WLC but the strength of the effect of e-mail use on WLC is equal for segmentors and integrators.
Figure 12 Moderated relation of e-mail Figure 13 Moderated relation of social
use and WLC media use and WLC
Note: N=203. *significant at p<.05. Note: N=203. *significant at p<.05.
Table 5
The conditional direct effect of social media use on WLC, dependent on PBMS
PBMS b t p LLCI ULCI
low .15* 2.73 .007 0.04 0.26
average .05 1.21 .228 -0.03 0.12
high -.05 -0.92 .360 -0.17 0.06
Note: N=203. *significant at p<.05.
The model was significant for social media use (F(14,189)=4.40, p=.000, R2=.26) and explained 26% of the variance in WLC (see figure 13). PBMS had a negative direct effect (b=-0.34, CI 95% [-0.46, -0.22], t(189)=-5.65, p=.000), while social media use had no direct effect on WLC. However a significant interaction effect of social media use and PBMS on WLC was present (b=-0.08, CI 95% [-0.15, -0.0.2], t(189)=-2.45, p=.015). This implies that there is a conditional effect of social media use on WLC (see table 5). This effect was significant when PBMS is low (b=0.15, CI 95% [0.04, 0.26], t(189)=2.73, p=.007), corresponding with integration, but not when it was average (b=0.04, CI 95% [-0.03, 0.12], t(189)=1.21, p=.228) or high (b=-0.05, CI 95% [-0.17, 0.06], t(189)=-0.92, p=.360), corresponding with segmentation. This means that segmentors are expected to experience less WLC because of a direct effect, and that the strength of the effect of social media use on WLC is smaller for segmentors than for integrators. Based on these results hypothesis 2, the effect of CTU on WLC is stronger for integrators than for segmentors, is rejected, with the remark that hypothesis 2 held for social media.
The conditional effect of WLC on exhaustion dependent on PBMS. To test Hypothesis 3; the effect of WLC on exhaustion is stronger for segmentors than for integrators, a regression model with WLC as independent variable, PBMS as moderator and exhaustion as dependent variable (F(14, 189)=4.96, p=.000, R2=.30) was employed (See figure 14). Both PBMS (b=0.35, CI 95% [0.20, 0.450], t(189)=4.49, p=.000) and WLC (b=0.26, CI 95% [0.10, 0.42], t(189)=3.23, p=.000) had a positive effect on exhaustion. Meaning that experiencing WLC or being a segmentor leads to more exhaustion. Moreover, there was a significant interaction effect of PBMS and WLC on exhaustion (b=0.13, CI 95% [0.02, 0.24], t(189)=2.37, p=.002). This means that the effect of WLC on exhaustion varied with different values for PBMS (See table 6). When PBMS was low (b=0.09, CI 95% [-0.13, 0.31], t(189)=0.79, p=.431), corresponding with integration, the effect of WLC on exhaustion was smaller than when PBMS was average (b=0.26, CI 95% [0.10, 0.42], t(189)=3.23, p=.000), which was smaller than when PBMS was high (b=0.43, CI 95% [0.23, 0.62], t(189)=4.29, p=.000), corresponding with segmentation. The conditional direct effect of WLC on exhaustion is significant for segmentors but not for integrators. Hence, hypothesis 3, the effect of WLC on exhaustion is stronger for segmentors than for integrators, is accepted.
Figure 14 Moderated relation of WLC and exhaustion Note: N=203. *significant at p<.05.
