Introduction

Work practices and organizational lives have become increasingly diverse in the past few decades, giving rise to a multitude of flexible work arrangements where employment is not restricted to one particular place or to standard work hours. Despite a growing literature showing the value of workplace flexibility for accommodating employees’ needs to balance work, leisure, and family as well as for reducing gender inequalities (Carlson et al., 2010; Fan et al. 2015; Golden et al. 2006; Grzywacz et al. 2008; Kelly et al. 2014; Moen et al. 2016), there are also studies raising caution for the possibility of flexibility to fuel heightened job demands and to widen gender gaps in work and family outcomes (Chesley, 2014; Schieman & Glavin, 2008; Schieman & Young, 2010; Voydanoff, 2005). The ongoing COVID-19 pandemic, along with the remote work natural experiment it ushered in, has accelerated transformations in conventional work arrangements (Barrero et al., 2021; Fan and Moen 2022). But do flexible work arrangements indeed benefit workers’ emotional health? And if so, which type of flexible work arrangements, through what mechanism, and for which subgroup(s) of the working population? Understanding these questions is critical to developing health-promoting organizational and public policy in the face of a growing number of workers calling for more flexibility in the workplace (Parker et al., 2020).

To address these questions, we draw on data collected from the 2008 U.S. National Study of the Changing Workforce (NSCW) to investigate four questions. First, does access to flexible work arrangements (FWA) predict lower levels of psychological distress? We exploit the rich measures in the NSCW to examine access to three major types of FWA: flextime, flexplace, and culture of flexibility in the workplace. Despite the complexity involved in measuring well-being (Maggino & Facioni, 2017; Maggino & Zumbo, 2011), we focus on psychological distress given its centrality in shaping the quality of life. Second, do conflict and enrichment between work and family (in both directions) mediate the relationship between access to FWA and psychological distress? Third, does the relationship between access to FWA and psychological distress differ between workers with and without family caregiving responsibilities? Lastly, does the moderating effect of family caregiving responsibilities further differ between women and men?

Answering these questions allows us to contribute to the literature on work, family, and well-being in three ways. First, despite a burgeoning literature on flexible work arrangements and their implications for workers’ well-being (Allen et al., 2013; Fan and Moen 2023; Kossek and Lautsch 2018; Moen et al. 2016), with a few exceptions (Kim et al., 2020; Mennino et al., 2005; Russell et al. 2009), studies have generally focused on a specific type of flexible arrangements rather than comparing across alternative arrangements. We bridge this gap by assessing whether and in what ways different types of FWA matter for employee’s psychological distress. Empirically examining such variation is important because it sheds light on the type of FWA that is most likely to deliver desirable outcomes for workers. It also contributes to a deeper understanding of why some FWA works whereas others do not, thereby inspiring an evidence-based reimagination of healthy and happy workplaces. Second, drawing on the rich measures of the NSCW data, we begin to unpack the black box of FWA and well-being. We focus on one mechanism that has been frequently theorized in prior studies to explain the well-being implications of FWA—work–family conflict and enrichment (Kelly & Moen, 2021; Kossek et al., 2006; Maruyama & Tietze, 2012). Even though some studies have tested the mediating role of work–family conflict (e.g., Moen et al. 2016), far fewer have explicitly tested for both work–family conflict and work–family enrichment (in both directions), a task we tackle in this study.

Third, we empirically evaluate whether family caregiving obligations—including both caring for preschool children and caring for elders—shape the effect of FWA on psychological distress. Family caregiving has become an increasingly prevalent home demand that touches the lives of many U.S. workers, as Joan Williams (2020) argued in a recent Harvard Business Review essay, “today, a key divide is between parents and non-parents.” Similar to childcare, elder care has become increasingly common given the growing share of older adults (Schulz et al., 2020; Spillman et al. 2020). Accordingly, we examine whether having access to presumably more family-friendly work arraignments benefits care-providing workers the most. We further look at the intersection of family caregiving responsibilities in combination with gender, in view of the gendered stress process when it comes to the competing demands of work and family (Blair-Loy, 2009; Moen & Roehling, 2005; Williams, 2010).

Access to FWAs and mental well-being: previous studies

Prior studies on flexible work arrangements (FWA) have generally showed positive mental well-being outcomes among workers who have access to such arrangements. Drawing on a group-randomized trial conducted in a Fortune 500 IT firm, for example, Moen and her colleagues (2016) provide strong evidence that an intervention that promotes greater employee control over work time and work location leads to reduced burnout, perceived stress, and psychological distress. Using data collected from several businesses across a variety of industries, Grzywacz and his colleagues (2008) likewise report a negative relationship between engaging in formal flexible arrangements (flextime and compressed workweek) and stress and burnout. A similar finding is shown between perceived flexibility and family-to-work conflict, stress, and burnout (Hill et al., 2008).

