Good mental health despite work-family conflict? The within-domain and cross-domain buffering potentials of family and work resources

Objective: This article investigates whether within-domain and cross-domain buffering by family and work resources can help mitigate the negative mental health effects of work-tofamily conflicts (WFC) and family-to-work conflicts (FWC). Background: Most literature on the work–life interface stresses the need to maintain employees’ health and well-being by preventing the emergence of work–family conflicts. Since such conflicts tend to be an unavoidable concomitant of role expansion, we aim to put forward the debate on the conditions that might prevent their negative health consequences instead. Method: Fixed-effects linear regression analyses were applied to a sample of 4,920 employees in a three-wave employer–employee panel study in Germany. Using interaction analyses, we tested within-domain and cross-domain buffering of family (social support and relative bargaining power within partnerships) and work (job resources, support from direct supervisors or co-workers, formal and informal organizational support) resources in the relationship between strain-based and time-based WFC and FWC and mental health (SF-12). Results: Family resources and work resources somewhat mitigated the health risks of WFC and FWC. Overall, within-domain resources were more effective than cross-domain ones. Conclusion: It is important to consider resources in both the family and the work domains to determine the most effective ways of preventing the negative mental health consequences of work–family conflicts.


Introduction
To facilitate work-family integration, most of the literature suggests that conflicts due to competing demands in an employee's family and work roles need to be prevented right from the start (Amstad et al., 2011). Indeed, extensive research has explored that workfamily conflicts negatively affect an individual's well-being and mental health in particular (e.g., Chandola et al., 2004;Cooklin et al., 2016;Hämmig & Bauer, 2009;Jensen & Knudsen, 2016;Zhao et al., 2021). Within families, such conflicts play a role in two directions: the family setting can be the source of conflicts that then spill over into a person's work life (i.e., family-to-work conflicts [FWC]) (Gutek et al., 1991;Netemeyer et al., 1996), while conflicts that a person experiences at work can be played out within the family setting (work-to-family conflicts [WFC]) (Greenhaus & Beutell, 1985). Scholars increasingly recognize that reconciliation needs are heterogeneous -for example, regarding negotiation processes within dual-earner couples (Hammer et al., 2005;Yucel & Fan, 2019), specific challenges of single parents (Reimann et al., 2019), or fathers (Graham & Dixon, 2014). Thus, in many parts of the workforce and in many families, both WFC and FWC need to be seen as a common occurrence (Byron, 2005;Gallie & Russell, 2009;Winslow, 2005). Without relativizing the need to understand the sources of such conflicts or attempts to avoid them, we argue that they are an unavoidable concomitant of role expansion (Grönlund & Öun, 2010;Mauno et al., 2006). For those who strive to meet their life goals in both family and work life, conflicts are bound to arise at some point, in one way or the other. Moreover, individuals who hope to satisfy ambitious goals in both domains may consciously come to view these conflicts as acceptable.
Following this argument, our study takes one step further to investigate how, despite the presence of WFC or FWC, employees might still experience good mental health. Referring to job demands-resources theory (Bakker & Demerouti, 2007;Bakker et al., 2005) and supplementary considerations on family demands-resources (Voydanoff, 2005), we argue that the degree to which conflicts eventually lead to impaired mental health is partly shaped by resources already available in the family and/or work spheres. Since both life domains may provide resources to handle stress (Billings & Moos, 1982;Voydanoff, 2005), we propose that conflicts may be buffered in the domain where they emerge (within-domain buffering) -that is, FWC would be moderated by family resources and WFC by work resources. However, if the conflicts initially emerged within an area of life, resources in that specific domain might not be effective enough to mitigate negative health effects, because such resources would have prevented conflicts in the first place. Thus, the negative consequences must be compensated for in the other life sphere, respectively (cross-domain buffering) -that is, FWC would be buffered by resources in the work domain and, conversely, WFC would be buffered by resources in the family domain.
Research in the area of family resources has been limited and inconclusive, for instance when it comes to determining whether social support from family members helps to buffer the effects of WFC and FWC on health (e.g., Kalliath et al., 2015;Parasuraman et al., 1992). In the area of work resources, studies have shown that supervisor support and a family-friendly climate in the workplace have mitigated the negative health effects of WFC (Mauno et al., 2006;O'Driscoll et al., 2003) but not of FWC (Beauregard, 2011). However, studies to determine the effectiveness of other organizational policies, such as formal work-family organizational support or healthpromotion, are lacking. In sum, when it comes to the association between work-family conflicts and mental health, empirical evidence regarding within-domain buffering is scarce and is even more so when it comes to cross-domain buffering.
We aim to fill this research gap by clarifying (1) whether family resources or work resources are more powerful in buffering the negative effects of WFC and FWC and (2) whether within-domain or cross-domain buffering is more helpful. We look at several family resources (having a partner, appreciation from partner, relative bargaining power in the relationship) and at a number of work resources (job resources such as flexible working, home-based telework, job autonomy, and wages; support from direct supervisors and co-workers; formal work-family organizational support such as work-family policies and health promotion; informal work-family organizational support such as a workfamily supportive workplace culture). In this way we aim to provide a better sense of which specific resources are helpful not only in explaining how to avoid conflicts in the first place (Michel et al., 2011), but also in mitigating their negative consequences on mental health. Next, we differentiate not only between the two directions of conflictsfrom work to family and from family to work -but also between strain-based and timebased conflicts as two subdimensions of WFC and FWC (Carlson et al., 2000).
Thereby we aim to provide theoretical and empirical evidence on the so-far clearly neglected role of families in coping with different directions and forms of work-family conflicts. We emphasize the necessity to understand families not only as possible sources of demands in the emergence of conflicts (mostly due to higher caring responsibilities of parents as compared to childless persons), but that they also provide potential resources despite already existing conflicts. To overcome a perspective that reduces handling WFC and FWC to an individual responsibility, living conditions in the family may be the most complex structural dimensions that we need to understand. This research is also important from the viewpoints of social policy and of employers: Even if we acknowledge that we cannot prevent work-family conflicts completely, we should still seek to increase the resources that will reduce its negative health outcomes (Mauno et al., 2006).
Finally, by using three waves from a longitudinal linked employer-employee panel dataset representative of large work organizations (more than 500 employees) in Germany (LEEP-B3; for further information see Diewald et al., 2014), and by applying fixed-effects regression modeling, we were able to overcome the limitations of the still prevalent crosssectional work-family studies and to learn more about the direction of causality in the relationship between work-family conflicts and health (Nohe et al., 2015).

