Crafting behaviours and employees’ and their partners’ well-being: a weekly study

ABSTRACT

Organizations are increasingly aware of the relevance of employees’ well-being, due to its positive impact on both companies and society. Based on the Work-Home Resources (W-HR) model, this study aims to analyse the relationship between two resource-gaining behaviours – i.e., expansion-oriented job crafting (JC) and leisure crafting (LC) behaviours- and employees’ and their partners’ well-being. A quantitative longitudinal study using weekly online questionnaires (for four weeks) was conducted with 50 participants and their partners. Results of a multilevel sequential mediation model with random slopes provided empirical support for the three-path sequential mediation model (weekly expansion-oriented JC ➔ weekly LC ➔ employees’ weekly well-being ➔ partners’ weekly well-being) when contemporaneous effects were considered, but not when time-lagged effects were considered. Thus, this study shows that work and leisure domains can be integrated – i.e., positively related- rather than segmented – and that both types of crafting behaviours positively contribute to higher employee well-being in the weeks when more crafting behaviours occur. In addition, when employees’ well-being increases from one week to the next, the well-being of their partners also increases. Taken together, these results suggest that crafting is a resource that contributes not only to the employees’ well-being, but also to that of their partners.

KEYWORDS:

• Job crafting
• leisure crafting
• well-being
• spillover

Employees’ well-being is increasingly important to organizations, due to its impact on their performance and competitiveness and on society as a whole (Robledo et al., 2019). Their well-being is especially relevant if we consider that it might be transferred to other people in their close environment (for example, to their partner), thus multiplying its effects (Westman, 2001).

Based on the COR theory (Hobfoll, 2001), the Work-Home Resources (W-HR) model (ten Brummelhuis & Bakker, 2012) explains how resources are interconnected across work and non-work domains. Employees develop different personal resources – i.e., a set of resources that include physical (e.g., vigour), psychological (e.g., self-efficacy), intellectual (e.g., skills), affective (e.g., mood), and capital (e.g., time) resources (ten Brummelhuis & Bakker, 2012)- to foster goal achievement. These personal resources are relevant because organizational resources are generally scarce (e.g., Rotondo et al., 2008). Then, through a gain spiral, these personal resources generate new resources that spill over from the work domain to the non-work domain.

In this study, we analyse job crafting (JC) and leisure crafting (LC) as proactive behaviours that build a pool of useful resources for work and non-work domains. Whereas JC builds resources in the work domain, LC builds resources in the leisure domain. On the one hand, JC is defined as “a proactive employee behavior consisting of seeking resources (i.e., asking managers or colleagues for advice) and seeking challenges (i.e., asking for more responsibilities)” (Petrou et al., 2012, p. 1122). The goal of seeking resources and challenges is to accumulate external and internal resources that enable employees to grow and find meaning (Petrou et al., 2012). Thus, through JC behaviours, employees increase their personal resources and their sustainable work abilities (e.g., van den Heuvel et al., 2015). On the other hand, LC is defined as “the proactive pursuit of leisure activities targeted at goal setting, human connection, learning, and personal development” (Petrou & Bakker, 2016, p. 508). Based on this definition, individuals who engage in LC will obtain resources such as greater self-efficacy in achieving the objectives, a wider social network, or more knowledge and skills development.

The relationship between JC and LC is suggested to be based on a positive spillover process of motives, efforts, and outcomes between work and non-work domains (de Bloom et al., 2020). The spillover process (also called generalization) is defined as “similarity between a construct in the work domain and a distinct but related construct in the family domain” (Edwards & Rothbard, 2000, p. 180). In our case, the spillover relates the work domain and a broader non-work domain to consider the leisure domain. The rationale behind the spillover process is that individuals are consistent in their behaviours across domains according to their personal values, skills, and needs. In this regard, previous studies have shown a positive relationship between needs satisfaction at work and at leisure (e.g., Walker & Kono, 2018), job satisfaction has been related to family satisfaction (e.g., Gutek et al., 1988), and work values have been linked to family values (e.g., Piotrkowski, 1979). Based on the W-HR model (ten Brummelhuis & Bakker, 2012), we suggest that there is a positive spillover of crafting behaviours – i.e., between JC in the work domain and LC in the non-work domain. Moreover, we predict that both crafting behaviours engage in a resource gain spiral that contributes to employees’ well-being. We also go a step further by studying the positive spillover from employees’ well-being to their partners’ well-being at home.

There is evidence that job crafting behaviours change over time and situations, and that intra-individual variability in job crafting (within-subjects variance) is related to intra-individual variability in relevant outcomes such as work engagement (Bakker & Oerlemans, 2019), work performance (Petrou & Xanthopoulou, 2021), work enjoyment (Tims et al., 2014), or meaningful work (Hulshof et al., 2020). In this study, we focus on within-person weekly fluctuations in JC and LC behaviours because both are dynamic behaviours that may fluctuate over weeks (Petrou et al., 2017). For instance, in the case of LC, individuals usually participate in leisure activities –e.g., sports, cultural events, voluntary work, etc.- once or several times a week, but not every day (Hubbard & Mannell, 2001). Similarly, individuals can also perform JC on a weekly basis -e.g., changing tasks or participating in training. Specifically, we propose that weekly fluctuations in JC levels contribute to explaining the weekly well-being of employees and their partners through a weekly spillover from a work (job-crafting) domain to a non-work (leisure crafting) domain. The proposed research model is shown in Figure 1.

Figure 1. Proposed conceptual model of the study.

This study contributes to the W-HR model (ten Brummelhuis & Bakker, 2012) by analysing the dynamic relationship between two proactive behaviours -JC and LC- and employees’ and their partners’ well-being. Specifically, we aim to provide evidence of the positive spillover relationship between JC and LC. Second, we provide initial empirical evidence for the mechanisms through which JC is related to the well-being of employees and their partners. By doing so, the positive crossover is highlighted, moving beyond the negative crossover (e.g., distress) traditionally analysed in the literature (Hobfoll et al., 2018). We explicitly focus on flourishing as an indicator of well-being because it allows us to assess the degree to which individuals achieve meaning and purpose in their lives, which would be the main aim of JC and LC behaviours.

