The (in)compatibility of identities: Understanding gender differences in work – life conﬂict through the ﬁt with leaders

Women’s concerns about work – life balance are cited as a key factor underlying their continued underrepresentation in particular domains and roles. This gendered pattern is often attributed to factors in the home, such as women’s disproportionate share of domestic work and childcare responsibilities. We offer an additional explanation that focuses on workplace identities. Across four studies, we demonstrate that perceptions of work – life balance are not only a matter of balancing time, but also a matter of balancing identity, and that the availability of attainable leaders plays a key role in determining these processes. More speciﬁcally, a survey study (Study 1, N = 1223) among participants working in a historically male-dominated profession shows that gender differences in work – life balanceperceptionsare,inpart, explainedbywomen’sperceived lackofﬁtwith leaders and, in turn, their perceptions of incompatibility between who they are at home and who they are at work. In Studies 2 ( N = 207), 3a ( N = 209), and 3b ( N = 191), we demonstrate that gender differences in anticipated work – life balance can be ameliorated through exposure to attainable female leaders. These ﬁndings have implications for organizations that seek to recruit and retain women and demonstrate that issues of identity are crucial for facilitating work – life balance.

Notably, in studies where gender differences do exist, they are often explained in terms of the different demands on women's and men's time, particularly in the domestic sphere. For example, Greenhaus and Beutell (1985) distinguish between three dimensions of work-life conflict: (1) behaviour-based conflict, where the behaviours required in one domain are incompatible with the behaviours required in the other; (2) time-based conflict, which contrasts time spent at work and time spent at home; and (3) strain-based conflict, where strain from one domain affects performance in the other. Results are mixed with regard to the experiences of men and women of behavioural conflict (e.g., Carlson et al., 2000;Dierdorff & Ellington, 2008), but as women still shoulder a disproportionate amount of caring responsibilities at home (Craig & Mullan, 2010) and are expected to do so (Park et al., 2008), it follows that women experience more time-and strain-based conflicts and in turn report lower levels of work-life balance than men (Charles & James, 2005;Emslie & Hunt, 2009).
Here, we introduce an additional dimension of work-life conflict. Extending work on identity approaches to the interface of work and life outside of work (e.g., Lewis & Cooper 2005;Ramarajan & Reid, 2013), we take a social identity approach to understanding issues related to work-life conflict (Tajfel & Turner, 1979;Turner et al., 1987). Focusing on the flipside of work-life conflictthat is, work-life balancewe suggest that work-life balance is not only about balancing time or strain, but also about balancing identity. While we do not dispute the fact that unequal distribution of domestic and childcare work and expectations that mothers should be primary caregivers greatly affect gender differences in work-life balance, we do not think that this fully captures why women may experience more work-life conflict.
Social identities, identity compatibility, and work-life conflict Social identities are those parts of the self that are based on social groups to which we belong, such as gender, occupation, or nationality (Tajfel & Turner, 1979). Through the internalization of group norms and stereotypes, social identities provide us with a sense of who we are, as group members (e.g., as women, as employees, as Europeans), are and who we ought to be. The degree to which different social identities affect cognition and behaviour depends both on the strength of identification and the salience in a specific context or situation (Turner et al., 1987).
These multiple identities can either be relatively harmonious or in conflict with one another. Within an organizational context, Horton et al., (2014) describe different types of identity conflict that can occur, including conflict within or between individuals, and at different levels or the organization. Of most relevance here is intra-individual identity conflict, which fits well with conceptualizations of the interface between work life and private life. More specifically, while this interface has been conceptualized in many ways (e.g., Edwards & Rothbart, 2000;Powell et al., 2019), it is often conceptualized in terms of conflict (e.g., Netemeyer et al., 1996).
We argue that this conflict can stem from an incompatibility of different identities. Work and non-work identities cannot be seen of as separate or distinct aspects of the self (e.g., Kanter, 1977;Ramarajan & Reid, 2013), and thus, work-life balance is in part a matter of integrating the different facets of one's life (e.g., Lewis & Cooper 2005;Ramarajan & Reid, 2013). However, the ease of integrating these identities can vary greatly, and it is likely to be dependent on their content.
Who one is outside of work can vary in the degree to which it contrast with the stereotypes associated with one's workplace identity. For example, religious identities can be more or less connected to occupational identities, depending on the degree of fit between religious and occupational values and behaviours (Héliot et al., 2020). Similarly, 'identity disconnects' can occur when stigmatized identities are not disclosed at work (Ragins, 2008).
In the context of understanding gender differences in the work-life interface, we argue that the two relevant social identities are gender identity (what it means to belong to one's gender group) and workplace identity (e.g., what it means to belong to one's occupation or organization; Ashforth & Mael, 1989;Haslam, 2004). Given the relatively chronic salience of gender (Deaux & Major, 1987), one's gender identity is likely to play an important part of who one is outside of work (and indeed, how one experiences the workplace; e.g., Ely, 1995). We suggest that when these two identities are incompatible with each other, integration will be less feasible, and perceptions of work-life balance will be reduced.
Gender, identity (in)compatibility, and work-life balance Gender identity and workplace identity are likely to be more compatible for men than for women, as the content of these identities (i.e., the stereotypes associated with the two identities) vary between women and men. More specifically, stereotypes of women tend to be in conflict with workplace stereotypes (e.g., Heilman, 1983), particularly at more senior levels of the organizational hierarchy (Schein, 1973).
Women continue to be stereotyped as communal (e.g., gentle, warm) and lacking agency (e.g. submissive, timid). In contract, men are stereotyped as agentic (e.g., independent, assertive) and lacking communality (e.g., aggressive, cold; Eagly & Wood, 2012). Thus, stereotypes of men are very similar to those associated with the ideal worker, who is expected to be independent, competitive, and ambitious, while stereotypes of women are not (Park et al., 2010). It follows that women are more likely to experience identity incompatibility compared to men and this is likely to be particularly pronounced for stereotypically masculine roles and domains. Such identity incompatibility may result in general feelings of uncertainty and anxiety (Hirsh & Kang, 2016), lower engagement (Ahlqvist et al., 2013), and poorer well-being following transitions (Iyer et al., 2009). Moreover, Hodges and Park (2013) demonstrated that when women switched between a parent identity and a workplace identity (what we would see as incompatible identities), they showed depleted cognitive resources, but the same depletion was not seen in men. Given these findings, we argue that holding incompatible identities will likely be taxing and therefore negatively affect perceptions of work-life balance. The extent to which this incompatibility occurs, however, is likely to depend on a range of factors.