Role strain: Negative effects of work-family conflicts on mental health
The family and work life spheres have become increasingly blurred, partly as a result of digital technologies that enable more flexible work arrangements (Kossek, 2016;Wajcman et al., 2010). These two life domains have different demands, values, and functions (Burkart, 2005;Parsons, 1949;Voydanoff, 2002), and because they compete for individual resources (especially time and energy), they often come into conflict with each other (Kahn et al., 1964). The bidirectional conflicts of WFC and FWC can include time-based and strain-based subdimensions: Time-based conflict refers to a person's multiple roles competing for limited time (i.e., time allotted to the person's role in one domain cannot be used for that person's role in the other domain). Strain-based conflict refers to the strain induced by the demands of a person's role in one domain that consequently reduce that person's ability to deal with role requirements in the other domain. In modern societies, both the directions and forms of work-family conflicts are part of everyday life; this situation is no less true in the German workforce (Abendroth & Reimann, 2018;Reimann et al., 2017;Beham & Drobnič, 2010;Gallie & Russell, 2009;Steinmetz et al., 2008), although the WFC tend to be greater than the FWC (Nohe et al., 2015;Seery et al., 2008).
Based on the assumptions posed in stress theories (Lazarus & Launier, 1978;Thoits, 1995), both WFC and FWC may be experienced as stressors that threaten a person's mental health (Greenhaus et al., 2009;Mauno et al., 2006). Manifold empirical studies confirm the negative effects of WFC and FWC on various health outcomes (for overviews, see Allen & Martin, 2017;Allen et al., 2000;Nohe et al., 2015), including mental health (e.g., Chandola et al., 2004;Cooklin et al., 2016;Hämmig & Bauer, 2009;Jensen & Knudsen, 2016;Sekine et al., 2006). However, most of these studies have been crosssectional, and the few longitudinal studies have been restricted to highly selected occupational groups or specific cohorts (Allen et al., 2000;Nohe et al., 2015). Moreover, the effects of strain-based and time-based conflicts are only rarely differentiated (Carlson et al., 2000). Based on the reviewed literature, we assume that strain-based as well as timebased WFC and FWC have negative effects on mental health.
However, work-family conflicts are not universal experiences. Especially gender, parenthood, and the intersection of both are among the most investigated groups in work-life interface research because they differ in their conflict perceptions (Bianchi & Milkie, 2010;Parasuraman & Greenhaus, 2002), and differences between these groups are often explained by differences in family and work demands and resources (Voydanoff, 2002). In contrast, findings on gender differences in the consequences of WFC and FWC are controversial (Borgmann et al., 2019;Frone, 2000). Though gender is not a focus of our study and we do not have gender-specific hypotheses, we offer gender-sensitive modeling and discussion of the results.

Role expansion: Can family and work resources help to mitigate the negative effects of work-family conflicts on mental health?
In contrast to the negative effects of role conflicts, the role expansion hypothesis assumes that the combination of dual roles in paid work and in the family setting can have positive effects on an employee's health (Barnett & Hyde, 2001;Jaga et al., 2013;Mullen et al., 2008), albeit this hypothesis has so far received less research attention. The feeling of being needed and appreciated in different contexts can strengthen a person's self-esteem and create a sense of security, since problems and failures in one sphere can be compensated for by success and satisfaction in the other. Those positive work-family interactions have been described as "positive spillover" (Poelmans et al., 2008), "facilitation" (Grzywacz et al., 2007), and "enrichment" (Greenhaus & Powell, 2006). Moving beyond the competing theoretical roles of role conflict and role expansion, our research is inspired by the work of Grönlund and Öun (2010), who propose that employees might experience high levels of both work-family conflicts and well-being at the same time. Moreover, we agree with the general criticism of use of the term "work-life balance" in work-life interface research, since it distinguishes between work and "life outside work" and suggests that the two need to be balanced. Instead, having a job and occupational success can be important parts of a person's life goals despite the demands they impose and despite the risk of conflicts (Guest, 2002;Lewis et al., 2007). We assume that work-family conflicts are accepted by employees at least in part as a way of fulfilling their goals in different life domains simultaneously. Accepting that work-family conflicts are often unavoidable, we will now address the question of what conditions might mitigate the negative effects of WFC and FWC on mental health.