Expansion-oriented job crafting

JC is a proactive behaviour designed to improve the fit between work and employees’ needs and motivations (Wrzesniewski & Dutton, 2001). Therefore, JC is an ideal tool to increase the meaning of employees’ work (Wrzesniewski et al., 2013) –i.e., when employees derive meaning from a work activity to find purpose in life (Ignelzi, 2000) and engagement – high involvement in relevant and valued activities- (Rana, 2015). The Job Demands- Resources (JD-R) theory (Bakker & Demerouti, 2007, 2014) operationalizes the concept and defines JC by focusing on job characteristics such as “the changes that employees may make to balance their job demands and job resources with their personal abilities and needs” (Tims et al., 2012, p. 174). Job demands have physical, psychological, or social work characteristics that involve physical or psychological costs, whereas job resources are job characteristics that lead to goal achievement, reduce job demands, and contribute to personal growth, learning, and development (Bakker & Demerouti, 2007). Therefore, JC is defined as seeking resources, seeking challenges, and reducing demands (Petrou et al., 2012). Specifically, seeking resources involves certain behaviours, such as asking for feedback, improving interpersonal communication, or developing new skills. Seeking challenges includes behaviours such as asking for more responsibilities after finishing assigned tasks or expanding the scope of job responsibilities. Finally, reducing demands includes behaviours designed to minimize the demanding aspects of the job (Demerouti & Bakker, 2014). Csikszentmihalyi and Nakamura (1989) argued that employees seek challenges in order to grow professionally while maintaining motivation and avoiding boredom. This definition is consistent with the expansion-oriented vs. contraction-oriented JC approach proposed by Laurence (2010). Expansion-oriented JC consists of increasing the number or complexity of task and social interactions, whereas contraction-oriented JC consists of reducing task and social interactions. Relating the two perspectives, seeking resources, and seeking challenges would be expansion-oriented JC, and reducing demands would be contraction-oriented JC (e.g., Demerouti et al., 2020). Whereas expansion-oriented JC proactively promotes individuals’ outcomes and focuses on meaning-making, contraction-oriented JC can have a detrimental effect on individuals’ well-being (e.g., Rofcanin et al., 2019). A similar suggestion was also proposed by Zhang and Parker (2018), who identified two crafting perspectives: approach-oriented vs. avoidance-oriented JC. Approach-oriented JC consists of expanding the job boundaries, whereas avoidance-oriented JC consists of reducing the job boundaries. Because we are interested in proactive behaviours that promote employees’ well-being, we focus on expansion-oriented JC.

Leisure crafting

Individuals not only craft their work, but they also craft other life domains, such as leisure (Petrou & Bakker, 2016). As in the case of JC, LC is a tool that can increase the meaning of employees’ non-work and work engagement (Iwasaki et al., 2018). For example, leisure induces meaning because it promotes and maintains a joyful and connected life. It also helps to discover individuals’ unique talents and characteristics, provides them with control and autonomy to liberate opportunities to change or adjust their lives, and enhances a sense of empowerment (Iwasaki et al., 2018). Therefore, just as employees use JC to redesign the tasks and relational boundaries of their jobs (Wrzesniewski & Dutton, 2001), they use LC to redesign the tasks (e.g., looking for new challenges) and relational (e.g., building new and inspiring relationships) boundaries of their leisure activity (Tsaur et al., 2020). Moreover, LC allows them to take advantage of leisure activities that involve proactivity, intention (e.g., Fritsch et al., 2005), control (Sonnentag & Fritz, 2007), goal setting, human connection, learning, and personal development (Snir & Harpaz, 2002), which are relevant personal resources that promote employees’ well-being (Iwasaki et al., 2018).

Hence, when individuals engage in LC, they behave in a proactive way and with intention. Moreover, through LC, individuals learn new things and develop in order to improve their feelings of control. Finally, LC experiences involve fellowship and the creation and strengthening of interpersonal relationships.

The spillover between weekly expansion-oriented JC and LC

Because the changing nature of work –e.g., more technology at work, more flex-time and flex-space work- has blurred the boundaries between work and non-work, there is a growing interest in the relationship between JC behaviours and other crafting behaviours in other domains (e.g., de Bloom et al., 2020; Demerouti et al., 2020).

From a spillover perspective, in this paper, we propose that individuals would generalize crafting behaviours in the work and leisure domains. Different arguments encourage us to consider this spillover effect between the work and leisure domains. For instance, the Integrative Needs Model (INM) of crafting behaviour (de Bloom et al., 2020) suggests that an individual’s identity in the work and non-work domains shares the crafting of motivation, effort, and outcomes. Moreover, the W-HR model (ten Brummelhuis & Bakker, 2012) posits that the non-work domain can benefit from resources –e.g., the skills, positive affect, and psychosocial capital- obtained from the work domain (ten Brummelhuis & Bakker, 2012). Bakker (2015) argued that employees who perform JC behaviours are likely to increase their job resources, leading to a gain spiral of resources that would motivate the generation of resources in other domains such as leisure. These resources would be created through LC behaviours. In line with this rationale, Hu et al. (2019) recently provided a perspective that integrates JC and the COR theory and highlights the role of JC in generating not only work resources (e.g., person-job fit), but also psychological resources, such as meaningfulness (Tims et al., 2016) and needs satisfaction (Deci & Ryan, 2000). These are energy resources that motivate individuals to put more effort into their crafting actions – e.g., in either work or leisure time- in order to achieve a meaningful life (Hu et al., 2019). Using the same reasoning, the literature on proactive behaviours at work suggests that positive affective states –i.e., energy resources- contribute to both initiating and sustaining proactive behaviours in either domain (Parker et al., 2010). Therefore, for instance, when employees engage in JC, positive emotions and attitudes arise (N. P. Podsakoff et al., 2007), making them particularly motivated to strive and sustain this proactive behaviour in the non-work domain through LC.

Despite this theoretical rationale for the spillover of crafting behaviour, little previous research has provided empirical evidence about the positive relationship between JC and crafting behaviours in other life domains. For instance, Demerouti et al. (2020) showed that JC and home crafting are related on a daily basis. Specifically, they found a spillover of expansion-oriented JC –i.e., seeking resources and seeking challenges- into the home domain. In the present paper, we follow previous research on the positive relationship between crafting behaviours in work and non-work domains by showing evidence of the spillover effect of crafting behaviours. Specifically, we hypothesize that:

Hypothesis 1.

Employees’ weekly expansion-oriented JC is positively related to employees’ weekly LC.

Expansion-oriented JC, LC, and well-being

The most studied outcomes in the JC literature have been work-related outcomes –e.g., work engagement, burnout, and job performance-, whereas the main outcomes in non-work crafting have focused on general well-being (de Bloom et al., 2020). Because we are interested in the spillover between work and non-work, we analyse employees’ well-being. We rely on Diener et al. (2010) concept of flourishing to define well-being as positive human functioning in relevant areas such as relationships, self-esteem, purpose, and optimism. Well-being has been used extensively because it predicts other positive outcomes in a wide variety of life domains (Gravador & Teng-Calleja, 2018), including the perception of employees’ quality of life or, in other words, how people feel about their existence.

Key resources for well-being are provided when individuals engage in activities that contribute to their identity or are perceived as meaningful. Therefore, we aim to analyse how two weekly proactive behaviours from different domains – i.e., expansion-oriented JC and LC- are related to achieving a meaningful life – i.e., well-being-. Several theoretical frameworks explain the relationship between crafting behaviours and well-being. Specifically, the identity research suggests that when individuals enhance their identities, they increase their behavioural repertories and, therefore, their well-being (Thoits, 1983, 1986). Moreover, according to the resource investment principle in the W-HR model (ten Brummelhuis & Bakker, 2012), individuals are especially motivated to acquire, protect, and develop their resources, and those who possess more resources are better than others in terms of resource gain. Whereas JC is involved in the gain spiral at work, LC is involved in the gain spiral outside work.

JC is suggested to be a resource gain strategy (Hu et al., 2019). By performing JC, employees not only gain work resources, but also psychological resources, such as meaningfulness and needs satisfaction. These energy resources contribute to a more purposeful and meaningful job and would ultimately be integrated to create the work identity, an important facet of an individual’s identity. In a recent meta-synthesis of qualitative JC research, Lazazzara et al. (2020) justified the positive relationship between JC and esteem-enhanced occupational identity, job satisfaction, and meaningfulness.