Leaders and identity (in)compatibility
Leaders hold a unique position that allows them to shape and define the group in a number of ways (e.g., Haslam et al., 2010;Steffens et al., 2014). They are often seen as prototypical group members who embody 'who we are', such that the traits and characteristics of a leader become defining of the group as a whole (Steffens, et al., 2013). For example, if an innovative and ambitious leader is seen as prototypical of their organization, the organization too will be seen as innovative and ambitious. Moreover, employees who identify strongly with the organization will value innovation and ambition, and innovative and ambitious employees are likely to succeed. However, such identity processes are not simply a static adoption of the leader's characteristics by the group, and the leader can also be an 'identity entrepreneur', actively defining the characteristics that are important to the group (Haslam et al., 2010).
Extending this line of reasoning, we suggest that perceived fit with leaders is likely to also impact perceptions of work-life balance because a sense of fit directly impacts perceptions of identity compatibility. That is, to the extent that individuals perceive a fit between themselves and their leaders, who are defining of organizational identity, we suggest they are more likely to perceive a compatibility between who they are at home and who they are at work, which should result in more positive perceptions of work-life balance.
Importantly, the issue of leadership and identity becomes more complex once gender is introduced into the equation. Leadership positions, especially at the higher echelons, are still overwhelmingly occupied by men, which likely affects their perceptions by women and men. Indeed, to the extent that women perceive lower fit between themselves and leaders, they also report lowered organizational and professional identification and reduced ambition and career motivation (e.g., Peters et al., 2012). Similarly, to the extent that women feel a lack of fit with leaders, they are likely to experience lower identity compatibility and, in turn, poorer work-life balance.
However, gender is likely not the only factor affecting fit with leaders. As described above, leaders can take an active role in crafting identity. As part of this process, they can act as role models and in turn affect identity compatibility and work-life balance. Indeed, the availability of role models has been shown to decrease identity incompatibility and increase individuals' identification with a domain or role (e.g., Dasgupta, 2011;Rosenthal et al., 2013). However, not every leader will make a good role model. One of the attributes which is likely to impact fit with leaders is their perceived attainability, that is, their perceived future similarity to the self (Morgenroth et al., 2015). We argue that the attainability of leaders is one of the factors that shapes perceived fit with these leaders. This is important as it shows avenues to increase leader fit that do not solely rely on increasing the number of female leaderswhich is certainly important, but unlikely to be achieved in the short term.
The current project Bringing all of these lines of research together, in this project we investigate the role of identity in understanding gender differences in work-life balance, focusing on the role that leaders play in shaping identity compatibility. We predict the following: H 1. Women will report lower levels of work-life balance compared to men.
H 2. The relationship between gender and work-life balance perceptions will be, in part, explained by (lack of) fit with leaders and, in turn, identity (in)compatibility.
H 3. The indirect effect of gender on work-life balanced will be moderated by leader attainability such that being exposed to an attainable leader will reduce the effect.
We test these hypotheses across four studies: A correlational survey study with a large sample of women and men from a historically male-dominated field tests H1 and H2, and three experimental studies (one exploratory, one pre-registered, the third a replication with additional measures) with employed participants from various fields additionally test H3.