Within-domain and cross-domain buffering
As can be found in the literature on job stress, coping resources can serve to buffer the negative effects of stressors (Cohen & McKay, 1984;Lazarus, 1966). In work-family conflicts, these negative consequences can be buffered either in the life domain where they had emerged (within-domain buffering) or in another life domain (cross-domain buffering). Borrowing from general assumptions of job demands-resources theory (Bakker and Demerouti 2007;Bakker, Demerouti, and Euwema 2005) and supplementary considerations on family demands-resources (Voydanoff, 2005) there are possible manifold resources in both life domains. Social support is probably the most established resource that has been proposed as a potential buffer against stress (Cohen, 2004;House et al., 1988). According to the utility perspective, social support is most likely to buffer strain in those particular circumstances in which support is needed or perceived to be useful (Cohen & McKay, 1984), indicating its potential for both cross-domain and withindomain buffering. Although it would seem that WFC should be buffered by work resources and FWC by family resources, the conflicts may have resulted from a lack of resources within the particular life domain itself, indicating that there were fewer possible resources within that environment in the first place (Allen, 2001). In this case, conflicts need to be handled in the domain into which the conflict is spilling over.
This situation becomes most evident when it comes to WFC. Macro theories of functional differentiation assume a division of labor between the subsystems of gainful employment and the family. Here, an important function of the family is to provide recreation from work -that is, as a source of compensation for the strains imposed by work as a way to achieve well-being and good mental health (Burkart, 2005;Parsons, 1949). From this viewpoint, primarily specific qualities of family life would be required for work strains to be buffered successfully by supportive relationships or other household resources, thus pointing to the greater importance of family resources in buffering WFC (cross-domain buffering).
A similar case can be made for FWC. If strains and conflicts in family life cannot be avoided and are transferred to the workplace, one should first ask how certain characteristics of the workplace can be shaped to buffer their potentially negative health effects. For example, such protective factors can be sought in the form of supportive workplace relationships or formal or informal organizational work-family support.

Figure 1:
Within-domain and cross-domain buffering in the relationship between workto-family and family-to-work conflicts and mental health Source: Own illustration.
Overall, we distinguish two different buffering pathways and propose two hypotheses linked to each pathway (see Fig. 1), as follows:

Within-domain buffering:
Hypothesis 1: Family resources buffer the negative effects of FWC on mental health. Hypothesis 2: Work resources buffer the negative effects of WFC on mental health.

Cross-domain buffering:
Hypothesis 3: Family resources buffer the negative effects of WFC on mental health. Hypothesis 4: Work resources buffer the negative effects of FWC on mental health.
In the following, we discuss specific resources that might moderate the relationship between WFC/FWC and mental health: family resources (social support within partnerships) and work resources (job resources, support from direct supervisors and coworkers, formal and informal organizational work-family support).

Family resources
An important resource in the family domain is social support by family members. It may buffer the effects of stressors on mental health through practical help, providing information, or emotional encouragement (Cohen & McKay, 1984;House et al., 1988). In line with the role expansion hypothesis, the feeling of being needed and appreciated in different contexts may protect against the negative consequences of stressful experiences in a particular life domain (Greenhaus & Powell, 2006). Having someone available to talk to about problems may help to focus on the more positive aspects of the two life spheres or may provide a reserve of positive feelings of belonging and being valued, thereby reducing the stressful experience of WFC or FWC. In the first instance, this support can be provided by being in a partnership. Empirical evidence is inconclusive: Some studies find that family resources similarly buffer the negative consequences of both FWC and WFC (French et al., 2018), whereas others show that family or spousal support can act as a buffer to mitigate the negative consequences of FWC, but a buffering of WFC's negative effects cannot be confirmed (Kalliath et al., 2015;Parasuraman et al., 1992). Some studies have observed that family or spousal support has no buffering effect whatsoever (Burke & Greenglass, 1999;O'Driscoll et al., 2004).
In this context, support may be emotional support and appreciation from a partner, but it may also depend on one's negotiating power within a relationship. As assumed in bargaining models, the allocation of available resources within the household can be understood in terms of an individual's position vis-à-vis one's spouse or co-habiting partner, the one with the superior position having the greater power to negotiate favorable resources (Bittman et al., 2003;Cooke, 2006;Lundberg & Pollak, 1996). In the case of combining work and family demands, a balanced relationship can be found if both partners work full-time and have a similar occupational status or education level. It may be imbalanced if one partner works full-time and the other partner works only part-time or is not employed at all, or if the partners differ in their occupational status or education level. Although a balanced relationship may profit when demand levels are similar and mutual understanding is greater (Yogev & Brett, 1985), in relationships in which the status of one partner is inferior, resources may be allocated to the superior partner and may serve to reduce role strain due to WFC and FWC. In comparison, dual-earner couples with a balanced relationship need to balance their resources and thus may be less able to prevent the consequences of conflicts (Hill et al., 2006). In sum, the relative position in a relationship may not be a direct resource that helps to handle WFC and FWC, but it can be understood as an indirect indicator of the distribution of valuable resources in the relationship that in turn are directly helpful in managing conflicts.

Job resources
We considered flexible working practices (flexible working hours, home-based telework), job autonomy, management position, and wages as examples of job resources. Flexible working arrangements are one of the most discussed job resources in work-family conflict literature. Though their actual benefit is regarded as controversial (for example because they can act as stressors if they mostly serve the flexibility interests of employers or if they are used for getting additional work done, Kelliher & Anderson, 2010), scholars predominantly emphasize the opportunities involved in flexible working hours or homebased telework (Allen et al., 2013). Also, they are regarded as important factors in employee health, mainly because they are associated with a greater latitude to cope with stressful situations (Bakker et al., 2005). Such measures allow the working day to be adjusted so the employee can deal with conflicts carried over from family life and are therefore helpful in preventing or dealing with conflicts and in buffering against negative effects on mental health (Bakker et al., 2005;Mauno et al., 2006). Similarly, being in a management position can be useful for handling WFC and FWC in that managers generally have greater access to company resources, not only when it comes to flexible working or autonomy, but also in drawing on other capabilities such as decision making and control, which can serve to reduce strain (Kelly & Kalev, 2006). Thus, having a management position can be a resource to effectively handle existing WFC and FWC. In this sense, acquiring a management position may be especially helpful because it provides employees with new opportunities for dealing with conflicts and it may also be understood as a gift for tolerating conflicts. However, higher status positions are not solely associated with higher resources but also with more demands such as increased responsibilities and more overtime work. As discussed in regards to the stress and resources of higher status hypotheses (Badawy & Schieman, 2020;Schieman & Reid, 2009), this may affect work-family conflicts as well as health. Therefore, it remains an open question whether having a management position can buffer the negative health effects of already existing WFC and FWC.
Higher wages can be a monetary compensation for an employee's investment in a job that has high demands (French & Dunlap, 1998). In the context of work-family conflicts, an increase in wages may also be understood as a form of compensation for employees who tolerate conflicts as they arise if they also exhibit a greater commitment to the organization by putting more effort in their job as well as their family life. Thus, wage increases act as a potential buffer for all types of conflicts.