The same rationale could be applied to LC. LC may be a strategy to initiate a gain spiral process outside work, where resources are accumulated and generate new resources such as “detachment, relaxation, autonomy, mastery, meaning and affiliation” (de Bloom, 2018). In line with this rationale, Kujanpää, Weigelt, et al. (2021) demonstrated that employees may craft off-job experiences of meaning and affiliation to enhance their vitality, an indicator of well-being. Moreover, despite the gain spiral, resources developed by engaging in LC could also buffer the negative effect of demands –e.g., work and non-work- on exhaustion (Abdel Hadi et al., 2021). It is reasonable to assume that individuals who perform weekly LC behaviours are able to obtain resources- e.g., self-efficacy, social networks- in the non-work domain, allowing them to restore and accumulate resources that might be related to their well-being. Moreover, engaging in LC would take advantage of leisure activities in terms of needs satisfaction and meaning-making, which are psychological energy-driven resources that facilitate the process of creating a meaningful non-work identity (Iwasaki et al., 2018). In fact, the scarce literature on LC has demonstrated the positive relationship between LC and meaning-making (Petrou et al., 2017), which resembles our well-being construct. Therefore, and considering the rationale presented above, we propose:

Hypothesis 2a.

Employees’ weekly expansion-oriented JC is positively related to their weekly well-being.

Hypothesis 2b.

Employees’ weekly LC is positively related to their weekly well-being.

The mediating role of LC in the relationship between expansion-oriented JC and Well-being

According to the W-HR model (ten Brummelhuis & Bakker, 2012), and despite JC’s potential to enhance employees’ well-being (e.g., Lazazzara et al., 2020), we propose that LC plays a role in explaining why expansion-oriented JC behaviours are related to employees’ well-being. First, LC has the potential to allow employees to recover from work, restoring the loss of resources in the work domain. In this regard, recent studies (e.g., Abdel Hadi et al., 2021) have shown that engaging in LC promotes health by counteracting the negative effects of highly demanding work and family environments. Second, LC increases the capacity to accumulate resources, satisfying individuals’ needs beyond work. Evidence of resource accumulation is provided by Kuykendall et al. (2018), who found that people have greater well-being when they engage in leisure activities that fulfil the need not only for detachment-recovery from work, but also for an increase in autonomy, meaning, mastery, and affiliation.

The process of accumulating resources between work and non-work settings was suggested by the W-HR model (ten Brummelhuis & Bakker, 2012), where resource gains at work can trigger a “gain spiral” of personal resources that may help individuals to perform better outside work. Gaining resources increases the resource pool, which makes it more likely that additional resources will subsequently be obtained. Therefore, the resource gain spiral that could start at work would trigger the gain spiral process in the non-work domain, motivating the individual to engage in LC behaviours to recover from work and gain energy resources to take part in leisure activities that lead to achieving purpose in life or psychological well-being. We propose that:

Hypothesis 3.

Employees’ weekly expansion-oriented JC is positively related to their weekly well-being through their weekly LC (mediation hypothesis).

The positive relationship between employees’ well-being and their partners’ well-being

The W-HR model (ten Brummelhuis & Bakker, 2012) explains how work resources generated by expansion-oriented JC and LC behaviours are related to employees’ well-being. This process may not only occur within the individual. Related to the W-HR model (ten Brummelhuis & Bakker, 2012), the COR theory suggests that resources can be transferred to other individuals in the proximal environment (Hobfoll et al., 2018). In this way, both negative and positive experiences of two individuals may be linked. According to the crossover model, Westman (2001) demonstrates that a positive direct crossover may occur when experiences, affective states, and resources are transmitted from one partner to the other. In this regard, it is well known that resources generated at work –e.g., performance, self-esteem, job-related self-efficacy (Neff et al., 2012), and work engagement (Bakker et al., 2005)- can be transferred not only to co-workers, but also to life partners.

Along with the association between weekly expansion-oriented JC and LC and employees’ well-being, we propose that employees’ weekly well-being is related to their romantic partners’ weekly well-being. Several mechanisms could explain this relationship. First, as self-expansion theory suggests (Aron et al., 1991), couples share an intimate relationship that makes them incorporate the other’s resources, perspectives, and identities as a way to expand their self. Second, according to the broaden and build theory (Fredrickson, 2001), positive feelings cross over via empathy. For instance, Sanz-Vergel and Muñoz (2013) found evidence of an indirect crossover effect of daily work enjoyment on the partner’s well-being via the employee’s well-being. Finally, an emotional contagion process (Fredrickson, 2001) may also play a role. Therefore, we hypothesize that:

Hypothesis 4.

Employees’ weekly well-being is positively related to their partners’ weekly well-being.

The mediating role of employees’ well-being in the relationship between Expansion-Oriented JC, LC, and partners’ well-being

Our research model also proposes that both crafting behaviours performed by employees will be positively and indirectly related to the well-being of employees’ partners through employees’ well-being. Lee et al. (2019) showed that resources produced by an organizational intervention–e.g., schedule control- increased partners’ well-being by reducing their level of stress. Although this effect has not been analysed in the LC literature, the research on couples’ leisure activities showed that they foster couples’ well-being (e.g., Claxton & Perry-Jenkins, 2008) by creating a couple identity (Hickman-Evans et al., 2018). That is, sharing an identity as a couple allows both employees and their partners to balance their roles as spouses with their personal needs and identities (Hickman-Evans et al., 2018).

The spillover literature also provides evidence about the positive crossover effect of employees’ work resources on their partners’ well-being. Hammer et al. (2005) showed that employees’ positive spillover perception –i.e., the transmission of positive emotion, affect- decreased their partners’ depression. Similarly, Liu et al. (2016) found that employees’ work-family enrichment –i.e., the perception that resources generated at work increase non-work effectiveness- increased the partners’ well-being through the perception of employees’ support. Recently, Walter and Haun (2020a) showed that employees’ positive work-related thinking benefits both the employees themselves and their partners.

Taken together, the W-HR model (ten Brummelhuis & Bakker, 2012) and the crossover model (Hobfoll et al., 2018) suggest that resources generated by expansion-oriented JC would spill over into non-work, motivating LC. Because both crafting behaviours contribute to the resource gain spiral and to employees’ work and non-work identity, they should be positively related to employees’ well-being (de Bloom et al., 2020). Finally, through a self-expansion (Aron et al., 1991) or emotional contagion (Fredrickson, 2001) process, employees’ well-being should be positively related to their partners’ well-being. Therefore, we hypothesize that:

Hypothesis 5a.

Employees’ weekly expansion-oriented JC is positively related to their partners’ weekly well-being through employees’ weekly well-being (mediation hypothesis).

Hypothesis 5b.

Employees’ weekly LC is positively related to their partners’ weekly well-being through employees’ weekly well-being (mediation hypothesis).

Considering the arguments presented above, our model proposes a three-path sequential mediation model in which individuals’ weekly expansion-oriented JC behaviours (i.e., seeking resources and seeking challenges) are related to the non-work domain in the form of LC behaviours. Weekly LC behaviours, in turn, make individuals more prone to obtaining resources in the non-work domain, allowing them to restore and accumulate resources that influence their weekly well-being, which is ultimately related to their partners’ weekly well-being. Thus, we propose:

Hypothesis 6.