Several reasons may account for the positive association between FWA and well-being. First, by offering workers greater autonomy in where, when, and how they work, FWA can help working families resolve work–family conflicts that are either caused or exacerbated by the incompatibility of work and family schedules and demands; a similar process can also operate for work–family enrichment (Fan and Moen 2023; Kelly and Moen 2021; Kossek et al. 2006; Maruyama and Tietze 2012). A second reason for the positive relationship between FWA and mental well-being is that organizations providing FWA benefits are likely perceived positively by workers, which, in turn, can produce positive employee outcomes. Indeed, one long-held finding in the social support literature is that perceived availability of social support oftentimes has a stronger effect on mental health than actual receipt of social support (Thoits, 2011). Consistent with this thesis, some research shows that access to flexibility policies can have a greater impact on workplace outcomes than actual usage of these policies (e.g., Allen et al. 2013). Accordingly, some scholars compare flexibility to an insurance policy—just knowing that flexibility is there for them appears to be sufficiently reassuring to employees (Galinsky et al., 2011).

In contrast to the first two explanations that view FWA as intrinsically valuable, the third explanation emphasizes the uneven provision of and access to flexibility across social groups. An accommodation model typically dominates in practice, in which a flexible work arrangement needs to be negotiated individually by an employee with their manager (Kelly & Moen, 2021; Perlow & Kelly, 2014). As a result, socioeconomic status such as educational attainment, along with important social locational markers such as gender, age, race and ethnicity, affect access to flexibility profoundly (Golden, 2001, 2008; Glass & Noonan, 2016; Kossek & Lautsch, 2018). The uneven social processes result in deeply skewed advantages for privileged workers, with men, whites, the more educationally advantaged, and those in professional and managerial jobs more likely to enjoy various forms of flexible work arrangements. The selective nature may confound the true well-being benefits associated with FWA. To examine and control for selection to the extent possible, we investigate the relationship between access to flexible work arrangements and mental well-being while adjusting for a wide range of socio-demographic and socio-economic characteristics that are predictive of FWA access.

We also note that flexible work arrangements are not found to promote well-being in all studies. One factor that may complicate the well-being implications of FWA is job flexibility being geared to business rather than employee needs, thereby further blurring the boundaries between work and home (Chesley, 2014; Golden et al., 2006; Schieman & Glavin, 2008; Schieman & Young, 2010; Voydanoff, 2005). With communication technologies making it possible and often expected to be available 24/7, employees with FWA may only find it leads to greater intrusion of work into family life (Chesley, 2014; Kelly & Moen, 2021) and even more distress.

Access to FWAs, work–family conflict and enrichment, and mental well-being

Given prior findings, our first research question examines whether access to flexible work arrangements is associated with lower levels of psychological distress. We build on and contribute to previous studies by examining the respective well-being implications of three different types of FWA: flextime (when work is completed), flexplace (where work is completed), and culture of flexibility in the workplace.

With a few exceptions, existing studies on workplace flexibility have largely ignored the potential heterogeneous well-being effects of different arrangements. One notable exception is a study conducted by Russell and her colleagues (2009), who use a 2003 national survey of employees in Ireland to show that the well-being effects differ by type of flexibility. While part-time work and flextime tend to reduce work pressure and work–life conflict, working from home is associated with greater levels of work pressure and work–life conflict. Similarly, a recent study by Kim et al. (2020) shows mixed patterns. Having the ability to take time off during the workday to attend to personal or family matter benefits workers, evidenced by their greater job satisfaction and lower job stress, daily fatigue, and work-to-family conflict. In comparison, flexibility in daily start and end times—along with working from home—shows inconsistent effects on worker well-being. In another study that compares availability of flexibility policies and a flexible workplace atmosphere, Mennino et al. (2005) reveal that family supportive workplace cultures matter more than availability of company policies (such as flextime) in lowering workers’ negative work–home spillover. Combined, these studies indicate a clear need to not treat flexible working arrangements as a package deal but to consider its discrete effects.

Given these findings, we expect a flexible workplace culture to have the most potential to deliver emotional well-being benefits, followed by flextime, whereas flexplace may have smaller benefits. This expectation is consistent with the “flexibility stigma” literature (Blair-Loy & Wharton, 2002; Williams, 2010), which shows that employees who have access to flexible arrangements may nevertheless believe they would pay a price if they use it, with the stigma even more strongly attached to employees who ask for flexplace accommodations than those asking for flextime accommodations (Munsch et al., 2014). In addition, as described above, flextime or flexplace may in practice lead to longer working hours and unpredictable schedules that are not in the best interests of many workers (Fan and Moen 2022; Noonan and Glass 2012; Lewis et al. 2007). Combined, the mere existence of FWA policies may not be sufficient, if workers do not perceive their organizational and workplace culture to be truly supportive when work–life issues arise (Allen, 2001; Kelly & Moen, 2021). Therefore, we hypothesize that:

Hypothesis 1: Compared with access to flextime or flexplace, a flexible workplace culture is more likely to be associated with lower levels of psychological distress.