STUDY 1
In this study, we test H1 and H2 in a large sample of women and men working in the UK veterinary profession. Veterinary medicine is a historically male-dominated field where the numerical balance has recently shifted such that women now make up the overwhelming majority of veterinary students. However, gender discrimination is still in evidence, with women being under-represented in leadership positions and prestigious sub-specialties, a gender pay gap favouring men, and evidence of experiences of gender discrimination in the field (Begeny, Ryan, Moss-Racusin, & Ravetz, 2020).

Method
All studies reported in this paper were approved by the Ethics Committee at one of the authors' institution.

Participants
We recruited participants as part of a larger survey sent out to members of a large association of veterinarians. A total of 1661 participants took part in our study. We excluded three participants who did not indicate their gender, 421 participants who did not indicate how many hours they worked per week, and 14 participants who indicated that their number of hours worked per week were either below 1.31 or above 92.71 (equal to the mean AE 3 SDs) 1 . This left us with a final sample of 1223 (65.90% female). 2.45% indicated they were under 25, 30.17% were between 25 and 34, 23.30% between 35 and 44, 23.14% between 45 and 54, 16.93% between 55 and 64, and 3.76% were 65 and over; Three participants did not indicate their age. Almost all participants (99.98%) worked in the United Kingdom.

Measures
We collected the data as part of a larger survey of individuals working in the UK veterinary profession. Here, we only analyse and report the measures relevant for this project. All constructs were measured using three items with seven-point response scales from 1 (strongly disagree) to 7 (strongly agree). We measured fit with leaders using the items 'When I look at successful vets, I have a lot in common with them', 'I think that people like me have made it to the top of the veterinary career', and 'I see myself as quite different from those who have made it in the veterinary career' (reverse scored; α = .67; adapted from Peters et al., 2012). We measured identity compatibility using items developed specifically for this study: 'Who I am at work is compatible with who I am outside of work', 'Who I am at work is similar to who I am outside of work', and 'Who I am at work is in conflict with who I am outside of work' (reverse scored; α = .87). Finally, we measured work-life balance using the items 'I have a good work-life balance', 'I feel that I have a good balance between work and my life outside of work', and 'I struggle to maintain balance in my life' (reverse coded; α = .91; adapted from Peters, Ryan, & Haslam 2015). Additionally, we measured hours worked per week and years of work experience as control variables for our analyses.

Results
Descriptive statistics and correlations are displayed in Table 1. We first tested H1 comparing levels of work-life balance of women to those of men using an ANCOVA, controlling for hours worked per week and years of work experience. In line with predictions, men reported higher levels of work-life balance than women, F(1, 1111) = 11.91, p = .001, η 2 p = .01. Two further ANCOVAs showed that women reported lower levels of fit with leaders than men, F(1, 1098) = 13.39, p < .001, η 2 p = .01, but that identity compatibility did not differ between women and men, F(1, 1112) = 0.46, p = .500, η 2 p < .01. Next, we tested H2 using PROCESS (v3.2, Model 6; Hayes, 2018) with gender as the predictor (0 = women and 1 = men), fit with leaders as the first mediator, identity compatibility as the second mediator, and work-life balance as the outcome. We controlled for hours worked per week and years of work experience. Results can be seen in Figure 1. As predicted, the indirect effect through both mediators was positive B = .04 [.01, .06] 2 , indicating that men, compared to women, felt that they fit more with leaders in their field, in turn experienced higher levels of identity compatibility and, in turn, work-life balance.