Employees’ weekly expansion-oriented JC is positively related to their partners’ weekly well-being through employees’ weekly LC and weekly well-being (sequential mediation hypothesis).

By testing these hypotheses, we aim to understand the dynamic relationship between the variables in our research model. Nevertheless, these hypotheses refer to contemporaneous “effects” of “X on M1, M1 on M2 and M2 on Y” over time. To obtain stronger evidence of possible causal relationships, we additionally explore whether our hypotheses are supported when investigating lagged associations between the variables involved: for example, whether expansion-oriented JC one week is associated with LC the following week. Because we only collected data for four weeks, for each hypothesis we always link the hypothesized predictor at T_i (starting at Time 1) with the hypothesized outcome at T_{i +1}, with the exception of Hypothesis 2a, where, in order to simultaneously test the full model, employees’ well-being at t + 2 is regressed on JC at t.

Method

Procedure and participants

Employees from various Spanish organizations participated in this study in 2016 between November and December. They were recruited by 50 Master’s degree students who each approached two employees using their personal contacts. Each employee had to provide the student with the email address of their partner in order to send them the link to the weekly online questionnaire. In doing so, each employee had a code to match the weekly questionnaires of their partner. Demerouti and Rispens (2014) suggest that the student-recruited sampling method has several advantages, including heterogeneity of the sample, cost reduction, elaborate research designs, and student learning, if conducted carefully. In the present study, students acquired data from working men and women from different professional backgrounds, with and without family responsibilities.

Before the data collection, each student received training in the implementation of the weekly research method. Data were collected via online questionnaires. First, participants had to fill in one general questionnaire, and then, together with their partners, they had to fill in one weekly questionnaire at the end of the work week (from Friday afternoon to Sunday at midnight) for four weeks. That is, the questionnaires were sent on Friday afternoon (i.e., between 3 p.m. and 5 p.m.), and the questionnaires were closed on Sunday at midnight. In order to ensure respondents’ anonymity, the system did not allow access to the email addresses from which the questionnaires were sent. Similarly, to make sure that the questionnaires were not responded to after the last day of the week in question, questionnaires were not available after 12 midnight on Sunday. Studies using a similar procedure have demonstrated that it leads to valid responses with sufficient variability and reliability (e.g., Meier et al., 2013).

Of the 100 online questionnaires sent, after four weeks, 72 were returned by employees, and 68 were returned by their partners. However, after linking all the codes of employees and their partners, we were able to keep data from 50 employees and their partners over the four weeks, which resulted in a total of 200 observations. Thus, we were able to use 50 weekly on-line questionnaires (50%) for our study. This rate is acceptable based on the overall response rate for online surveys, which is above 30% (Lindemann, 2021). Of the final 50 participants, 24 were males and 26 females, they had a mean age of 38.60 years (S.D. = 12.53), and they worked a mean of 34.40 hours per week (S.D. = 9.96). Regarding marital status, most of the participants were married or living with their partners (94%). In terms of family burden, 12.1% of the respondents had children between the ages of 0 and 3, 23.5% had children between the ages of 4 and 12, and 47.6% had children older than 13. More than half of the participants had obtained post-secondary degrees (67.3%), 20.4% had completed secondary school, and 12.2% had completed elementary school. In terms of working status, most of the participants were salaried (83.7%), whereas 16.3% were self-employed. Participants had a mean of 15.40 years (S.D. = 13.13) of work experience. We did not collect sociodemographic information about the partners.

Measures

General measures

Before collecting the main variables of this study, participants reported some demographic variables, such as gender, age, marital status, number of children, level of education, work experience in years, and working hours per week.

Weekly measures

The weekly measures were implemented once a week (at the end of the work week) for four consecutive weeks. We modified all the scale items to refer to the week of data collection (e.g., “This week, … ”).

Weekly expansion-oriented job crafting

We used two sub-dimensions of the JC scale developed by Petrou et al. (2012). Specifically, nine items were used to measure seeking resources (6 items; “This week, I have asked my supervisor for advice”) and seeking challenges (3 items; “This week, I have asked for more responsibilities”). Items were rated on a 5-point graded scale (1 = never; 5 = often). Omega reliability coefficients ranged from .78 to .84 across the four time points.

Weekly leisure crafting

We used the nine-item scale developed by Petrou and Bakker (2016). A sample item was “This week, I’ve tried to increase my learning experiences through leisure activities”. Items were rated on a 5-point graded scale (1 = not at all; 5 = very much). Omega reliability coefficients ranged from .91 to .95 across the four time points.

Employees’ and partners’ weekly well-being

We used the eight-item scale developed by Diener et al. (2010) to measure well-being. It is a unidimensional scale, and all the items are positively worded. A sample item was “This week, I’ve led a purposeful and meaningful life”. Items were rated on a 7-point Likert scale (7 = strongly agree; 1 = strongly disagree). For employees’ weekly well-being, Omega ranged from .79 to .94 across the four time points. For partners’ weekly well-being, Omega ranged from .83 to .93.

Control variables

Two demographic variables, gender and age, were considered potential control variables because of their influence on well-being experiences (e.g., Batz & Tay, 2018; Diener & Suh, 1998; Fields et al., 2022) and the inconclusive findings about their role. For example, focusing on gender, Stevenson and Wolfers (2009) found that men experience higher levels of well-being than women, whereas other studies have shown the opposite (e.g., Fujita et al., 1991). Focusing on age, the relationships observed have ranged from positive to negative and even U-shaped curves (Diener & Suh, 1998; Fields et al., 2022; Ulloa et al., 2013).

However, in our sample, age did not show significant correlations with any of the variables in the research model, whereas gender showed significant correlations with partners’ well-being. Given the small sample size and the complexity of the model, for the sake of parsimony, we decided to only control for gender in our model.¹ Although gender is considered a stable variable that can only have an effect on the between-person relationships we are controlling for when testing the within-person relationships of interest, we included this relevant covariate as a way of increasing power, given the modest sample size and number of measurement time points in our study. In fact, the inclusion of relevant covariates (at either level of analysis) has been suggested as a good option to increase power (González-Romá & Hernández, 2017; Pituch & Stapleton, 2012; Scherbaum & Ferreter, 2009).

Confirmatory factor analysis and within-individual reliability

We conducted a multilevel CFA with longitudinal data using Mplus 8.8 (L. K. Muthén & Muthén, 2017) to examine the factor structure of our four measures and be able to separate within-person reliability from between-person reliability by means of within-person omega coefficients and between-person omega coefficients (Bolger & Laurenceau, 2013). Because we are especially focusing on within-person fluctuations over time, the within-person coefficient is of special interest.