To the extent that any of the three FWAs—flextime, flexplace, or culture of flexibility—predicts lower levels of psychological distress, we expect work–family conflict and work–family enrichment to mediate these associations. Previous studies have largely shown beneficial effects of flexible work arrangements on work–family outcomes such as work–family conflict (Butts et al., 2013; Carlson, Grzywacz, and Kacmar 2010; Kelly et al. 2014). For example, drawing on experiment data collected from a U.S. IT workplace, Kelly and her colleagues (2014) report that participating in an initiative that provides employees with more control and support over when and where they work leads to reduced work-to-family and family-to-work conflict. Similarly, a meta-analysis shows a significant association between flexible work cultures and work-to-family conflict as well as family-to-work conflict (Mesmer-Magnus & Viswesvaran, 2006). Given that work-to-family conflict and family-to-work conflict have long been shown to precipitate negative well-being outcomes (Greenhaus & Allen, 2011; Grzywacz, 2000), we expect them to mediate the relationship between FWA and psychological distress.

Previous research indeed shows that flextime, by enabling workers to adjust the starting or ending times of work or allowing them to work a schedule different from the traditional 9 − 5 one, makes it easier for workers to arrange their work schedules around their home demands (Gajendran & Harrison, 2007; Golden et al., 2006; Versey, 2015), thereby leaving more time for family life and other activities (Kelly & Moen, 2021). Similarly, flexplace—by empowering workers the ability to adjust the location of work—reduces pressures associated with commuting time and housing costs (Evans et al., 2002; Hansson et al., 2011; Hill et al., 2001; Hilbrecht et al., 2008), results in less work–family conflict, and contributes to increased job satisfaction and quality of life (Bloom et al., 2015; Fan and Moen 2023; Gajendran and Harrison 2007; Golden et al. 2006).

Parallel to work–family conflict, given the positive association between flexible work arrangements and work–family enrichment (Chen et al., 2018; Carlson et al., 2010) and the roles of work-to-family and family-to-work enrichment in promoting well-being (Grzywacz, 2000), the relationship between FWAs and subjective well-being is likely mediated by work–family enrichment. Using a sample of 726 employees in China, Chen et al. (2018) show that work–family enrichment mediates the relationships between flexible work arrangements (flextime and compressed work week) and job satisfaction. A similar finding is reported in a U.S.-based study (McNall et al., 2010), as well as in other research which finds that work–family enrichment mediates the relationship between FWAs and a wide set of outcomes including organizational commitment, employee engagement, family satisfaction, and burnout (Carlson et al., 2010; Ivanauskaite, 2015; Wattree, 2020). These studies provide valuable insights, though almost all of them are based on non-representative, convenience samples. Extending these studies by using nationally representative data, we hypothesize that:

Hypothesis 2: Work–family conflict and work–family enrichment mediate the relationships between access to flextime, flexplace, or culture of flexibility and psychological distress.

Access to FWAs and mental well-being: differences by family caregiving responsibilities and gender

The association between FWAs and mental well-being likely differs between workers with and without family caregiving responsibilities. Given that many flexible work programs are developed to assist workers in managing the demands of paid work and unpaid care work, we expect workers who provide care to children, elderly relatives, or other adults who are unable to care for themselves to derive the most emotional well-being benefits of FWAs. Previous studies provide supportive evidence, showing that workers most in need of flexible policies—including parents of young children and caregivers for the elderly or disabled (Michel et al., 2011)—are most likely to use such policies (e.g., Blair-Loy and Wharton 2002) and possibly benefit the most from such policies (Erickson et al., 2010; Ruppanner, Lee, and Huffman 2018).

Hypothesis 3: The negative associations between access to flextime, flexplace, or culture of flexibility and psychological distress are stronger for workers with childcare or elder-care responsibilities.