Support from direct supervisors and co-workers
As discussed earlier, social support is assumed to be one of the most likely potential buffers of work stress. Therefore, supportive relationships with one's direct co-workers and supervisors could also be helpful in buffering the effects of all types of conflicts that affect a person's mental health. Nevertheless, existing studies have focused mainly on WFC, and evidence for the moderating role of such support has been mixed. Overall, supervisor support seems to be more important than co-worker support and is relevant mostly for WFC (Bakker et al., 2005;Beauregard, 2011;O'Driscoll et al., 2003). Regarding the two different forms of WFC, support from supervisors was found to buffer the effect of time-based WFC, but not of strain-based WFC, on employees' emotional exhaustion (Karatepe & Kilic, 2015). Others have reported that neither supervisor support (Mauno, 2010) nor co-worker support (Kalliath et al., 2015) had a moderating effect on the association between WFC or FWC and an employee's well-being. In our research, we consider both the support from direct supervisors as well as from direct co-workers to be potential buffers of WFC and FWC.

Formal work-family organizational support
Organizations are situated within institutional fields that provide sets of rules and meanings that can be used to legitimize resource allocation (Avent-Holt & Tomaskovic-Devey, 2010). Thus, workplace characteristics can alter how experienced conflicts impinge on an individual's well-being. Responsibility for the integration of work and family life has increasingly been shifting from the private sphere to the organizational sphere, which is referred to as the institutional/organizational turn (Moen, 2015). As a consequence, many large organizations in Germany offer a variety of benefits to help employees meet familyrelated demands. Most such benefits are intended to provide greater schedule flexibility or to help with dependent care and have been described as formal work-family organizational support (Behson, 2005), which is also designed to prevent health impairment caused by work overload linked to competing demands in both these life spheres. If organizations offer these policies as way of investing in the employability of persons who have considerable family obligations, this action might signal the company's support for employees' commitment to both work and family life, as well as an appreciation of employees' tolerance of ensuing conflicts. Thus, the offer of such formal work-family organizational support may buffer the negative effects of both WFC and FWC on an employee's mental health.

Informal work-family organizational support
We looked at workplace culture as a more subjective dimension of organizational support, which has been described as informal work-family organizational support (Behson, 2005). A workplace with a work-family supportive culture -that is, one that encourages employees to make use of its supportive measures and that acknowledges that employees have to handle both work and family demands -can be a potential resource for preventing negative health consequences. Moreover, this informal work-family organizational support can also be represented by a low-demand workplace culture that values employee's high commitments to both work and family life by emphasizing that it is not important to always be available for work (even outside regular working hours), or that employees do not need to do extra work all the time. Perceived informal organizational support was found to buffer the negative effects of WFC on depressive symptoms (Hao et al., 2016), and a family-supportive climate was found to buffer WFC's negative effects on well-being (Mauno et al., 2006). Thus, we consider informal work-family organizational support as a potential buffer of WFC and FWC.

Data
The empirical analyses are based on three waves of a linked employer-employee dataset that was collected between 2012 and 2019 as part of the study "Interactions between Capabilities in Work and Family Life" (LEEP-B3). The study consisted of an employer survey and an employee survey. In line with the purposes and the methodological design of the study (for more information and details on the sampling process and data structure, see Diewald et al., 2014), the sample was restricted to large organizations with 500 or more employees required to pay social security contributions. In collaboration with the Institute for Employment Research (IAB), the population of all corresponding organizations was derived from mandatory administrative employer data in Germany (3,934 organizations). To cover all industrial sectors (exceptions are agriculture, forestry, and fishing industries, and mining industries) and organizations from North and West Germany, a disproportional stratified random gross sample of 539 employers was drawn. 115 employers were interviewed successfully. Multivariate analyses comparing the gross and the net samples with regard to industrial sector, size and location revealed that there are only marginal differences between the population and realized interviews . Of all 115 employers, 100 agreed to the employee survey. For each of those, a simple random sample of employees was drawn based on the administrative data. Multivariate analyses comparing gross and net samples regarding employee characteristics (e.g., gender, age, education, work experience, earnings, tenure, industrial sector, size of the organization) reveal that the employees who participated in the survey are representative of those who work in large organizations (>500 employees) in Germany . Employee interviews were conducted using computer-assisted telephone interviews (CATI), whereas the employers were interviewed mainly face-to-face. The first wave, t1 (2012/2013), comprised 100 organizations and 6,454 employees, and the response rate was 29%. Of these, 4,000 employees also participated in the second wave (t2) (2014/2015) (a response rate of 73.3% for panel respondents, as calculated based on AAPOR [The American Association for Public Opinion Research, 2015]). To take into account systematic and non-systematic panel mortality, new employers and employees were added to the sample during both wave t2 and wave t3 (2018/2019), which resulted in a total of 160 organizations participating in at least one of the three waves. For our purposes, we restricted our sample to all employees who had participated in at least two consecutive panel waves (5,016 cases). Missing values in all waves on any of the dependent (2 cases) or independent variables (96 cases) led to a further reduction, so the sample for the final analysis included a total of 4,920 panel cases: 2,317 employees who participated in all three waves, 1,607 who participated in waves t1 and t2, and 996 who participated in waves t2 and t3. Of the employees who participated in all three waves 399 had missing values on any of the independent values in t1 or t3. Those cases were kept in the analyses as they still had valid data for two consecutive waves, leading to 11,359 observations in the fixed-effects regression analyses. The final analysis sample and the missing sample (= including any observations of cases that were dropped due to missing values) were compared using mean comparison and t-test statistics for socio-demographic characteristics (see Table A1 in Online Appendix). Though there are some statistically significant (p<0.05) differences between the samples, those differences are marginal and mostly explained by the large number of observations and/or by the fact that the analysis sample includes panel cases only whereas the missing sample also includes employees that only participated in 2012 (e.g., higher average age and earnings in the panel sample). However, it must be noted that the analysis sample is marginally but statistically significantly healthier. This might be explained by a survival bias of healthier workers in longitudinal data (e.g., Buckley et al., 2015) as we do not find these differences when comparing data from all three waves separately.