Given the large number of parameters to be estimated and the small number of level-2 units (50 individuals), we reduced the number of items by creating three aggregate indicators for the seeking resources and seeking challenges JC and LC scales, respectively, and two aggregate indicators for employees’ and partners’ well-being scales. This practice can even improve the measurement properties of the model without affecting the comparison of models of theoretical interest (Vandenberghe et al., 2002). Bayesian estimation methods with default priors were used, which are considered more suitable for small sample sizes (McNeish, 2016). The four-factor solution was adequate. First, to evaluate the fit of the four-factor model, we examined the posterior predictive p-value (PPP) and the potential scale reduction (PSR) values. The criterion for interpreting the PPP index suggests that values greater than .05 or .10 are considered acceptable (Asparouhov & Muthén, 2021). For the PSR index, values less than 1.05 indicate an acceptable fit to data (Zyphur & Oswald, 2015). The four-factor model showed an acceptable fit to the data (PPP = .11; PSR = 1.04). Then, we paid attention to the standardized factor loadings, which were statistically significant and greater than .40 (Fabrigar et al., 1999). Finally, as expected, the four factors were positively and significantly correlated (with correlations ranging between .20 and .52). These correlations between the factors are not high enough to threaten discriminant validity (<.70) (Hair et al., 2010).

The within-person omega coefficients, estimated by means of Multilevel CFA, were .74, .84, .88, and .83 for JC, LC, employees’ well-being, and partners’ well-being, respectively. These coefficients support the reliability of the scales in capturing individual change over time (the between-person omega coefficients range from .76 for JC to .98 for LC).

Results

Preliminary analyses

Before testing our hypotheses, we examined the within-person variance components of the dependent variables by calculating the intraclass correlation coefficient (ICC). The within-person variance in expansion-oriented JC was 55%, in LC, 54%, and in employees’ well-being and partners’ well-being, 36% and 55%, respectively. Means, standard deviations, and within-level and between-level correlations among all the study variables are reported in Table 1. Between-person correlations were calculated using the aggregated scores of each person over the available time points (four weeks for contemporaneous relationships and three weeks for lagged relationships -the only exception was for the lagged correlation between JC_t and employees’ well-being_t + 2), where the scores were aggregated over two weeks (1 and 2 for JC and 3 and 4 for employees’ well-being).

Table 1. Means, standard deviations, and correlations between the study variables at both levels.

Variable	M	SD	1	2	3	4	5	6	7	8	9	10
1. Gender	.52	.50	-	.07	−.03	−.14	−.22	−.32*	-	-	-	-
2. Age	38.60	12.53		-	−.04	.03	.15	−.21	-	-	-	-
3. Weekly Expansion-Oriented JC	3.02	.80			-	.27	.11	.06	.27	-	.04	-
4. Weekly LC	3.25	.91			.32**	-	.26	.01	-	.15	-	-
5. Employees’ weekly well-being	5.44	.97			.32**	.46**	-	.25	-	-	-	.15
6. Partners’ weekly well-being	5.53	.88			−.00	.10	.26**	-	-	-	-	-
7. Weekly LC^t+1	3.27	.76			−.15	-	-	-	-	-	-	-
8. Employees’ weekly well-being^t+1	5.45	.78			-	−.10	-	-	-	-	-	-
9. Employees’ weekly well-being^t+1ª	5.46	.93			.07	-	-	-	-	-	-	-
10. Partners’ weekly well-being^t+1	5.48	.78			-	-	−.19*	-	-	-	-	-

N = 50 participants and partners, and N = 200 data points. JC = Job Crafting. LC = Leisure Crafting. Within-person correlations are below the diagonal and between-person correlations are above the diagonal. ^t+1 indicates the values where we substituted the time 1 with time 2, time 2 with time 3, as well as time 3 with time 4 (N = 150 data points). ªWhen the predictor is Weekly Expansion-Oriented JC, ^t+1 is replaced by ^t+2 (N = 100 data points).

*p< .05, **p< .01.

Hypothesis testing

We used multilevel structural equation modelling (MSEM) to test the hypothesized 1-1-1-1 multilevel model.² We examined within-person relationships while controlling for between-person relationships and tested for the significance of the within-person indirect relationships by following Preacher et al.’s recommendations (Preacher et al., 2010). In addition, considering the order of causal precedence assumed, we also analysed the dynamic dependence for adjacent time units by modelling lagged relationships (T_i with T_i+1) between each pair of variables in the model and testing for the significance of the indirect effects. The analyses were run by means of Mplus 8.8 (L. K. Muthén & Muthén, 2017). Because all our hypotheses were directional, we ran one-tail tests. The instructions used to run all the models tested are provided as supplementary online materials.

Within-person contemporaneous relationships

First, we fitted a random-effects model (with free covariances among random slopes) and compared it to a fixed-effects model. Random-effects models can only be tested via the Bayesian estimation method, which is the most commonly used method for small sample sizes when models are complex (McNeish, 2016). The deviance information criterion (DIC) (B. Muthén, 2010) index was used to compare the two models. This index assesses the balance between the models’ parsimony and fit (Gill, 2008). The criterion for interpreting the DIC index suggests that lower values indicate a close fit to data. Because the DIC difference between the two models was substantial, that is, greater than 10 (Raftery, 1995), the results support the random-effects model. The DIC index of the random-effects model indicated a better model-data fit (DIC = 1627.32) than the DIC index of the fixed-effects model (DIC = 1686.28).

Next, we compared two random-effects models: one with covariances between random slopes fixed at 0 and the previous model, where covariances between random slopes were freely estimated. This step allowed us to determine whether it was necessary to incorporate the covariance between the random slopes to obtain the indirect effects (Bauer et al., 2006). The DIC difference between the two models was close to zero (diff = .37 < 10) (Raftery, 1995). In addition, none of the freely estimated covariances were significantly different from zero. Thus, we chose the random-effects model with covariances between random slopes fixed at zero as the best fitting model (DIC = 1627.69). Using the parameter estimates obtained from this model, we employed Selig and Preacher’s (2008) online interactive tool to test for the significance of the indirect relationships. This tool provides confidence intervals (CI) for the indirect effects using the Monte Carlo (MC) method (Preacher et al., 2010). Because our mediation hypotheses are directional, we used one-tailed tests, α = .05, and obtained the 90% CI (Cho & Abe, 2013; Preacher et al., 2010). Specifically, we used 20.000 replications for the MC simulation (Ruxton & Neuhäuser, 2010). The mediation effects were calculated using the means of within-level random effects across participants.

Before testing our hypotheses, we controlled for the effect of the employees’ gender. Results showed no significant gender differences in LC or personal well-being and significant differences in partner well-being, with women employees’ partners showing lower levels of well-being than men employees’ partners (B = −.39, p= .03).

Results are shown in Table 2 and summarized in Figure 2. Employees’ weekly expansion-oriented JC was positively related to employees’ weekly LC (B = .34, p< .001). This result supports H1. In addition, employees’ weekly expansion-oriented JC was significantly and positively related to their weekly well-being (B = .21, p= .03). This result supports H2a. Similarly, employees’ weekly LC was also significantly and positively related to their weekly well-being (B = .36, p< .001). This result supports H2b. Finally, consistent with Hypothesis 4, employees’ weekly well-being was significantly and positively related to their partners’ weekly well-being (B = .19, p< .01). Therefore, the results provide full support for the synchronous contemporaneous spillover of crafting behaviours and the effect of the relationship between the two types of crafting behaviours on employees’ well-being, which is finally related to their partners’ well-being.

Figure 2. Research model results.

*p< .05, **p< .01, ***p< .001

Table 2. Within-level estimates averaged for research model predicting week-level LC, and week-level employees’ and partners’ well-being.