We further expect childcare or elder-care obligations to intersect with gender to moderate the relationship between FWA and psychological distress, given that expected roles and practices around work and family are heavily shaped and guided by gender norms, which relegate most of the family care work to women (Acker, 1990; Blair-Loy, 2009; Moen & Roehling, 2005; Williams, 2010). In light of the still prevalent male-breadwinning/female-homemaking norm, women are socially expected to create and sustain a satisfactory family life, whereas men usually assume a breadwinning role. Therefore, women with heavy caregiving responsibilities, such as women in the sandwich generation, may benefit the most from access to FWA because they are in particular need of flexible arrangements to combine work, family, and personal lives. Consistent with this expectation, analyses of an organizational intervention designed to promote greater employee control over work time and place of work reveal that the intervention brought greater benefits to employees with heavier family demands—women and mothers in particular (Kelly et al., 2014; Moen et al., 2016). Similarly, other research shows that workplace flexibility—having control over when and where to work—is associated with a greater reduction in work–family conflict especially for mothers (Hill et al., 2008).

Conversely, however, the tensions and contradictions between the ethic of care and employment responsibilities (Blair-Loy, 2004), as well as the intensive mothering ideology (Hays, 1996), may limit the extent of benefits women with heavy caregiving responsibilities gain from FWA. Women with access to flexplace, for example, are found to replace the time saved from commuting with childcare and household chores rather than personal leisure (Hilbrecht et al., 2008), compared with non-remote working women (Noonan et al., 2007). Given these compensating mechanisms, women with intensive family caregiving obligations may not derive as much emotional benefit from FWA as men in general or non-care-providing women (Fan and Moen 2023). In light of these alternative possibilities, we tentatively hypothesize that:

Hypothesis 4: The moderating effects of childcare or elder-care responsibilities on the associations between flextime, flexplace, or culture of flexibility and psychological distress differ between women and men.

Data and sample

This study uses data from the 2008 wave of the National Study of the Changing Workforce (NSCW). The NSCW is an ongoing nationally representative study of the U.S. workforce. Building on the U.S. Department of Labor’s 1977 Quality of Employment Survey, the NSCW was designed and conducted by Families and Work Institute in 1992, 1997, 2002, 2008, and 2016. We use the 2008 wave for this study, which includes information from 2,769 workers. We limit the sample to workers who are between 18 and 64 years old (n = 2,577), and remove 260 workers who do not have a supervisor because questions related to flexibility culture were not asked for these workers. Similarly, we remove 69 workers who mainly work from home because these workers were not asked question on access to flexplace. After further removing 15 respondents who have missing values on psychological distress, our final sample consists of 2,233 workers.

Dependent variable

We use psychological distress as our outcome variable (for methodological challenges in well-being studies, see Maggino and Facioni 2017; Maggino and Zumbo 2011). Psychological distress is constructed based on five questions. Three questions ask respondents how often in the last month they have (1) been bothered by minor health problems such as headaches, insomnia, or stomach upsets, (2) had trouble sleeping to the point that it affected their performance on and off the job, or (3) felt nervous and stressed (Minnotte et al., 2013; Schieman & Glavin, 2011; Voydanoff, 2005). Response options range from 1 (never) to 5 (very often). In addition, respondents are asked whether, during the past month, they have been bothered by (4) feeling down, depressed, or hopeless, and (5) having little interest or pleasure in doing things. The answer categories range from 1 (yes) to 2 (no), which we reverse code. Given the different answer categories, we standardize and then average these five items to create a composite scale, with higher values indicating higher levels of psychological distress. The Cronbach’s alpha of this scale is 0.75, indicating high internal reliability.

Independent variables

Our main independent variable is workplace flexibility, operationalized as access to flextime, access to flexplace, and culture of flexibility. To construct these three measures, we follow Galinsky et al. (2011) who provide details on the rationale of these measures and the NSCW items used to create these measures. Access to flextime is measured by three questions: (1) Are you allowed to choose your own starting and quitting times within some range of hours? (2) Are you able to temporarily change your starting and quitting times on short notice when special needs arise (if you check with your supervisor or manager)? (3) Are employees in your organization allowed to work a compressed workweek for part or all of the year? Responses to all three items are coded as dichotomous variables (1 = Yes, 0 = No); we reverse code and sum the responses to create a scale for access to flextime, with higher values indicating greater access to flextime. The Cronbach’s alpha is 0.73, indicating high internal reliability.

Access to flexplace is measured by the question “Are you allowed to work part of regular paid hours at home?” We distinguish those who responded “yes” to the question from those who responded “no.”