Mental health
The outcome variable "mental health" was measured using the SF-12 summary scales for mental and physical health (German Socio-Economic Panel version; see Andersen et al., 2007). Each of the two superordinate dimensions of the SF-12 -mental health (MCS) and physical health (PCS) -is represented by six items. Exemplary items for MCS are "During the last 4 weeks, how often did you accomplish less than you would have liked as a result of any mental or emotional problems" and "During the past 4 weeks, how often did you have a lot of energy" (reversed). The MCS and PCS scores were generated by conducting a confirmatory factor analysis in structural equation modelling on all twelve items for mental and physical health and based on the used panel sample (e.g., t3: χ 2 (45) = 906.92, p<0.001, RMSEA = 0.069, CFI = 0.952, TLI = 0.930). In contrast to the conventional orthogonal scoring method, this type of analysis allowed us to correlate the factors, thus reflecting the more realistic assumption that mental and physical health may influence each other (Tucker et al., 2014). In accordance with the original approach, the scores were standardized to a sample mean of 50 and a standard deviation of 10, with higher values indicating better mental health.

Work-family conflicts
The four subdimensions of work-family conflicts (i.e., time-based and strain-based WFC and time-based and strain-based FWC) were measured using the work-family conflict scale developed by Carlson et al. (2000). The original scale was translated into German for the purposes of the LEEP-B3 survey, with each subscale being represented by three items. Examples of statements included for the four subdimensions are as follows: time-based WFC (e.g., "I have to miss family activities owing to the amount of time I must spend on work responsibilities"); strain-based WFC (e.g., "I am often so emotionally drained when I get home from work that it prevents me from contributing to my family"); time-based FWC (e.g., "Because of family appointments, I often have problems getting my work done"); and strain-based FWC (e.g., "Because I am often stressed owing to family responsibilities, I have a hard time concentrating on my work").
Three subscale items were used to measure how often time-based and strain-based WFC and FWC, respectively, were experienced, with responses ranging from 0 (Never) to 4 (Very often). For each direction and type of conflict, the scores for the three items were added up and divided by the number of items and response options, resulting in a range between 0 and 1, with higher values indicating higher levels of conflict. Moreover, the measures for WFC and FWC overall were computed similarly but included all six items each. The internal consistency of all dimensions was good, with Cronbach's alpha coefficient (of reliability) exceeding the conventional level of acceptability (time-based WFC: α = 0.85; strain-based WFC: α = 0.78; time-based FWC: α = 0.75; strain-based FWC: α = 0.69). All conflict dimensions are correlated but distinct (see Table A2 in Online Appendix) in line with the original validation of the scale and previous research (Carlson et al., 2000;Nohe et al., 2015;Seery et al., 2008). Correlations are lower (0.137 to 0.231) between the two directions (WFC and FWC) of conflicts, and comparably higher (0.370 to 0.575) between the two types (strain-based and time-based) of conflicts within the same direction.

Family resources
Within the family: To assess resources within a family, we used whether respondents had a partner (1 = Yes), appreciation from partner (0 = No partner, 1 = Low appreciation, 2 = High appreciation), and an indicator for the relative bargaining power in the relationship based on the number of working hours and the CASMIN education level (Comparative Analysis of Social Mobility in Industrial Nations): 0 = Balanced (if both work equal hours and CASMIN is equal); 1 = Inferior ([if partner works more hours or if both work equal hours and CASMIN is lower); 2 = Superior (if partner works fewer hours or if both work equal hours and CASMIN is higher); 3 = No partner.