Discussion
In Study 1, we found support for our predictions that women experience lower work-life balance than men and that this can be explained by the fact that they experience lower fit  2 Numbers in brackets refer to 95% confidence intervals.
with their leaders and, in turn, feel that their work identity and outside-of-work identity are less compatible. This pattern held even when controlling for hours worked per week, indicating that work-life balance is not simply a balancing of time or strain, but also a balancing of identities. While this first study provides initial evidence for our hypotheses, its correlational nature does not allow us to make causal claims. In the next three studies, we sought to replicate the initial findings and further investigate the causal mechanisms underlying these relationships.

STUDY 2
As we have demonstrated in Study 1, fit with leaders plays an important role in shaping work-life balance perceptions. Importantly, women generally report lower levels of leader fit, particularly in male-dominated professions (Peters et al., 2012). However, as we argued above, as identity entrepreneurs, leaders can also actively increase fit. They can do so in many ways. Here, we focus on attainability (i.e., potential future similarity) as one of the factors affecting fit with leaders. While leader gender likely plays an important role in these processes as well, we chose a different focus (a) as the number of female leaders is not easily changeable in real organizational contexts, and (b) to illustrate that presenting women with female leaders is not a panacea. Instead, fit with leaders is likely determined by a range of attributes, gender being just one of them.
In this exploratory study, we recruited a sample of employed participants working in various sectors and organizations and presented them with fictional information about a new female leader. We aimed to portray this leader as either particularly attainable to women in particular (she was relatable and worked hard for her success) or unattainable (not as relatable, and found success easy), or no information was given. As this study was exploratory, we included a range of measures and explored our data extensively. Here, we only report the measures analyses relevant to our hypotheses (for full materials for this study as well as Studies 3a and 3b, see online supplement).

Participants
We recruited 277 employed participants online using the Prolific website. After excluding those who had not completed the survey and those who failed the manipulation check, our final sample consisted of 207 (40.10% female). This sample size gives us 80% power to detect small to medium effect sizes (f = .20) in our primary analyses. Of all participants, one participant was between 18 and 24, 99 were between 25 and 34, 76 were between 35 and 44, 23 were between 45 and 54, and eight were between 55 and 64. The majority of participants were British (54.59%), from the United States (20.29%), or various European countries (17.87%).

Design and procedure
We introduced the study as an investigation of reactions to changes in the workplace.
Participants were asked to imagine that their organization had announced a new leader for a top management position. We then presented them with information about this leader, who was always female and presented as highly competent and successful. In the two experimental conditions, this was followed by a brief interview with the new leader. In this interview, she was portrayed as either (a) having low levels of attainability by emphasizing how easy she had found her career so far and how she had not struggled to become successful or (b) having high levels of attainability, talking about how she had found her career challenging but had been able to overcome these challenges to become successful. In a control condition, we omitted the interview. We randomly assigned participants to one out of these three conditions, resulting in a 2 (Participant gender: Female vs. Male) × 2 (Type of leader: Unattainable vs. Attainable vs. Control) betweenparticipants design.
We then measured leader fit and anticipated work-life balance on scales from 1 (strongly disagree) to 7 (strongly agree). We measured leader fit using nine items (e.g., 'I have a lot in common with this leader' α = .88) and anticipated work-life balance using four items (e.g., 'It would be easy to achieve a good work-life balance', α = .79). Lastly, participants provided demographic information and were debriefed.