Table 6
The conditional direct effect of WLC on exhaustion, dependent on PBMS
PBMS b t p LLCI ULCI
low .09 0.79 .431 -0.13 0.31
average .26* 3.23 .002 0.10 0.42
high .43* 4.29 .000 0.23 0.62
The conditional indirect effect of CTU on exhaustion dependent on task interdependency and PBMS. When combining these results, a model with CTU as dependent variable, WLC as a mediator, exhaustion as a dependent variables, and depending on the technology, task interdependency and/or PBMS as (a) moderator(s) emerged. This model, which is depicted in figure 15 was significant for smartphone use (F(14, 189)=4.86, p=.000, R2=.29) and explained 29% of the variance in exhaustion. The effect strengths differ slightly from the models depicted earlier. The reason being that the moderator variables task interdependency or PBMS were entered as covariates when they were not used as independent variables in previous instances. Since both fullfill the role of moderator in this model, they could not be entered as covariate. In this current model smartphone use had as positive effect on WLC (b=0.16, CI 95% [0.08, 0.24], t(190)=4.07, p=.000). Furthermore, there was significant interaction effect of task interdependency and smartphone use (b=0.08, CI 95% [0.02, 0.14], t(189)=2.46, p=.015) on WLC. For conditional direct effects of smarthpone use on WLC, see table 3. WLC had a positive effect on exhaustion (b=0.30, CI 95% [0.13, 0.46], t(189)=3.55, p=.001), as did PBMS (b=0.35, CI 95% [0.20, 0.51], t(189)=4.54, p=.000). Moreover, there was a significant interaction effect of WLC and PBMS on exhaustion (b=0.14, CI 95% [0.03, 0.25], t(189)=2.52, p=.013). For the conditional direct effect of WLC on exhaustion, see table 6. The indirect effect of smartphone use on exhaustion was thus dependent on two moderators. For an overview of the conditional indirect effect of smartphone use on exhaustion see table 7. The conditional indirect effect was only significant when both task interdependency and PBMS were atleast average and was strongest when both task interdependency and PBMS were high.
Figure 15 Moderated indirect effect of smartphone use on exhaustion Note: N=203. *significant at p<.05.
The conditional indirect effect of smarthpone use on exhaustion. Low task interdepdency Average task interdependenct High task interdependency PBMS b LLCI ULCI b LLCI ULCI b LLCI ULCI Low .01 -0.00 0.04 .02 -0.01 0.10 .03 -0.02 0.09 Average .02 -0.01 0.06 .05* 0.02 0.09 .07* 0.03 0.13 high .04 -0.01 0.11 .08* 0.03 0.14 .11* 0.05 0.20 Note: N=203. *significant at p<.05.
The combined model for laptop use is depicted in figure 16, was significant (F(15, 188)=4.65, p=.000, R2=.30) and explained 30% of the variance in exhaustion. For previously stated reasons, slight fluctuations in effect strength occur when compared to earlier depicted models. Laptop use had a positive effect on WLC (b=0.14, CI 95% [0.06, 0.22], t(189)=3.58, p=.000). WLC had a positive effect on exhaustion (b=0.26, CI 95% [0.09, 0.42], t(188)=3.15, p=.002), as did PBMS (b=0.35, CI 95% [0.19, 0.51], t(188)=4.42, p=.000). Moreover, there was a significant interaction effect of WLC and PBMS (b=0.13, CI 95% [0.02, 0.24], t(188)=2.35, p=.020) on exhaustion. For the conditional direct effect of WLC on exhaustion see table 6. Logically following, the indirect effect of laptop use on exhaustion was conditional as well (see table 8). When PBMS was high (b=0.06, CI 95% [0.02, 0.12], corresponding with segmentation, or average (b=0.04, CI 95% [0.01, 0.08] this effect was significant. When PBMS was low (b=0.01, CI 95% [-0.02, 0.05], corresponding with integration, this effect was not significant.
Figure 16 Moderated indirect effect of laptop use on exhaustion Note: N=203. *significant at p<.05.
Table 8
The conditional indirect effect of laptop use on exhaustion, dependent on PBMS
low .01 -0.02 0.05
average .04* 0.01 0.08
high .06* 0.02 0.12
Note: N=203. *significant at p<.05.
The combined model for e-mail use is depicted in figure 17, was significant (F(15, 188)=4.69, p=.000, R2=.30) and explained 30% of the variance in exhaustion. E-mail use had a direct effect on WLC (b=0.22, CI 95% [0.15, 0.30], t(189)=6.17, p=.000). WLC had a direct effect on exhaustion (b=0.23, CI 95% [0.07, 0.40], t(188)=2.76, p=.007), as did PBMS (b=0.36, CI 95% [0.20, 0.51], t(188)=4.54, p=.000). In addition, there was a significant interaction effect of WLC and PBMS on exhaustion (b=0.13, CI 95% [0.02, 0.24], t(188)=2.31, p=.022). For the conditional direct effect of WLC on exhaustion see table 6. Logically following, the indirect effect of e-mail use on exhaustion was conditional as well (see table 9). When PBMS was high (b=0.09, CI 95% [0.04, 0.16], corresponding with segmentation, or average (b=0.05, CI 95% [0.02, 0.10] this effect was significant. When PBMS was low (b=0.02, CI 95% [-0.03, 0.07], corresponding with integration, this effect was not significant.