	Weekly LC			Employees’ weekly well-being			Partners’ weekly well-being
Predictor	B	SE	p	B	SE	p	B	SE	p
Weekly Expansion-Oriented JC	.34***	.13	<.001	.21*	.13	.03
Weekly LC				.36***	.12	<.001
Employees’ weekly well-being							.19**	.09	<.01
Between-person (Level 2) residual variance (SD)	.47***	.13	<.001	.37***	.11	<.001	.42***	.11	<.001
Within-person (Level 1) residual variance (SD)	.31***	.04	<.001	.35***	.05	<.001	.32***	.04	<.001

N = 200 data points. JC = Job Crafting. LC = Leisure Crafting. Unstandardized Coefficients; one-tailed p values.

*p< .05, **p< .01, ***p< .001.

Regarding the indirect relationships hypothesized, we obtained the following results. Hypothesis 3 (employees’ weekly expansion-oriented JC ➔ employees’ weekly LC ➔ employees’ weekly well-being) was supported. The indirect “effect” was .11, which was statistically significant because the interval does not include the value zero (90% CI [.03, .23]). In addition, because employees’ weekly expansion-oriented JC was significantly and positively related to their weekly well-being (B = .21, p= .03) after controlling for LC, the results support partial mediation. Hypotheses 5a (employees’ weekly expansion-oriented JC ➔ employees’ weekly well-being ➔ partners’ weekly well-being) and 5b (employees’ weekly LC ➔ employees’ weekly well-being ➔ partners’ weekly well-being) were also supported. The indirect “effects” were .04 (90% CI [.01, .08]) and .07 (90% CI [.02, .14]), respectively. Finally, the sequential mediation proposed in Hypothesis 6 (employees’ weekly expansion-oriented JC ➔ employees’ weekly LC ➔ employees’ weekly well-being ➔ partners’ weekly well-being) was also supported. The three-path sequential mediation “effect” was .02, which was statistically significant (90% CI [.004, .05]).

Within-person lagged relationships

We followed the same steps as in the within-person contemporaneous analysis. The comparison of the DIC index of the random slope effects model (with free covariances among random slopes) (DIC = 1468.14) and the DIC index of the fixed-effects model (DIC = 1752.22) indicated that the random-effects model was preferable. Then we compared the two random-effects models: one with covariances between random slopes fixed at 0 and the previous one with free covariance estimates. The DIC index for the model with fixed covariances among the slopes was 1442.27. In this case, the DIC difference between the two random-slopes models was larger than 10, but it favoured the random-effects model with covariances between the slopes fixed at zero. Thus, this model was selected for parameter interpretations. Results are shown in Table 3. In this case, none of the hypotheses were supported. The lagged relationships proposed in H1, H2a, H2b were not statistically significant. The only significant relationship was found for H4. However, the lagged relationship between employees’ well-being and their partners’ well-being one week later was negative (B = −.19, p= .02). Because the hypotheses about the direct lagged relationships were not supported, we did not go on to test the significance of the indirect effects.

Table 3. Time-lagged within-person.

Predictor	Weekly LC^t+1			Employees’ weekly well-being^t+1ª			Partners’ weekly well-being^t+1
	B	SE	p	B	SE	p	B	SE	p
Weekly Expansion-Oriented JC	−.14	.14	.17	−.02	.19	.47
Weekly LC				−.01	.16	.46
Employees’ weekly well-being							−.19*	.09	.02

N = 200 data points. JC = Job Crafting. LC = Leisure Crafting. Unstandardized Coefficients; one-tailed p values. ^t+1 indicates the values where we substituted the time 1 with time 2, time 2 with time 3, as well as time 3 with time 4 (N = 150 data points). ^ªWhen the predictor is Weekly Expansion-Oriented JC, ^t+1 is replaced by ^t+2 (N = 100 data points).*p< .05.

Together, the results show that the hypothesized relationships are supported when we focus on within-person contemporaneous relationships, but not when we focus on lagged relationships from the previous week to the next. These results are discussed in the next section.

Discussion

Drawing on the W-HR model (ten Brummelhuis & Bakker, 2012), and focusing on within-person variability over time in a weekly study with employees and their partners, this study proposed that individuals change their work in order to grow and find meaning by performing expansion-oriented JC, and that this behaviour is generalized to non-work behaviours by engaging in LC. In addition, we proposed that these two resource-gaining behaviours (JC & LC) are not only related to higher employee well-being, but also, indirectly, to their partners’ well-being.

The results supported all the hypotheses in our research model when we focused on contemporaneous “effects” within the same week. Thus, employees and their romantic partners experience higher well-being during weeks when employees craft their work and leisure domains more, compared to weeks when they show fewer crafting behaviours in both domains. In addition, employees show more LC behaviours during the weeks when they show more JC. However, the results do not support the hypotheses about lagged relationships over time; that is, the levels of expansion-oriented JC shown by employees in a particular week do not systematically covariate with the levels of LC the following week. In addition, neither of these crafting behaviours in a particular week systematically covariates with employees’ well-being the following week.

When we focus on contemporaneous relationships during the same weeks, the evidence shows that weekly crafting behaviours at work are generalized to the leisure domain. This finding makes a relevant contribution to the literature because research combining these two settings is scarce (e.g., van Hooff & Geurts, 2015; Walker & Kono, 2018), even though both contribute to shaping individuals’ well-being. Second, the results reveal that weekly LC can help to explain how weekly JC behaviours contribute both directly and indirectly to employees’ weekly well-being because we found evidence for the partial mediating role of LC in the relationship between employees’ weekly JC and their weekly well-being. Third, we expand the crossover literature by showing that the positive relationship between employees’ well-being and their partners’ well-being was linked to employees’ weekly JC and LC behaviours.

Theoretical contributions

Findings of this study expand the W-HR model (ten Brummelhuis & Bakker, 2012) by providing evidence of the weekly spillover between JC and LC. We demonstrate that the resource gain spiral should be extended to non-work domains. Thus, this study highlights the need to use a holistic approach in understanding well-being. In this regard, our results strengthen the idea that we should not only consider domain segmentation as a way to cope with difficulties in work or non-work domains, but also the spillover between work and non-work behaviours, in order to understand how to increase well-being (see, for example, Walter & Haun, 2020b). Therefore, generalizing behaviours that encourage the development of personal resources to different domains may be an optimal way to increase employees’ well-being. Specifically, our results highlight the role of the leisure domain in well-being, expanding scarce previous research on the spillover of JC and non-work crafting behaviours. Moreover, we found that weekly JC and LC are two interrelated behaviours, providing support for the Integrative Needs Model of crafting behaviour (INM) (de Bloom et al., 2020). Accordingly, individuals with an affiliation need would satisfy it at work and in the leisure domain by generating social resources that, in turn, would ultimately impact their well-being.