Results
Descriptive statistics and correlations for Studies 2 and 3a can be found in Table 2.
We first ran two 2 (Participant gender: Female vs. Male) × 3 (Type of leader: Attainable vs. Unattainable vs. Control) ANOVAs with anticipated fit with the leader and work-life balance as the DVs. For leader fit, we found an effect for gender, F(1, 201) = 6.88, p = .009, η 2 p = .03, such that women (M = 4.18, SD = 1.28) anticipated lower levels of leader fit than men (M = 4.62, SD = 1.08). In addition, there was an effect for type of leader, F(2, 201) = 16.41, p < .001, η 2 p = .14, and post-hoc tests showed that participants anticipated higher levels of fit with the attainable leader (M = 5.03, SD = 1.08) than with the unattainable leader (M = 4.02, SD = 1.17), p < .001, or the leader in the control condition (M = 4.41, SD = 1.08), p = .004. Levels of leader fit were not different for the unattainable leader and the leader in the control condition, p = .075. Importantly, and in line with H3, we found a significant interaction between type of leader and gender, F(2, 201) = 3.48, p = .033, η 2 p = .03, which is illustrated in Figure 2. The gender difference was primarily driven by participants exposed to the unattainable leader. Here, but not in the other two conditions, women anticipated lower levels of leader fit than men, p < .001. Moreover, the manipulation had a stronger effect on female participants: They anticipated lower levels of fit with the unattainable leader compared to the attainable leader, p < .001, and the leader in the control condition, p = .037, and lower fit with the leader in the control condition, compared to the attainable leader, p = .006. Men also reported lower levels of fit with the unattainable leader compared to the attainable leader, p = .023, but not compared to the leader in the control condition, p = .750. Similar to women, fit with the leader in the control condition was lower compared to the attainable leader, p = .042. For anticipated work-life balance, we also found an effect for gender, F(1, 201) = 19.57, p < .001, η 2 p = .09. In line with H1, women (M = 3.98, SD = 1.27) anticipated lower levels of work-life balance than men (M = 4.75, SD = 1.09). In addition, there was an effect for type of leader, F(2, 201) = 15.18, p < .001, η 2 p = .13, and post-hoc tests showed that participants anticipated lower levels of work-life balance when exposed to an unattainable leader (M = 3.91, SD = 1.25) than when exposed to the attainable leader (M = 4.84, SD = 0.94), p < .001, or the leader in the control condition (M = 4.69, SD = 1.21), p < .001. Levels of anticipated work-life balance were not different in the attainable leader and control condition, p = .709. In line with H3, we found a significant interaction between type of leader and gender, F(2, 201) = 5.72, p = .004, η 2 p = .05 (see Figure 3). Simple effects analyses revealed that, as predicted, the gender difference was driven by participants in the unattainable leader condition, p < .001, and the control condition, p = .004. In both conditions, but not in the attainable leader condition, women anticipated lower levels work-life balance than men. Moreover, the manipulation once more had a stronger effect on female participants. Women anticipated lower levels of work-life balance when exposed to the unattainable leader compared to the attainable leader, p < .001, and the leader in the control condition, p = .007, and lower levels in the control condition compared to the attainable leader condition, p = .017. For men, only the difference between the unattainable leader and the leader in the control condition was significant, p = .040.
Finally, to test H2 and H3, we ran a moderated mediation analysis using PROCESS (v3.2, Model 7) with participant gender as the predictor, leader fit as the mediator, type of leader as the moderator of the path from gender to leader fit, and anticipated work-life balance as the outcome (see Figure 4). We used indicator coding to code type of leader and, as we were particularly interested in the beneficial effects of an attainable leader, selected the attainable leader condition as the reference category. Gender was coded as 0 = female and 1 = male.  Table 3). This pattern lends partial support to H3. In line with predictions, compared to an unattainable leader, exposure to an attainable leader does eliminate gender differences in fit with the leader and in turn work-life balance. However, this was not the case when comparing the attainable leader to a leader about whom little information was provided.

Discussion
This exploratory study provided support for H1 and partial support for H3. In line with H1, we found that, overall, women anticipated lower levels of work-life balance than men. This was the case when they were exposed to an unattainable leader or a leader about whom little information was provided.  Despite these effects, gender did not directly affect anticipated fit with the new leader in the control condition and, in turn, we did not find the predicted moderated mediation effect suggesting that exposure to an attainable leader could eliminate the gender difference in fit with leaders which we observed in Study 1. Nevertheless, we found clear evidence that fit with leaders is malleable and that attainability is one of the factors underlying leader fit. More specifically, exposure to an attainable leader can eliminate the gender difference evident when exposed to an unattainable leader.
As this study was exploratory, it is important to replicate the findings. We do so in Studies 3a and 3b.

STUDY 3A
In Studies 1 and 2, we found support for the notion that (a) the gender difference in work-life balance is explained, in part, by fit with leaders and in turn identity compatibility and (b) that exposure to an attainable or unattainable leader can alter this effect. In this study, we aim to replicate both findings, using a similar design to Study 2, but including a measure of identity compatibility. We pre-registered our hypotheses, target sample size, measures, and analyses (see https://osf.io/gmw7e/) and adhere to this preregistration throughout this study.

Participants
We recruited 208 participants using the Prolific website, restricting access to the survey to those who were employed. We excluded 8 participants who indicated they were not employed, one participant who did not indicate their gender, and 10 who failed the attention check. After recruiting additional participants, our final sample size was 209 (48.80% female; M age = 31.9; SD = 9.43). Participants came primarily from the United Kingdom (28.71%), Poland (14.83%), the United States (7.66%), Portugal (7.18%), Canada (4.78%), and various European countries (29.66%). Note. Values in brackets refer to 95% confidence intervals. We used dummy coding to code type of leader with the attainable leader as the reference category. Gender was coded as 0 = female and 1 = male.