Figure 17 Moderated indirect effect of e-mail use on exhaustion Note: N=203. *significant at p<.05.
Table 9
The conditional indirect effect of e-mail use on exhaustion, dependent on PBMS
PBMS b LLCI ULCI
low .02 -0.03 0.07
average .05* 0.02 0.10
high .09* 0.04 0.16
The combined model for social media use is depicted in figure 18 (F(14, 189)=5.27, p=.000, R2=.30) and explained 30% of the variance in exhaustion. At hypothesis 1 and 2 it became clear that the direct effect of social media on WLC was not significant (b=0.05, CI 95% [-0.02, 0.12], t(190)=1.28, p=.201). However social media use combined with task interdependency (table 4) or PBMS (table 5) did yield some significant effects. When these different moderators are combined in one model, both the interaction effect of social media and task interdependency on WLC (b=0.08, CI 95% [0.01, 0.14], t(189)=2.36, p=.019) and the interaction effect of social media and PBMS on WLC (b=-0.08, CI 95% [-0.15, -0.02], t(189)=-2.60, p=.010) remain significant. Table 10 shows an overview of the direct effect strength of social media on WLC for different combinations of PBMS and task interdependency; when task interdependency gets higher and when PBMS gets lower, the effect strength becomes stronger. WLC had a positive effect on exhaustion (b=0.26, CI 95% [0.11, 0.40], t(189)=3.51, p=.001), as did PBMS (b=0.35, CI 95% [0.20, 0.51], t(189)=4.54, p=.000). Moreover, there was a significant interaction effect of WLC and PBMS (b=0.15, CI 95% [0.04, 0.26], t(189)=2.69, p=.008) on exhaustion. For the conditional direct effect of WLC on exhaustion, see table 6. The indirect effect of social media use on exhaustion was thus dependent on two moderators. The direct effect of social media on WLC was moderated by task interdependency and PBMS and the direct effect of WLC on exhaustion was moderated by PBMS. An important remark is that the two moderating effects of PBMS are in opposite direction. Being an integrator strengthens the effect of social media use on WLC, but weakens the effect WLC on exhaustion. Being a segmentor weakens the effect of social media use on WLC, but strengthens the effect of WLC on exhaustion. Hence only when PBMS was average, a significant effect was found. For an overview of the conditional indirect effect of social media use on exhaustion see table 11.
Figure 18 Moderated indirect effect of e-mail use on exhaustion Note: N=203. *significant at p<.05.
Table 10
The conditional direct effect of social media use on WLC. Low task interdepdency Average task interdependency High task interdependency
PBMS b LLCI ULCI b LLCI ULCI b LLCI ULCI
Low .06 -0.06 0.20 .15* 0.05 0.26 .24* 0.11 0.38 Average -.04 -0.14 0.07 .05 -0.03 0.13 .14* 0.03 0.24 high -.15* -0.28 -0.01 -.06 -0.17 0.05 .03 -0.11 0.16 Note: N=203. *significant at p<.05.
Table 11
The conditional imdirect effect of social media use on exhaustion. Low task interdepdency Average task interdependency High task interdependency
PBMS b LLCI ULCI b LLCI ULCI b LLCI ULCI
Low .00 -0.01 0.04 .01 -0.02 0.05 .02 -0.03 0.07 Average -.01 -0.05 0.02 .01 -0.01 0.04 .03* 0.01 0.07 high -.06* -0.15 -0.00 -.03 -0.09 0.02 .01 -0.05 0.08 Note: N=203. *significant at p<.05
Discussion Conclusion and theoretical contributions
The research problem that the current study is addressing is that CTU leads to WLC and subsequently exhaustion in some situations but not in others. The aim of this research is therefore to determine conditions under which CTU leads to WLC and exhaustion. This research paper helps to locate employees who are at risk for experiencing WLC and exhaustion as a result of CTU. By locating these individuals, organizations become enabled to administer WLC and exhaustion counter measures such as, but not limited to: reducing work pressure, restricting CTU, and training in time management or stress reduction. In addition, locating these individuals can help in the forming of new organizational policies around CTU and shaping organizational culture.