The third dimension of FWA—culture of flexibility—is measured by seven questions following Galinsky and her colleagues (2011). Two questions concern the workplace: (1) At my place of employment, employees have to choose between advancing in their jobs or devoting attention to their family or personal lives. (2) At the place where you work, employees who ask for time off for personal or family reasons or try to arrange different schedules or hours to meet their personal or family needs are less likely to get ahead in their jobs or careers. Five additional questions are about supervisors: (3) My supervisor or manager is fair and doesn’t show favoritism in responding to employees’ personal or family needs. (4) My supervisor or manager is responsive to my needs when I have family or personal business to take care of—for example, medical appointments, meeting with child’s teacher, etc. (5) My supervisor or manager is understanding when I talk about personal or family issues that affect my work. (6) I feel comfortable bringing up personal or family issues with my supervisor or manager. (7) My supervisor or manager really cares about the effects that work demands have on my personal and family life. The answer categories for each item range from 1 (strongly agree) to 4 (strongly disagree). After reverse coding items 3–7, we obtain the average of the seven items, with higher values representing a more supportive culture of flexibility. The Cronbach’s alpha is 0.80, indicating high internal reliability.

Mediating variables

We use work–family conflict (in both directions) and work–family enrichment (in both directions) as mediating variables. Following prior research (Minnotte et al., 2015; Yucel 2017; Jung Jang et al. 2012), the work-to-family conflict (WFC) scale is based on five items (Minnotte et al., 2015). Some example items are “In the past three months, how often has work kept you from doing as good a job at home as you could?” or “In the past three months, how often have you not been in as good a mood as you would like to be at home because of your job?” (1 = very often to 5 = never). We reverse code these items and obtain the average to create a scale for work-to-family conflict (Cronbach’s alpha = 0.86).

Similarly, five items are used to create the scale for family-to-work conflict (FWC) (Minnotte et al., 2015). Some examples of the items are “In the past three months, how often has your family or personal life drained you of the energy you needed to do your job?” or “How often has your family or personal life kept you from concentrating on your job?” (1 = very often to 5 = never). We reverse code these items and obtain the average to create a scale for family-to-work conflict (Cronbach’s alpha = 0.82).

In addition, we use two items to create the work-to-family enrichment scale (WFE) (Yucel 2017; Hill 2005). One sample item is “In the past three months, how often have you been in a better mood at home because of your job?” (1 = very often to 5 = never). After reverse coding, we average these two items to create a scale for work-to-family enrichment. In parallel, for the two-item family-to-work enrichment scale (FEW) (Yucel 2017; Hill 2005), one sample item is “In the past three months, how often have you had more energy to do your job because of your family or personal life?” (1 = very often to 5 = never). We reverse code these items and obtain the average to create a scale for family-to-work enrichment.

Moderating variables

Our analysis tests two moderating variables: gender (1 = male, 0 = female) and family caregiving obligations, including both child care and elder care. Specifically, to capture preschool child care responsibilities, we distinguish between respondents who have a preschool child living in the household and those who do not. Elder care responsibility is coded 1 for respondents who cared for any relative/in-law older than 65 years of age in the past five years. We then create a three-category variable consisting of (1) those with preschool child at home and caring for elders (“sandwich generation”), (2) those with preschool child at home or caring for elders (but not both), and (3) those with neither childcare nor elder-care responsibilities (the reference category).

Control variables

Following prior research (Jung Jang et al. 2012; Ray and Pana-Cryan 2021; Yucel and Fan 2019), we control for the following variables that are related to workplace flexibility and/or psychological distress: age, educational attainment, race, work hours, work schedule, occupation, personal income (logged), relationship status, partners’ employment status, and number of minor children living in the household. Age is treated as a continuous variable. We experimented with adding a squared term of age in the models but it was not significant; we therefore do not include age squared in our analysis. Education is measured as the highest level of schooling, including high school graduate or less (reference), some college, Bachelor’s degree, and more than Bachelor’s degree. We distinguish three racial categories that respondents identify themselves with: White (reference category), Black (Black or African American), and Other (non-White, non-Black racial groups). Work hours measures respondents’ usual hours worked per week, work schedule indicates whether respondents work in regular hours, and occupation is a dichotomous variable distinguishing those in a professional or managerial occupation from those in other occupations. Respondents’ income is measured as hourly earnings at their main job. This variable is skewed, so we take the log before entering it into our models. Relationship status is a dichotomous variable, distinguishing those living with a partner or spouse from those in other arrangements. We also include a dichotomous variable to denote whether respondents’ partner or spouse has a paid job. Lastly, we control for the total number of children who are younger than 18 years old living in the household (0, 1, and 2 or more).

Analytical strategy

Descriptive findings for our independent and dependent measures, as well as control variables, are displayed in Table 1, first for the whole sample and then for men and women separately. To determine whether there are any significant differences between men and women in continuous and categorical variables, we use t-tests and chi-square tests, respectively. The analytical steps of the multivariate analyses are as follows. First, given that our outcome variable, psychological distress, is continuous, we use OLS regression models to estimate the total effect of each dimension of FWAs on psychological distress (without the mediators) (Model 1 in Table 2).