Design and measures
The design and procedure were similar to Study 2. However, to simplify the design, we did not include a control condition, resulting in a 2 (Participant gender: Female vs. Male) × 2 (Type of leader: Unattainable vs. Attainable) between-participants study. We measured leader fit using the same items as in Study 2 (α = .91) and identity compatibility using the same items as in Study 1, adapted for the context of this study (α = .83). We measured anticipated work-life balance using the three items from Study 1 and the four items from Study 2 (α = .89).

Results
Descriptive statistics and correlations can be found in Table 2. We first ran a series of 2 (Participant gender: Female vs. Male) × 2 (Type of leader: Unattainable vs. Attainable) ANOVAs with anticipated leader fit, identity compatibility, and work-life balance as the dependent variables. For anticipated leader fit, we found an effect for type of leader, F(1, 205) = 33.98, p < .001, η 2 p = .14, such that participants anticipated lower fit with the unattainable leader (M = 3.92, SD = 1.17) than with attainable leader (M = 4.81, SD = 1.10). There was no main effect for gender (p=.278), but, in line with H3, we found a significant interaction between type of leader and gender, F(1, 205) = 6.58, p = .011, η 2 p = .03, which is illustrated in Figure 5. Simple effects analyses revealed that men anticipated a significantly higher fit with the unattainable leader than women, p = .010, while there was no gender difference for those in the attainable leader condition, p = .300. Both women, p < .001, and men, p = .021, anticipated a higher fit with the attainable leader than with the unattainable leader.
For anticipated identity compatibility, we once more found an effect for type of leader, F(1, 205) = 25.41, p < .001, η 2 p = .11, such that participants anticipated lower identity compatibility when exposed to an unattainable leader (M = 4.38, SD = 1.36) than when exposed to an attainable leader (M = 5.19, SD = 1.08). There was also main effect for gender, F(1, 205) = 7.48, p = .007, η 2 p = .04, such that women (M = 4.57, SD = 1.40) anticipated lower levels of identity compatibility compared to men (M = 4.98, SD = 1.15). The two factors did not interact, p = .191. Finally, for anticipated work-life balance, we only found an effect for type of leader, F(1, 205) = 19.13, p < .001, η 2 p = .09, such that participants in the unattainable leader condition (M = 4.03, SD = 1.12) anticipated lower levels of work-life balance than those in the attainable leader condition (M = 4.66, SD = 1.08). The effect of gender, p = .082, and the interaction, p = .067, were not significant. These findings do not support H1.
Next, to test H2 and H3, we ran a moderated serial mediation analysis using PROCESS (v3.2, Model 83) which is illustrated in Figure 6. In line with predictions, results revealed evidence for moderated mediation, index of moderated mediation = −.06 [−.14, −.01]. The indirect effect of gender on work-life balance through leader fit and, in turn, identity compatibility was significant in the unattainable condition, B = .04 [.005, .10], but disappeared in the attainable leader condition, B = −.02 [−.06, .01], supporting H2 and H3 (for additional details, see Table 4).

Discussion
In this study, we replicated the finding that gender differences in work-life balance can in part be explained by differences in leader fit and identity compatibility. However, this was only the case when exposed to an unattainable leader. Exposure to an attainable female leader eliminated these gender differences.

Gender
Work-life balance Fit with leader Identity compatibility Type of leader  Note. Values in brackets refer to 95% confidence intervals. Gender was coded as 0 = female and 1 = male. Condition was coded as 0 = low attainability and 1 = high attainability.

STUDY 3B
Study 3b was a replication of Study 3a with the addition of a manipulation check and some additional measures to examine (a) how the attainable and unattainable leaders were perceived more generally and (b) whether the effect on anticipated work-life balance was indeed due to fit with the leader and not because the attainable leader was seen as more supportive of policies aimed at increasing work-life balance. We did not pre-register this study as the key measures, target sample size, and hypotheses were identical to Study 3a. Thus, other than the changes (described above and below), we adhere to the preregistration for Study 3a.