Second, the findings support the relationships between JC, LC, and employees’ well-being, showing that individuals generalize their weekly proactive behaviours in work and leisure domains, and that this generalization is positively related to their weekly well-being. Until now and starting with the Petrou et al. (2017) study, crafting was understood as a zero-sum situation, with JC and LC compensating for each other. For example, when employees have fewer JC opportunities, they compensate by engaging in more LC behaviours. Our findings provide evidence that crafting can be generalized between domains: when people have JC opportunities, they can extend them to LC behaviours or vice versa. In this regard, we provide initial evidence that two specific crafting behaviours could be important ways to increase general well-being, defined as optimal functioning and experience. JC implies an awareness and motivation on the part of employees at work that they can use to enhance their leisure activities. This result is relevant because it goes beyond the domain-specific well-being usually analysed in the job-crafting and work-nonwork literature. Specifically, these results support bottom-up models of well-being that posit that domain experiences are causal antecedents to both domain-specific well-being and general well-being. These models explain that well-being is a consequence of positive and negative life events (Tay et al., 2014), and it may change as domain experiences change (Tay & Kuykendall, 2013). Therefore, as our study results show, during the weeks that employees experience positive events motivated by JC and LC, they also experience more general well-being.

Third, in this study, we expand the recent line of research that focuses on the specific benefits of LC (e.g., Petrou & Bakker, 2016) by providing initial evidence that LC could explain how JC is positively related to employees’ and their partners’ well-being. Based on previous findings, crafting leisure activities contributes to the development of the individual’s identity, relatedness needs, and motivation, due to its voluntary, autonomous, and intrinsically motivating nature (e.g., Newman et al., 2014). Moreover, LC would promote leisure seriousness (Kelly et al., 2020). –i.e., persevering in the face of difficulties, maintaining involvement, and generating an identity linked to the voluntarily chosen activity, which are useful resources that enhance well-being. Considering that individuals engage in serious leisure by engaging in LC, we can understand that LC would not only trigger the resource gain spiral in the non-work domain, but it would also reinforce the resource gain spiral initiated by JC. Future studies may consider how specific aspects of the use of leisure could explain the role of LC in the resource gain mechanism for improving individuals’ well-being.

The present research also demonstrates that both expansion-oriented JC and LC are positively related to partners’ well-being through their relationship with employees’ well-being. This result can be explained by the self-expansion (Aron et al., 1991) and emotional contagion processes (Fredrickson, 2001), so that the positive emotions or well-being produced in one life domain can have an impact on other domains. As far as we know, this is the first study to demonstrate that employees’ resource gain behaviours are linked to their partners’ well-being.

Finally, as we indicated above, our results do not support the lagged relationships between JC, LC, and well-being over time. One possible explanation is that, as meta-analytical results have shown, lagged relationships tend to be weaker than the corresponding concurrent ones. The difference becomes stronger as the interval between sampling moments increases (Hulin et al., 1990), and, thus, the tests would have enough power to detect the hypothesized lagged relationships. However, we believe that a more solid theory of time in organizations is needed (Roe et al., 2008), and particularly when dealing with crafting constructs and anticipating their impact on well-being and other relevant consequences. We carried out a weekly study because some leisure activities and tasks at work occur once or several times a week, but not daily (Hubbard & Mannell, 2001). However, one week may be too long to observe crossover and spillover effects over time. First, as Kosenkranius et al. (2021) showed, psychological needs satisfaction -e.g., mastery, meaning and affiliation- motivation, is related to crafting behaviour in both work and non-work domains. It is feasible that these psychological needs are more prone to having greater salience in some weeks than in others, which would contribute to both crafting behaviours being activated in both work and leisure domains during those weeks. In addition, the satisfaction of needs in the work and leisure domains has an immediate effect -e.g., weekly- on the well-being of employees (Kujanpää, Syrek, et al., 2021). Therefore, during those weeks when employees do more crafting behaviours, they will experience higher satisfaction. Thus, even if our results seem to reflect ongoing stable relations between measures, future research using shorter periods of time, such as daily data, or studies that clearly refer to leisure time after completing the workday and/or before starting the workday, may be helpful to provide further insight in this regard.

One surprising result in the lagged analysis that deserves attention is the negative lagged relationship between employees’ well-being and their partner’s well-being. Employees’ well-being shows an unexpected negative relationship with their partners’ well-being the following week. Given that the concurrent relationships are positive, the negative lagged relationship may indicate that, as the Adaptation Theory (Brickman & Campbell, 1971) suggests, employees will react to good or bad job and leisure experiences by changing their affective states and well-being. These changes, in turn, will affect partners’ feelings and well-being by means of contagion processes. However, after some time has passed (i.e., a week), they should return to their general affective states and well-being. Another possible hypothesis stems from social comparison theory. As in the previous explanation, in an initial stage – i.e., during the same week – there may be emotional contagion between the employee and the partner. However, as time goes by – i.e., next week – a social comparison process could be activated, and the partner may observe that the employee feels better than the partner does -upward social comparison- because, for example, the employee has obtained more resources -by performing JC and LC- to face certain demands, and this comparison is reflected in the partner’s lower well-being. Although some previous research shows that upward comparison has more positive emotional consequences, recent studies have shown that reactions to downward comparison depend on the type of attachment within the couple (Thai et al., 2016). Recent studies have also shown that social comparison can produce harmful consequences in the integration between work and personal life, especially in dual-income couples (Carlson et al., 2022). The different explanations are rather speculative at this point. Thus, future research should take these possibilities into consideration.

Limitations and future research

Despite our best efforts to capture the complex weekly processes hypothesized, the study has some limitations. First, in this study, all the variables collected over time, except employees’ and partners’ well-being, were collected through self-report measures provided by employees, which raises the issue of common method variance (P. M. Podsakoff et al., 2003). However, correlations among the main variables were low, which suggests low common method variance. Similarly, results of the CFA show that the measures in this study showed discriminant validity with moderate intercorrelations. Nevertheless, future studies may include other sources of information. For example, the employees’ partners could report on their perceptions of employees’ well-being, or co-workers might provide information about employees’ job crafting behaviours. Second, although our sample is heterogeneous in terms of sectors, position of responsibility in the organization, work status (i.e., salaried, or self-employed), and family burden, and it is adequate for robust estimations according to the criteria of Maas and Hox (2004), a larger number of participants should be included in future studies. An increase in the number and heterogeneity of the participants would allow us to systematically analyse the effect of individual differences such as gender, age, or personality. First, leisure literature suggests that age is a moderator of the relationship between leisure engagement – i.e., diversity of activities and time devoted to these activities- and well-being. Mental health literature (e.g., Matud et al., 2019) demonstrates that the greatest difficulties appear between the third and the fifth decade of life, especially for women, and mainly in caregiving activities. From this age, people would benefit from engaging in a large variety of leisure activities because these activities would help them to structure their leisure time. In future studies, different age ranges could be included. Second, it would be important to consider the effects of personality in future studies. For example, a proactive personality may predispose individuals to behave proactively in different domains (McCormick et al., 2019) because they would generalize their capacity for initiative, identification of opportunities, and perseverance in pursuing goals (Rudolph et al., 2017).

Third, although the random-effects model of mediation proposed by Kenny et al. (2003) is appropriate for testing within-subjects mediated relationships in longitudinal data, the only model that received support was the one that “specifies contemporaneous effects of X on M and of M on Y” (Maxwell & Cole, 2007, p. 34) over time. Although this model has several measurement points across time, we cannot establish causal relationships between variables assessed at the same moment or exclude reciprocal relationships. With regard to causality, the lagged within-person relationships were not supported by the data. These results might be attributed to the small number of observations per person, or to the fact that the one-week time-lag was too long to detect crossover and spillover effects. Thus, even if the contemporaneous relationships tested offer preliminary evidence about the mechanisms that explain why JC and LC contribute to employees’ and their partners’ well-being, future longitudinal studies with shorter time lags (or that clearly refer to LC after completing and/or starting the workday) would help to clarify the directionality and/or reciprocity of the relationships under study. The random-intercept cross-lagged panel model (Mulder & Hamaker, 2021) would be useful for this purpose. This model was too complex for our sample size (N=50). Nevertheless, following the anonymous reviewers’ suggestions, we tried to gain some additional insight into the directionality of the effects by testing additional models (see supplementary online materials).