Table 1 Descriptive Statistics for the Main Dependent and Independent Variables (N = 2,233)
Full size table
Table 2 Parallel Multiple Mediation Analyses from PROCESS MACRO (N = 2,233)
Full size table

Second, we use Model 4 from the PROCESS macro to conduct parallel mediation analysis (for an example of its application, see Cuc et al. 2022). The PROCESS macro, part of the SPSS software, is developed by Hayes (2013). The PROCESS macro has several advantages over conventional mediation methods. First, it allows us to examine the effects of multiple mediating variables simultaneously. In comparison, conventional methods for mediation analyses, such as Baron and Kenny’s approach (1986) or Sobel’s test (1982), cannot test for parallel mediation where multiple mediators (such as what we have in this study) are included in the analyses simultaneously.Footnote 1 Second, it allows us to decompose the mediation pathways, testing both the effects of our key independent variables (i.e., each dimension of workplace flexibility) on mediators and the effect of each mediator on psychological distress. Third, the program computes unstandardized and standardized coefficients for the direct, indirect, and total effects, making it easy to assess the mediation process and to compare the magnitude of the indirect effects associated with different mediators (Abu-Bader & Jones, 2021). Given that the PROCESS macro requires complete data, we first use Stata’s ice command (Royston, 2005) to impute the missing cases. Next, using 5,000 bootstrapped samples, we run parallel multiple mediation analyses where all four mediators (i.e., WFC, FWC, WFE, and FWE) are included in the analyses simultaneously. We examined the variance inflation factors (IVFs) of the four mediators before conducting the mediating analysis, finding multicollinearity is not an issue (results not shown but are available upon request). The results showing mediation analyses based on PROCESS macro are displayed in Table 2 (see Models 2–6). Model 2 shows the direct effect of each dimension of FWAs on psychological distress after adjusting for all four mediating variables, and Models 3–6 show the effect of each dimension of FWAs on each mediator separately.

Finally, for the moderation analyses, we use Models 1 and 3 from the PROCESS macro (for applications, see Rey et al. 2019; Chen et al. 2021),Footnote 2 with results presented in Table 3. We first add the interaction terms between each dimension of workplace flexibility and family caregiving responsibilities (see Models 1–3 in Table 3). Next, we present estimates for the three-way interaction terms between each dimension of workplace flexibility, family caregiving responsibilities, and gender (see Models 4–6 in Table 3).

Table 3 Moderation Analyses from PROCESS MACRO (N = 2,233)
Full size table

Results

Descriptive findings

On average, women report significantly higher levels of psychological distress compared with men (0.08 among women versus −0.10 among men—recall that psychological distress is centered at 0). 20% of our sample have access to flexplace, that is, they are allowed to work at least part of their regular paid hours at home. On average, respondents report relatively low levels of access to flextime (1.23 on a 0–3 scale) but tend to perceive the culture in their workplace as flexible (3.16 on a 1–4 scale). No gender difference is found in the reported workplace flexibility culture, but men seem to have greater access to flextime (1.31 versus 1.17, p < .01) and flexplace (24% versus 16%, p < .001) relative to women.

Respondents in the sample report moderate levels of work-to-family conflict (2.57 on a 1–5 scale), lower levels of family-to-work conflict (2.11 on a 1–5 scale), and moderate-to-high levels of work-to-family enrichment (2.68 on a 1–5 scale) and family-to-work enrichment (3.17 on a 1–5 scale). Only for family-to-work conflict do we find a significant gender difference, with women reporting higher levels of such conflict compared with men (2.15 versus 2.07 among men, p < .01). 44% of our sample have neither preschooler nor elderly care responsibilities, slightly more than half (51%) of the sample either have a preschooler living at home or provide care for elders, and 5% of the sample are sandwiched between preschooler childcare and elder care responsibilities.

For the overall sample, 45% are men. Whites make up 82% of our sample, 8% of the sample are Blacks, and the remaining 10% are respondents from other racial groups. On average, respondents are around 45 years old, with 67% of the sample married and living with their spouse (especially among men). Respondents have on average one minor child living in the household. About two out of five (44%) respondents have a college or advanced degree. The average work hours is 41 h per week, with men putting in significantly longer hours than women (43 versus 39, p < .001), and about 21% report working a shift schedule. Half of the respondents in our sample have a partner or spouse who is employed. Men earn higher income than do women, whereas more women than men in our sample are employed in professional or managerial occupations.

Predicting psychological distress—main effects and mediating effects

Model 1 in Table 2 presents estimates of the total effects (without adjusting for the mediating variables). Having access to flextime or flexplace is not associated with psychological distress, but a flexible culture in the workplace is associated with workers’ lower levels of psychological distress (b=-0.327, SE = 0.026, p < .001). Hypothesis 1 is therefore supported.