Participants
We aimed for the same sample size as in Study 3a and used the same recruitment method. Our final sample size was 191 (53.40% female) with an average age of 30.24 (SD = 8.61). Our sample came primarily from United Kingdom (20.94%), Poland (13.61%), Portugal (13.61%), Canada (6.28%), the United States (5.24%), and Greece (5.24%). The remaining participants were primarily from other European countries (25.65%).

Design and measures
The design and procedure were similar to Study 3a. We measured leader fit (α = .90), identity compatibility (α = .86), and anticipated work-life balance (α = .89) using the same items as in Study 3a. Participants then indicated the extent to which different attributes described the leader on a scale from 1 (does not describe her at all) to 7 (describes her very well), including our manipulation check (attainable, relatable, accessible, α = .84) as well as items measuring the leader's levels of success (successful, accomplished, competent, α = .75), warmth (likable, nice, approachable, α = .91), and authenticity (authentic, honest, truthful, α = .92). In addition, we measured the extent to which participants believed the new leader was supportive of policies aimed to support work-life balance (generous parental leave, flexible working hours, and working remotely; α = .88) on a scale from 1 (not at all) to 7 (very much).
We ran a series of 2 (Participant gender: Female vs. Male) × 2 (Type of leader: Unattainable vs. Attainable) ANCOVAs with anticipated leader fit, identity compatibility, and work-life balance as the dependent variables, controlling for authenticity, warmth, and policy support.
For anticipated leader fit, neither of the main effects were significant (all p>.261), but we found the predicted significant interaction between type of leader and gender, F(1, 184) = 5.74, p = .018, η 2 p = .03 (see Figure 7) Simple effects analyses revealed that women anticipated a significantly higher fit with the attainable leader than men, p = .029, while there was no gender difference for the unattainable leader, p = .253. In addition, women, p = .019, anticipated a higher fit with the attainable leader than with the unattainable leader, but there was no difference for men, p = .463.
For anticipated identity compatibility, neither the main effects nor the interaction were significant (all p>.322). Finally, for anticipated work-life balance, we only found the predicted effect for gender, F(1, 184) = 4.11, p = .044, η 2 p = .02, such that women  (M = 4.11, SD = 1.17) anticipated lower levels of work-life balance than men (M = 4.59, SD = 1.03). The effect of type of leader, p = .782, and the interaction, p = .857, were not significant. Next, we ran the same moderated serial mediation analysis as in Study 3a (see Figure 6), but controlling for authenticity, warmth, and policy support. In line with predictions, results revealed evidence for moderated mediation, index of moderated mediation = −.06 [−.14, −.01]. However, the indirect effect for gender on work-life balance through leader fit and, in turn, identity compatibility was negative, and significant only in the attainable leader condition,  Table 6). Interestingly, perceived support for work-life balance enhancing policies did not affect anticipated work-life balance.

Discussion
This study largely replicated findings from the previous studies. First, we found that, in line with H1, men reported higher anticipated work-life balance than women. Moreover, while results differed slightly from Study 3a, we once again found that exposure to an Note. Values in brackets refer to 95% confidence intervals. Gender was coded as 0 = female and 1 = male. Condition was coded as 0 = low attainability and 1 = high attainability.
attainable female leader affected women's, but not men's anticipated leader fit and, in turn, their identity compatibility and work-life balance. Importantly, this study included a manipulation check showing that we successfully manipulated attainability. Importantly, while our manipulation also affected a range of other perceptions, results held when controlling for these factors and, somewhat surprisingly, perceived support for work-life balance enhancing policies was not associated with anticipated work-life balance. This emphasizes the need to look beyond time-based factors and consider identity as an important predictor of work-life balance.
While this study showed that our manipulation was perhaps not the 'cleanest' in that it manipulated authenticity, warmth, and policy support as well as attainability, we do not see this as particularly problematic. Across these studies, we are examining the effect of fit with leaders on work-life balance and thus aimed for a manipulation that would increase/ decrease this variable. Our results show that we succeededand whether participants feel stronger fit because the leader is attainable, authentic, or warm, is not central to our research question. Indeed, we would argue that there are many factors that affect fit with leaders, attainability just being one of them.