Work resources
Job resources: We considered whether respondents used flexible working hours (1 = Yes) and/or home-based telework (1 = Yes) and whether they had a management position with supervisory responsibilities (1 = Yes). Job autonomy (Breaugh, 1985) was measured using respondents' scores on three items (e.g., "During my working hours, I have control over the sequencing of my work activities") using a 5-point scale ranging from 1 ("I totally disagree") to 5 ("I totally agree"). These values were added to an index that ranged from 3 to 15, with higher values indicating greater job autonomy (Cronbach's α = 0.72), dichotomized to 1 (high autonomy, 11 to 15) vs. 0 (low autonomy, 3 to 10). Also, hourly wages (logarithmized) were assessed. Support from direct supervisors and co-workers: To investigate social support from direct relationships at the workplace, respondents were asked whether their supervisors supported employees in their efforts to reconcile work and family life and whether coworkers helped one another to get their work done when an employee had to leave early or came to work late for personal reasons. Respondents were asked to rate each of these two measures using a 5-point scale that ranged from "Applies completely" to "Does not apply at all" (1 = High support [1 or 2]; 0 = Low support [3 to 5]).
Formal work-family organizational support: We used the number of available formal family-friendly policies out of a list of five possible benefits (e.g., company provides child support or makes home-based telework available), resulting in a range of 0 to 5, dichotomized to 1 (high level of family support, 5) vs. 0 (low or normal level of family support, 0 to 4). We used the number of available health-promotion policies out of a list of four possible policies (e.g., health circles or courses on health-promoting behavior), resulting in a range of 0 to 4, dichotomized to 1 (high level of health promotion, 4) vs. 0 (low or normal level of health promotion, 0 to 3).
Informal work-family organizational support: A family-supportive workplace culture was assessed by asking respondents, using a 5-point scale, to what extent are employees who make use of family-supportive measures viewed as being less committed to the company (1 = Family-supportive [4 or 5]; 0 = Not family-supportive [1 to 3]). Similarly, respondents were asked whether expectations were high within their organizations regarding three behaviors: availability for work, doing extra work, and the ability to withstand stress. The organization was considered to have a low-demand culture if expectations for all three behaviors were low or if only one of the three behaviors was expected (1 = Low-demand; 0 = High-demand).

Co-variates
With respect to health-relevant characteristics of the individual job, we controlled for employees' actual working hours per week, and experiences of physical strain (1 = Yes). Also, we controlled for having a partner (1 = Yes), age of the youngest child (0 = No children or children older than 25 years of age; 1 = ages 0 to 3; 2 = ages 4 to 6; 3 = ages 7 to 11, 4 = ages 12 to 18; and 5 = ages 19 to 25), and employee's age.
In addition, we used gender (1 = Male) and occupational skill level (0= Lower-/semiskilled; 1 = Skilled; 2 = Expert) for descriptive purposes only because time-invariant variables are controlled for in the applied fixed-effects regressions (see chapter 4.3 Analysis strategy).

Analysis strategy
To render these analyses as causally robust as possible, all multivariate models were performed using longitudinal fixed-effects regression models and data from the three waves of the survey (t1 + t2 + t3) to address unobserved heterogeneity as well as reversed causality in the conflict-health relationship (Nohe et al., 2015), for example regarding base line differences in individual mental health. Since fixed-effects modeling has the advantage of controlling for unobserved heterogeneity in time-invariant characteristics (Allison, 2009), time-stable variables on the individual level (e.g., gender, occupation) as well as on the organizational level (e.g., sector, industry) were excluded from the regressions.
Regressions were estimated in two steps. In the first step, effects on mental health of all six of the work-family conflict variables (i.e., WFC, FWC, time-based WFC, strainbased WFC, time-based FWC, and strain-based FWC) were estimated stepwise (see Table  2). In the second step, the buffer hypotheses were tested by including interactions between moderator variables and the WFC variables separately and stepwise for one conflict variable at a time (see Table 3 and the interaction plots in Figures 2 through 6; also see detailed models in Table A3-1 through A3-14 in Online Appendix). Owing to the lack of time variance, family-friendly measures and health-promotion measures offered by the companies were used from wave t1 only. Given that the dependent variable MCS was already standardized, unstandardized coefficients were used to make interpretation of the results more plausible.
In addition, we performed several robustness checks and sensitivity analyses to test the consistency of our results (available on request). First, the multilevel structure regarding the clustering of employees in organizations and in occupations was estimated, but we found only a low degree of variation in each case. Second, to address possible methodological issues due to different time spans in the data collection, we replicated all analyses separately for changes between t1 and t2 and between t2 and t3, respectively. All conclusions remained the same, and the range of minor differences in effect sizes and significant levels can be traced back to differences in the sample sizes. Third, because previous studies were controversial with respect to gender effects (Borgmann et al., 2019;Frone, 2000), we tested for gender differences in the effects of conflict variables on mental health as well as in the moderation effects by using three-way interactions. No gender differences were found. Fourth, we tested the fixed-effects model against the alternative random-effects model; the Hausman test supported use of the fixed-effects model.

Descriptive results
Mental health was slightly better for males than for females and for higher-skilled employees than for lower-or semi-skilled ones (t-test p<0.001). All four dimensions of work-family conflicts were widely experienced among heterogeneous groups of employees in Germany (see Table 1). In comparison, the average levels of WFC were higher than those of FWC. Moreover, time-based WFC was the dimension of conflict experienced most often, whereas time-based FWC was reported least often. Employees in higher status positions, such as managers and skilled workers, had slightly higher mean levels of strain-based and time-based WFC compared with lower-status employees, but there were fewer differences between these two groups for both subdimensions of FWC. Although these results provide evidence in support of the stress of higher status hypothesis (Schieman et al., 2006), conflicts should obviously not be viewed as a risk factor only in particular occupational or status groups; rather, they are a problem of everyday life that affects almost every employee in Germany to some extent. Consequently, possible health risks may affect large parts of the workforce.  Note: Averaged over three panel waves; 'age' was re-coded to a catergorial version for the purposes of presentability; SD = standard deviation; N = 4,920. LEEP-B3 Survey, own calculations.