We then add the mediating variables. As shown in Model 2 in Table 2, the direct effects for flextime and flexplace remain non-significant after adding mediators. In comparison, the direct effect for culture of flexibility remains significant, predicting lower levels of psychological distress (b=-0.116, SE = 0.025, p < .001), though note that the magnitude of the effect is reduced by almost two-thirds after adding the mediating variables (-0.116 relative to -0.327).

Next, in Models 3–6, we show the effects of flextime, flexplace, and culture of flexibility on each of the mediating variables. Flextime is associated with higher levels of work-to-family enrichment, whereas flexplace is associated with higher levels of family-to-work conflict. Given the null total effect for flextime or flexplace, we do not test for mediation effects for these two dimensions of FWA.

Different from access to flextime or flexplace, culture of flexibility is a significant predictor of all mediators. Coupled with the finding that higher work–family conflict (in both directions), as well as lower work–family enrichment (in both directions), are all associated with higher levels of psychological distress, work–family conflict and enrichment are significant mediators of the relationship between culture of flexibility and psychological distress. Specifically, the indirect path through work-to-family conflict is -0.133 (95% bootstrap CI [-0.161, -0.106]), the indirect path through family-to-work conflict is -0.046 (95% bootstrap CI [-0.064, -0.030]), the indirect path through work-to-family enrichment is -0.023 (95% bootstrap CI [-0.039, -0.008]), and the indirect path through family-to-work enrichment is -0.009 (95% bootstrap CI [-0.017, -0.001]). These estimates indicate that work-to-family conflict, family-to-work conflict, work-to-family enrichment, and family-to-work enrichment all partially mediate the effect of culture of flexibility on psychological distress. Thus, Hypothesis 2 is partially supported.

We also compared the size of the four indirect effects using standardized coefficients. Results showed that the two conflict mediators (work-to-family conflict and family-to-work conflict) play a larger role in mediating the relationship between culture of flexibility and psychological distress compared with the two enrichment mediators (results not shown but are available upon request).

Moderating effects by family caregiving responsibilities and gender

Table 3 tests whether child/elder care responsibilities moderates the effects of flextime, flexplace, and culture of flexibility on psychological distress. Models 1–3 show results testing the moderating role of family caregiving responsibilities. The insignificant interaction terms between flextime or flexplace and caregiving responsibilities indicate that the effect of flextime or flexplace on psychological distress does not vary by respondents’ responsibilities regarding child care or elder care (see Models 1 and 2). However, Model 3 shows a negative interaction term between culture of flexibility and being sandwiched between elder care and preschooler care responsibilities (b=-0.230, p < .05). Therefore, the benefit associated with culture of flexibility in lowering psychological distress is significantly stronger among those with both childcare and elder care responsibilities, relative to those with neither obligation. Figure 1 shows the pattern well where the downward slope indicting the negative effect of culture of workplace flexibility on psychological distress is steepest for individuals with both childcare and elder care responsibilities. Thus, Hypothesis 3 is partially supported.

Fig. 1
figure 1

The Effect of Culture of Workplace Flexibility (CWF) on Psychological Distress, by Family Caregiving Obligations

Full size image

Lastly, we include three-way interaction terms between access to FWA, family caregiving responsibilities, and gender in Models 4–6. We do not find caregiving responsibilities and gender combine to moderate the effect of flextime or flexplace on psychological distress (Models 4 and 5). The positive and significant interaction term between culture of flexibility, gender, and sandwich generation (b = 0.535, p < .05) in Model 6, however, suggests that the effect of culture of flexibility on psychological distress varies by the intersection of gender and family caregiving responsibilities. As Fig. 2 shows, the downward slope indicting the negative effect of flexibility cultures on psychological distress is steepest for women with both child care and elder care responsibilities, whereas no such pattern is found among men. Hypothesis 4 is therefore partially supported.

Fig. 2
figure 2

The Effect of Culture of Workplace Flexibility (CWF) on Psychological Distress, by Family Caregiving Obligations and Gender

Full size image

Discussion

Using data from the 2008 wave of the National Study of the Changing Workforce, this study tests the effects of access to flextime, flexplace, and culture of flexibility on psychological distress. We further examine whether work–family conflict and work–family enrichment (in both directions) mediate the relationships between different types of FWA and distress, and whether these associations are moderated by family caregiving responsibilities as well as its intersection with gender. Results show that culture of flexibility—but not access to flextime or flexplace—is associated with lower psychological distress, with work–family conflict and work–family enrichment (in both directions) partially mediating the association. We also find that the benefit of culture of flexibility in lowering psychological distress is particularly salient among women (but not men) with both preschool childcare and elder care responsibilities.