One of the most important conclusions of this study is a division in the work force based on PBMS. Person-job fit theory (Kristof-Brown et al., 2005) and boundary management
theory (Ashfort et al., 2000) were used to hypothesize a stronger relationship between WLC and exhaustion for segmentors than for integrators. The analyses performed in this research paper support this notion. A significant interaction effect of WLC and PBMS on exhaustion was found. This interaction effect is the first theoretical contribution this study makes. This finding provides support for the integrator-segmentor division prevalent in boundary management theory and substantiates person-job fit theory. Person-job fit theory states that a misfit between an employee’s needs and desires and the actual job can result in exhaustion. Previous research focused on fits such as between an employee’s knowledge and abilities and the knowledge and abilities a job requires, or between an employee’s needs, for example personal development, and the resources available to fulfill this need (e.g. Kristoff-brown et al., 2005). To my knowledge the current study is the first that determines PBMS to be a need/desire that needs to be met to prevent exhaustion. For segmentors there is a strong relationship between WLC and exhaustion. Implying that for segmentors in particular, it is important to minimize WLC. Interestingly, integrators don’t seem to be at risk of WLC induced exhaustion at all, no significant relation between WLC and exhaustion exists for integrators. However, concluding that WLC does not have any negative effects for integrators would be premature, since WLC has also been related to higher turnover intentions, absenteeism, reduced performance and lower organizational commitment (Boles, Johnson, & Hair, 1997; Bond, Galinsky, & Swanberg, 1998; Kossek & Ozeki, 1998; Thompson, Beauvais, & Lyness, 1999). Future research will have to determine if integrators experience other negative effects as a result of WLC. Nonetheless, this study does imply that integrators are not at risk of exhaustion and burnout as a result of WLC.
Even though for segmentors the relation between WLC and exhaustion is stronger than for integrators, segmentors do not seem to be able to manage their technology use in a way that minimizes WLC. No significant interaction effects between smartphone use, laptop use or e-mail use, and PBMS on exhaustion were found in this study. This is an important theoretical contribution because it falsifies the assumption that PBMS is a factor that shapes the management of technology use in a way that an employee’s CTU corresponds with an employee’s PBMS. The assumption was based on research showing segmentors to experience less WLC and integrators to experience more WLC (Olson-Buchanan & Boswell, 2006; Park & Jex, 2011). Even though PBMS does not moderate the relation between CTU and WLC, the direct effect of PBMS on WLC found in this study does correspond with the findings in Olson-Buchanan and Boswell (2006) and in Park and Jex (2011). This implies that employees have rather limited control over their CTU, but are able to limit WLC, to a certain extent, in
other ways. The direct effect of PBMS on WLC however, is not strong enough to mitigate the strengthened effect of WLC on exhaustion. This is particularly problematic for segmentors, who suffer rather extreme effects of WLC, but seem to possess only limited measures to prevent WLC. The exception is social media use, which showed a significant interaction effect with PBMS on WLC. Only integrators seem to experience WLC as a result of social media use.
The third important conclusion of this research paper is that not all CTU affects WLC in the same way. Laptop use and e-mail use affect WLC regardless of task interdependency. The effect strength of smartphone use and social media use on WLC increases when task interdependency rises. The premise that task interdependency would strengthen the effect of CTU on WLC was based on a qualitative study that showed that the use of mobile e-mail devices in high task interdependency settings could lead to decreased autonomy in deciding when and where to work (Mazmanian et al., 2013). This study shows that this premise holds in quantitative research for smartphone use and social media use, but not for e-mail use and laptop use. The current study further adds to existing literature by nuancing previously found effects of smartphone use and social media use on WLC (Van Zoonen et al., 2016; Derks & Bakker, 2014; Yun et al., 2014), by determining conditions under which this effect is present and absent.
Concluding, the relation between WLC and exhaustion is strongest for segmentors, which means segmentors are at greater risk for CTU induced exhaustion than employees who score average on PBMS. While integrators might still experience CTU induced WLC, they are not at risk for CTU induced exhaustion at all, because for them WLC does not incur exhaustion. Organizations that are aiming to limit CTU induced exhaustion should thus focus their effort accordingly. Importantly, PBMS does not moderate the relation of CTU and WLC for smartphone use, laptop use or e-mail use. Implying that how an employee wants to manage the boundary between the ‘work’ and ‘life’ domain, does not influence the actual management of this boundary regarding CTU. Different technologies have different effects on WLC. In average task interdependency settings, e-mail use leads to the highest amounts of WLC, sequentially followed by smartphone use, laptop use and social media use. Task interdependency strengthens the effect of smartphone use and social media use, but not of laptop use and e-mail use. In high task interdependency settings, using a smartphone leads to more WLC than using e-mail does. In low task interdependency settings smartphones can be used without the risk of WLC, whereas social media can be used without the risk of WLC in low and average task interdependency settings. The current study thus shows that both PBMS
and task interdependency are moderators on the indirect relation of CTU and exhaustion through WLC and are thus conditions under which CTU can lead to WLC and exhaustion. The following section will discuss the practical implications of these findings.