Negative mental health effects of strain-based/time-based WFC and FWC
Increases in WFC and FWC, as well as in all four subdimensions of these conflicts, significantly predicted a decline in mental health. WFC was a stronger predictor than FWC, and strain-based conflicts were more strongly associated with mental health declines than were time-based conflicts (see Table 2). However, when all four conflict variables were included in one model, the effect of time-based WFC on mental health was no longer statistically significant. Therefore, in direct comparison, strain-based WFC is the dimension of conflict that has the largest health impairing effect. Note: * p< 0.05, ** p< 0.01, *** p< 0.001 (two-tailed test). Standard errors are shown in parentheses; N = 4,920. Controlled for age, actual working hours, physical strain, management position, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations. Table 3 shows an overview of all interactions estimated to test the within-domain and cross-domain buffering hypotheses; a summary of all the interactions tested can be found in the Table 3, detailed interaction models can be found in Table A3-1    High autonomy, high supervisor support, high co-worker support, low-demand culture, family-friendly culture Wages, high co-worker support, healthpromotion, low-demand culture, familyfriendly culture High autonomy, high supervisor support, high co-worker support, low-demand culture, family-friendly culture H3: Family resources buffer WFC (cross- Wages, health-promotion Family-friendly culture High autonomy, health-promotion Note: Significance of moderation based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.05; N = 4,920. WFC = work-to-family conflicts, FWC = family-to-work conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Family resources
If employees have a partner, an increase in strain-based FWC as well as in overall FWC was less strongly associated with decreases in mental health. None of the family resources moderated the effects of time-based FWC. Relative negotiating power, however, was a significant moderator of all dimensions of WFC: An imbalance in the relationship on the part of either the superior or the inferior member can help to mitigate the negative effects of WFC, whereas a balanced relationship even increased the mental health impairment (Fig. 2). With regard to family resources, these results provide partial support for both the within-domain buffering and the cross-domain buffering hypotheses (H1, H3); however, resources from the family domain only marginally mitigated negative mental health effects.

Figure 2:
Interaction plots for WFC#Relative negotiating power Note: Interaction plot based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.01; N = 4,920. WFC = work-to-family conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Work resources
When it comes to work resources, about half the resource indicators provided support for the within-buffering hypothesis (H2). If WFC increases, we found that a high level of autonomy (Fig. 3) and high levels of support from direct supervisors or co-workers (Fig. 4) were resources that mitigated negative effects on mental health. This was also the case for employees in a low-demand work culture or in a family-friendly work culture. Whereas the results were similar for strain-based and time-based WFC, the effect of time-based WFC on health was less severe if the employer offered diverse health-promotion policies.
The availability of such policies can also mitigate the negative effects of strain-based FWC (Fig. 5), and high job autonomy was also a statistically significant buffer of strain-based FWC. In comparison, an increase in wages was found to be most beneficial in mitigating FWC: If FWC increased but wages did also, the negative effects of FWC could be buffered almost completely as the wages got higher (Fig. 6). However, none of the work resources seemed to moderate the effects of time-based FWC on mental health. In sum, these results only moderately support the cross-domain buffering hypothesis in the area of work resources (H4). All in all, we found moderate support for all four of our hypotheses, both for withindomain and cross-domain buffering. However, within-domain buffering proved to be more effective than between-domain buffering in mitigating work-family conflicts.

Figure 3:
Interaction plots for strain-based WFC#autonomy Note: Interaction plot based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.001; N = 4,920. WFC = work-to-family conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Figure 4:
Interaction plots for time-based WFC#support from co-workers Note: Interaction plot based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.001; N = 4,920. WFC = work-to-family conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Figure 5:
Interaction plots for strain-based FWC#health-promotion Note: Interaction plot based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.10; N = 4,920. FWC = family-to-work conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Figure 6: Interaction plots for FWC#wages
Note: Interaction plot based on linear fixed-effects regressions with interaction terms and predicted margins; p<0.10; N = 4,920. FWC = family-to-work conflicts. Controlled for controlled for age, actual working hours, physical strain, hourly wages, age of youngest child in household. LEEP-B3 Survey, own calculations.