Our first contribution to the work, family, and well-being literature is to show the diverse well-being implications of different types of flexible work arrangements. Of the three dimensions of FWAs, only culture of flexibility is found to predict lower psychological distress. This finding is consistent with prior research showing that usage of flexible work arrangements typically depends on how strong supervisors or the organization indeed support flexibility (Kelly & Moen, 2021; Vega, 2015). Therefore, flextime or flexplace alone might not be sufficient to promote well-being in the absence of family-friendly supportive practices (Allen, 2001; Mennino et al., 2005; Mesmer-Magnus and Viswesvaran 2006). By revealing the heterogeneous effects of alternative flexible arrangements, our findings contribute to the literature, showing that not all FWAs are equal in terms of their emotional well-being implications.

A second contribution of our research is to illuminate the mediating effects of work–family conflict and work–family enrichment (in both directions) in understanding why flexibility matters for well-being. We show that working in an organization characterized by a flexible culture lowers psychological distress partly through decreasing workers’ work-to-family and family-to-work conflict while increasing their work-to-family and family-to-work enrichment. These findings support the theorization of flexible work arrangements as “boundary-spanning resources” (Voydanoff, 2005). That is, a flexible workplace culture allows for the “physical or social attributes of one domain (either work or family) … [to] be exploited by the individual to resolve demands arising in another life domain” (Carlson et al., 2010, p.332). Importantly, even though flexible workplace arrangements originate from the domain of work, they seem to be effective in addressing issues arising in either the work or the home domain, as evidenced by the significant mediating effects of both work-to-family and family-to-work conflict or enrichment. Overall, our findings on the mediating effects indicate that experiences at the work–family interface constitute key channels through which work resources such as FWA affect workers’ subjective well-being (Voydanoff, 2005).

A third contribution we make to existing studies is to show the importance of family caregiving responsibilities intersecting with gender in shaping the effect of FWA on psychological distress. We find that women in the sandwich generation (with both childcare and elder care responsibilities) derive the most well-being benefits from a flexible workplace culture, whereas family caregiving obligations does not appear to moderate the relationship between FWA and psychological distress for men. This gendered pattern supports findings from prior research (Hill et al., 2008; Kelly et al., 2014; Moen et al., 2016) and highlights the potential operation of traditional gender ideology, with women socially expected to and usually shouldering the majority of family caregiving responsibilities. Therefore, they stand to benefit the most from a flexible workplace culture. Men, in comparison, may be more susceptible to the flexibility stigma (Coltrane et al., 2013)—that is, men are at higher risk than women to be viewed as poor organizational citizens once they begin to request flexible arrangements, given the violation of the presumed link between manhood and paid work (Townsend, 2002). As a result, men may be more conscious of using flexible arrangements and benefit less from such arrangements, even in the face of heavy family caregiving obligations.

From a practical point of view, our findings suggest that, compared with providing formal policies such as flextime or flexplace, it is perhaps more effective, in terms of promoting workers’ well-being, to build a truly supportive and flexible workplace culture. In the absence of organizational or supervisor support for workers’ work–family matters, family-friendly policies might only be interpreted by workers as “window dressing.” In addition, workers with heavy family caregiving obligations appear to benefit the most from a flexible culture—at least among women—indicating the need to provide sufficient support to those workers to facilitate their addressing family demands while still achieving career goals.

There are several limitations of this research. First is the use of cross-sectional data which does not allow us to test more dynamic relationships between our key variables. Future research based on longitudinal or experimental data is necessary to better shed light on the relationship between FWA and subjective well-being (Maggino & Facioni, 2017; Maggino & Zumbo, 2011). Second, even with the rich measures in the NSCW, we cannot control for all psycho-social or organizational variables, such as personality or organizational contexts, which likely shape the way workers perceive flexibility arrangements or stress, as well as whether and how they can cope with such stress (Wayne et al., 2004). Further quantitative research with more detailed measures, as well as qualitative studies, are needed to better understand the social dynamics around flexible work arrangements and subjective well-being. Third, flexible work arrangements could have a potential crossover effect, that is, workers’ access to FWA may affect the subjective well-being of their significant others. The design of the NSCW—surveying only one person per household—does not allow us to test this possibility, a task that awaits future researchers when data are available.

The COVID-19 pandemic has fundamentally shaped the work and family environments for millions of workers, highlighting the necessity for more flexible work arrangements to better balance work and family responsibilities. Our results highlight the well-being benefits for employees when they work in a supportive, flexible workplace culture, especially for women in the sandwich generation. Findings helps elucidate the disparate effects on well-being associated with different types of FWAs, the underlying mechanisms, and the heterogeneous effects for different social groups. To build a healthy and happy workforce, more research is needed to understand the organizational and policy forces that can impede or promote workers’ health and well-being.