Practical implications
This study shows that CTU can lead to exhaustion for parts of the workforce. When organizations want to minimize CTU induced exhaustion, they should focus their efforts on employees with average to high PBMS scores. Employees with a low PBMS score can use communication technology without limits. They might experience WLC, but are not at risk for CTU induced exhaustion. Especially for employees with a high PBMS score, and to a lesser extent employees with an average PBMS score, it is important to create circumstances in which they are able to adequately manage the boundary between the ‘work’ and the ‘life’ domain in such a way that it corresponds with their PBMS. The control variable BMN offers insight in one way of achieving this. When an organization is able to create a culture in which segmentation of the ‘work’ and ‘life’ domain is the norm, their workforce as a whole will experience less WLC.
Another important way of reducing WLC for employees who score at least average on PBMS, is reducing task interdependency. Though this will only be effective in organizations where employees commonly use smartphones and social media. Due to the pervasiveness, and in some cases necessity, of these technologies, reducing task interdependency instead of limiting technology use should be a primary strategy. This can be achieved through knowledge sharing and structuring jobs in such a way that expertise and resources are redundant. For example, when an application manager running the organizational electronic system is the only one with the resources and knowledge to respond to problems, colleagues will be greatly dependent on this person. Hence, the application manager will receive many work demands, even when he/she is not working. When this responsibility is shared by more people, the demands will be divided among multiple employees and demands in personal time will be less common because there is a greater chance that at least one of the employees with knowledge and resources to respond to problems is working.
In some instances, reducing task interdependency will be unfeasible. In other instances, such as for organizations in which employees primarily use e-mail and laptops, reducing task interdependency will only have a limited effect. In these cases, limiting technology use of employees who score at least average on PBMS is another way of reducing their WLC and subsequently exhaustion. Organizations can do this in several ways. A rather
drastic option is limiting or forbidding CTU when these employees are not officially working. Though this option might not be very popular, it will probably have a rather strong effect. Hence, for organizations experiencing a gravely exhausted workforce this could be a solution. Another option is limiting or forbidding technology use for employees who are especially exhausted and/or close to a burnout. This is a targeted solution to prevent exhaustion and burnout, but makes constantly monitoring employees on signs of exhaustion a necessity.
It is important to understand that an exhausted employee is often the result of several different causes, some of which are work pressure, atmosphere on the work floor and thus CTU and WLC. There might be instances in which reducing CTU or WLC is infeasible. In these cases, addressing other causes for exhaustion, for instance reducing work pressure, or employing more general measures such as time management training or stress reduction training, could be effective.
To conclude, it is important to create circumstances in which employees who have an average to high PBMS score are able to limit their experienced WLC. Because of the rather extreme effect of WLC on exhaustion that segmentors will experience, failing to create these circumstances will result in a partially exhausted work force. This research paper suggests that limiting technology use or decreasing task interdependency are viable ways of achieving a decrease in WLC. Respecting the PBMS of employees is of the utmost importance.
Future research
Several findings in this study call for more research to be understood. In this study, no significant interaction of CTU and PBMS was found. However, a non-hypothesized direct relation between PBMS and WLC was found. Interpreting p-values of non-hypothesized effects can be problematic. Because a p-value represents the chance to obtain this specific sample if the null-hypothesis is true, running a great number of tests will always produce a few significant effects. Therefore p-values can only be used to test pre-analysis formed hypotheses. However, a direct effect of PBMS on WLC corresponds with earlier research (Olson-Buchanan & Boswell, 2006). In addition, the p-value for this effect was extremely low. Therefore, I deem it likely that there is an effect of PBMS on WLC. Since there is a direct effect, segmentors do experience less WLC than integrators, however this difference in experienced WLC can not be explained by a weaker effect of CTU on WLC. This implies that segmentors are able to reduce the experienced WLC, but that they do not achieve this by managing their technology use. Future research should investigate how segmentors alleviate