GENERAL DISCUSSION
Across four studies, we used an identity framework to investigate gender differences in real and anticipated work-life balance. In line with our predictions, we found that women generally reported and anticipated lower levels of work-life balance than men. This is in line with some, but not all previous findings. Our findings further shed light on this inconsistency: The sample of Study 1 consisted of men and women in a historically maledominated industry, the veterinary profession. It is therefore likely that women are mostly exposed to male leaders, whom they are likely to perceive as unattainable. In line with this, we found support for the notion that the gender difference in work-life balance disappears after exposure to an attainable female leader. In other words, we would not expect to find these differences in all circumstances.
We also found evidence for our prediction that fit with leaders and identity compatibility play important roles in shaping perceptions of work-life balance. Across all four studies, we found that gender differences in work-life balance were explained by the fact that women experience lower levels of fit with some leaders and in turn a larger incompatibility between their work and non-work identities. This was true when examining fit with leaders very broadly (Study 1) or examining fit with a specific leader (Studies 2, 3a, 3b).
Importantly, while the overall pattern was consistent across studies, some differences emerged. Results from Study 2 suggested that gender differences in work-life balance are specifically caused by unattainable leaders, not by leaders for whom little information is provided. This pattern was replicated in Study 3a, but Study 3b showed that indirect effects of gender only emerged when participants were exposed to attainable leaders when controlling for the leader's perceived authenticity, warmth, and support for work-life balance enhancing policies. More specifically, women reported higher anticipated work-life balance under these circumstances. Despite these inconsistencies, all three experimental studies show that women's perceived fit with leaders, identity compatibility, and work-life balance are particularly dependent on characteristics of the leader, while men's perceptions of these constructs are more stablelikely because they have no reason to perceive a lack of fit with leaders or identity incompatibility to begin with.

Theoretical and practical implications
Our findings advance the understanding of work-life balance. We argue that work-life balance is as much about balancing identity as it is about balancing time and that, for women, particularly in male-dominated and stereotypically masculine domains, this balancing is harder because stereotypes of women are less compatible with the stereotypes associated with their work role. Our findings support this notion and further suggest that fit with leaders is one aspect that shapes the perceived compatibility of work and non-work identities.
These insights also have practical implications. Most initiatives aiming to improve work-life balance focus on alleviating time-based conflicts. For example, organizations may offer part-time work, flexible working hours, or working from home to enable workers with caring responsibilitiesoften womento manage their different time demands more effectively. However, while these policies are likely beneficial for some employees, our findings indicate that time-based strategies overlook an important part of work-life balance, namely identity.
Organizations aiming to improve work-life balance should make active choices to create work-related identities that are more compatible with the diverse set of identities held by their employees. The visibility of attainable leaders is one route to achieving this goal. Importantly, as what is considered attainable will differ from employee to employee, having a diverse and visible leadership team is likely most beneficial for perceptions of work-life balance.
At the same time, these findings should not be taken as a prompt to only focus on identity and ignore the strain that long working hours and unequal distribution of domestic labour and caring responsibilities have on workersand on women in particular. Policies and interventions aiming to increase work-life balance should take both of these issues into account.

Limitations and future research
While our studies make an important contribution, they are not without limitations. First, it could be argued that instead of the causal pathway we proposed, identity compatibility could lead to fit with leaders and in turn affect attainability. For example, a match between someone's implicit theories of leadership and self-concept (identity compatibility) could lead to perceiving oneself as more similar to leaders and, in turn, increase the attainability of leadership positions (e.g., DeRue & Ashford, 2010;Junker & van Dick, 2014). However, this pathway seems less relevant here, as it is specifically relevant to the attainability of leadership positions and less relevant to the role of leader fit in perceptions of work-life balance.
It should be noted that the manipulation we used not only manipulated attainability, but also warmth, authenticity, and support for work-life balance enhancing policies. Future research should replicate our findings using a different, cleaner manipulation. For example, it would be interesting to know whether leader gender itself affects the attainability of and fit with the leader for women and for men. Another possibility would be to provide participants with feedback about whether or not they are similar to leaders in their organization to manipulate fit with leaders more directly.
Lastly, we did not measure strength of gender identification, gender self-stereotyping, or intersecting identities in these studies, and instead examined women as a group. However, women differ vastly in how they see themselves and how important gender stereotypes are in shaping their self-imagine. Similarly, women with intersecting minority identities (e.g., lesbians, women of colour) face different stereotypes and identity content that are likely to affect identity compatibility. Future research should investigate these nuances.

Conclusion
In this paper, we have investigated gendered work-life conflict using a social identity framework. While we have focused on gender, this approach can be applied to any identity and can therefore be used to understand how individuals from a diverse range of backgrounds may experience work-life conflict. It also has practical implications for those seeking to increase work-life balance: Instead of solely focusing on time, we need to create environments that enable all individuals to balance both their time and their identities.