First, from the within-subjects perspective, we ran the multilevel random-effects model (both for contemporaneous and lagged relationships), reversing the order of the variables in the model (partners’ well-being ➔ employees’ well-being ➔ LC ➔ expansive-oriented JC, with a direct link also from employees’ well-being to expansion-oriented JC). We tested this model because previous research has shown that well-being predicts job crafting (Hakanen et al., 2018), and, according to an anonymous reviewer, it is reasonable that well-being could also predict leisure crafting, which may affect job crafting in turn. Finally, it could be argued that job and leisure crafting are not causally related, but that employees’ well-being causes both. The model that tested reversed contemporaneous relationships (multilevel model with random slopes with covariances among random slopes fixed at zero, which was the best fitting model) showed that partners’ well-being was significantly and positively related to employees’ well-being. Employees’ well-being was significantly and positively related to employees’ LC, which, in turn, was positively related to employees’ expansion-oriented JC. These results, together with the previous results, suggest that the relationships between well-being and LC may be reciprocal, and that crafting behaviours may also spill over from non-work to work. However, employees’ well-being did not predict employees’ expansion-oriented JC. Measures of well-being at work may be more relevant than general well-being to predict job crafting. Regarding the model that tested reversed lagged relationships (multilevel model with random slopes with free covariances among random slopes, which was the best fitting model), none of the tested relationships were statistically significant. More detailed results are provided in the online supplementary materials.

Second, focusing on a between-subjects perspective, we tested whether expansion-oriented JC at Time 1 was related to LC at Time 2, which was related to employees’ well-being at Time 3, which, in turn, was related to partners’ well-being (measured at Time 4) (see time-lagged Model 8 in the supplementary materials). This sequence is congruent with the causal order proposed in our model. In addition, following our research model, employees’ well-being was also regressed on expansion-oriented JC. This model showed a satisfactory fit to data. However, none of the hypothesized relationships were statistically significant. A similar model but controlling for stability effects (by including the lagged DV as a predictor), was also fitted (see time-lagged Model 9 in the supplementary materials). In this case, none of the hypothesized relationships were significant either, although in this latter case, the model did not show an acceptable fit.

All these results together indicate the need to increase the number of participants and observations in order to overcome some of the methodological limitations of this study. We encourage the use of longitudinal studies with more heterogeneous populations and several data collections per day, in order to analyse the link between work, leisure time, and employees’ and their partners’ well-being over time, where antecedents precede the mediators, and the latter precede the outcomes. Longitudinal studies would also make it possible to analyse the possibility of reciprocal relationships via the random-intercept cross-lagged panel model (Mulder & Hamaker, 2021). Experimental or quasi-experimental designs with interventions designed to independently increase JC and LC would also be extremely helpful.

Finally, different research approaches have suggested that crafting behaviours are not exclusive to the workplace or to the leisure domain. For instance, Chen (2020) analyzes online leisure crafting; Demerouti et al. (2020) analyzes home crafting; and Sturges (2012) and, more recently, Dreyer and Busch (2021) analyse work-home balance crafting. These authors conceptualize these crafting behaviours by drawing on the previous JC concept. Accordingly, they differentiate the specific strategies people use to craft the specific domains – e.g., reduce demands, seek challenges, and seek resources-. Future studies could analyse the effect of the specific dimensions of these different crafting behaviours on individuals’ well-being and how they relate and interact with each other to predict individuals’ well-being.

Practical implications and conclusions

Our findings have important implications for both theory and practice. From an applied perspective, HR managers should be aware of the positive impact of expansion-oriented JC on employees’ well-being. For example, training or counselling programmes are needed to teach employees how to craft their job and domains (e.g., van den Heuvel et al., 2015). In addition, exploring and feeling the positive effects of expansion-oriented JC in these intervention programmes will lead to more expansion-oriented JC behaviours in the future (Tims et al., 2014). Specifically, organizations should consider that expansion-oriented JC not only has a positive impact in the work domain, but also in the non-work domain, which, as stated above, can benefit them.

According to the bottom-up model of well-being, organizations could help their employees to control, improve, and create opportunities to carry out plans, not only at work, but also in the leisure domain. To achieve LC opportunities, individuals can participate in vicarious experiences – e.g., activities organized by significant others such as family or friends and their own hobbies (Berg et al., 2010). Based on the literature on leisure activities, to take advantage of leisure, it is important to consider the types of activity individuals do. For instance, sports, socializing, or attending cultural events -as opposed to watching TV – support the fulfilment of psychological needs (Kuykendall et al., 2015). As the LC literature highlights, during leisure time, it would also be relevant to establish personal goals and promote personal relationships as well as learning and personal development (Petrou & Bakker, 2016). In sum, organizations should see employees as active agents who are capable of using both their work and leisure time as they see fit to increase their well-being experiences.

From a theoretical perspective, our findings demonstrate that expansion-oriented JC behaviours have beneficial implications beyond the work domain. Our results suggest that these behaviours may be expanded to the leisure domain via a positive spillover effect. This study implies that crafting expansion-oriented job behaviours could motivate employees and have a vigorous effect on leisure life. Additionally, we expand the crossover model of positive weekly well-being. Our findings highlight the key role of employees’ weekly LC as a resource gain behaviour that not only leads to their weekly well-being, but also to their partners’ weekly well-being, via a positive indirect crossover effect.

Together, the results of this study show that, beyond segmenting the work and personal life domains, integrating behaviours such as job and leisure crafting is quite positive, not only for employees, but also for their partners. Moreover, LC seems to play a more relevant role than JC in determining individuals’ well-being. Therefore, both organizations and employees should look for opportunities to successfully engage in these crafting behaviours.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Available Upon Request

Additional information

Funding

The work was supported by the Ministerio de Ciencia e Innovación y Universidades Spanish Ministry of Science, Innovation and Universities The project PID2019-110093GB-I00 was funding by the MCIN/AEI /10.13039/501100011033 3.

Notes

1. Given our focus on within-person relationships, and considering gender is a between-person variable, we conducted analyses excluding gender. When gender was omitted, the coefficients linking employees’ weekly well-being with their partners’ weekly well-being (both contemporaneous and lagged “effects” one week later) were not statistically significant, despite the coefficients’ sizes being similar. Detailed results are provided as supplementary online material.

2. Given the nature of our four-week data, the autoregressive model of change proposed by Maxwell and Cole (2007) would have been an adequate strategy to study mediation at the between-subjects level. However, this model had identification problems because the sample size was too small (N = 50) for such a complex model (62 free parameters) (see instructions in supplementary online material). According to Maxwell and Cole (2007), multilevel modelling with random intercepts and slopes is also a suitable model for longitudinal relationships at the within-subjects level.