Discussion and conclusion
Work-family conflicts need to be understood as a tolerated and often unavoidable part of role expansion that originate from employees' attempts to pursue life goals in both the family and the work domains (Grönlund & Öun, 2010). Our study aimed to tackle the question of how employees can still experience good mental health in spite of experiencing such conflicts. To add to previous knowledge, we used a longitudinal approach to investigate the within-domain and cross-domain buffering potentials of family resources and work resources and to explore how strain-based and time-based WFC and FWC are related to impaired mental health. We confirmed that the experience of WFC and FWC is indeed widespread among German employees across various work and family life situations. All four subdimensions of work-family conflicts -strain-based and time-based WFC and FWC -proved to be significant risk factors when it came to employees' mental health within a heterogeneous workforce. WFC was a stronger predictor than was FWC, and when directly compared, strain-based conflicts in both directions were more harmful for mental health than were time-based conflicts. Thus, our findings confirm the results of previous studies concerning the conflict-health relationship (e.g., Cooklin et al., 2016;Jensen & Knudsen, 2016;Kinnunen et al., 2004) by using a longitudinal design to study the causal direction of the relationship (Nohe et al., 2015). Moreover, we provide evidence that the still under-researched conflicts in the family-to-work direction should not be negelected when it comes to health consequences. We also found that the strength of the conflict-mental health relationship depended on within-domain buffering and cross-domain-buffering by resources both in the family and at the workplace. Only minor support was found for the buffering potential of job resources with the exceptions of high levels of job autonomy and higher wages. In contrast, we found interesting results for social relations within the family, for support from direct supervisors and co-workers in the workplace, and for the organizational environment in the form of formal and informal work-family organizational support.
In terms of the family, it turns out that having a partner helps to mitigate both FWC and WFC. However, an imbalanced negotiation relationship can be of advantage in using the resources that are available within the partnership, whereas a balanced relationship tends to increase the negative mental health effects of FWC and WFC. These results might indicate that when couples have a similar occupational status and both work fulltime, the partnership can be overloaded by the role strain experienced by both partners (Hill et al., 2006), and that a surplus of resources accumulated within the partnership does not sufficiently outweigh the accumulated strains to prevent negative consequences. In line with Becker (Becker, 1991), the traditional division of labor as a form of specialization within the household, as well as the avoidance of competition within the marriage (or partnership), seems to help in bundling resources to handle conflicts. However, this view has important implications. Current research on both the family and on the work-life interface shows that the traditional patterns in the division of labor within the family are increasingly undesirable: Dual-earner couples and dual-career families (Hammer et al., 2005;Yucel & Fan, 2019), single parents (Reimann et al., 2019), and fathers who want to engage in childcare (Graham & Dixon, 2014) are only some of the most commonly discussed developments that foster the increasing importance of an egalitarian or at least somewhat alternating division of labor within families. Consequently, the likelihood of WFC or FWC increases. But families, especially those with dual-career partnerships, are obviously unable to stem the risks that accompany these conflicts. Thus, individualizing the responsibility to buffer conflicts cannot be the solution if individual resources are overwhelmed. However, institutions that are still based on the traditional division of labor may not support families in handling conflicts either. Therefore, institutional solutions are necessary, on the side of both social policy and organizations. Though our study did not focus on distributional differences across different family forms or among women and men, it is already known from previous research that they largely differ in family and/or work demands/resources. For instance, though mothers still tend to shoulder most care demands, they also have less negotiation power and lower resources in the family domain as compared to fathers (Duxbury & Higgins, 1991;Gutek et al., 1991;Nomaguchi, 2012), and this applies especially to single parents despite their increased need for resources (Reimann et al., 2019). In the work domain, men are more likely than women to work in positions with higher status (Stojmenovska & England, 2021) and thus are more likely to have higher job autonomy as well as higher negotiation power to choose flexibility options (Wheatley, 2017). And even though women tend to choose occupations that offer formal flexible work arrangements more often, they receive fewer other resources. However, as the same resources can be helpful in avoiding WFC and FWC in the first place, but also in coping with conflicts when they did occur, further research is needed to clarify whether this might imply a double burden for women and especially mothers.
In terms of the workplace and organizational environment, supportive direct relationships with co-workers and supervisors, as well as a family-supportive or lowdemand workplace culture, can mitigate the negative effects of conflicts that emerge within the workplace. Health-promotion policies also have the potential for cross-domain buffering: They help prevent conflicts that were brought into the work sphere from the employees' family life. It must be kept in mind that these data were derived from large organizations only. By covering diverse employee groups, our dataset offers considerable advantages over previous studies that focused on very specific occupational groups or sectors (Allen et al., 2000;Nohe et al., 2015), but the conditions in small or medium-sized organizations may be very different, especially regarding the availability of work-family supportive and health-promotion policies. Nevertheless, our results have practical implications for employers as well. Among possible consequences, mental health plays a prominent role as an outcome, since, in the longer run, health is not only an outcome but also a prerequisite to coping successfully with the stressful task of reconciling competing demands in a person's work and family life (Bergs et al., 2017). Thus, mitigating the negative consequences of work-family conflicts is important for circumventing the onset of vicious circles. Both work-family conflicts and impaired employee health can be linked to poorer performance. If organizations have high expectations with respect to employees' availabiliy and flexibility, work-family conflicts are inevitable. Therefore, organizations also need to support their employees to prevent negative health consequences if they value and want to preserve employability.
Our findings have further implications for the current debates in family and work-life research. An obvious question is how family and work conditions can be both antecedents of conflicts and moderators of the conflict-health relationship at the same time. As our longitudinal analysis demonstrates, resources are indeed also moderators of pre-existing conflicts. Consequently, a focus that is solely on avoiding such conflicts by means of these resources would be far too simplistic and one-sided, since conflicts cannot be prevented completely in the first place and to some degree are even accepted by both employees and employers. Therefore, the typical antecedents of conflicts (i.e., long working hours with regard to WFC or caring for children with regard to FWC, Michel et al., 2011) might not be conditions they would want to change if they have ambitious goals in different areas of their lives. As a consequence, finding ways to successfully deal with conflicts becomes necessary if they want to stay healthy over time. In addition, though some family and work resources tend to have dual benefits in avoding conflicts in the first place as well as in buffering their negative health consequences, this does not seem to be an universal value among all resources. Even more so, some family and work characerstistics that appeared to be assets in the first instance (e.g., an egalitarian power relationship among partners), may impede dealing with already existing conflicts. Thus, further work should continue to draw a differentiated picture of which family and work conditions are beneficial as resources for avoding mental health risks and which ones turn out to be additional demands in the end.
One limitation of the present research is that the time span of all three panel waves was only about 6 years, which, from a life-course perspective, is just a small part of a lifetime during which the integration of family and work life is a crucial topic. A willingness to tolerate conflicts as well as the availability of resources may change over time, and the accumulation of previous experiences of conflicts may also play a part in confronting potential health consequences. Then again, impaired mental health due to work-family conflicts may also have long-term negative consequences as the health resources required for facing demands in family and work life are reduced. Thus, future research should look at the possible long-term effects of work-family conflicts over longer periods of the life course.

Data availability statement
The datasets analyzed during the current study are not publicly available due to data restrictions by the Federal Institute for Employment Research (Institut für Arbeitsmarkt und Berufsforschung (IAB). Data are only available on request for analyses to be conducted locally at Bielefeld University in cooperation with project members. Requests to access these datasets should be directed to martin.diewald@uni-bielefeld.de.