Explaining how gender affects attitudes toward the state remains a fundamental task for political science. While these attitudes are undoubtedly the result of many individual, sociological, and contextual factors and their interactions, this article focuses on how experiences related to gendered social roles adopted over the life course contribute to opinions about societal priorities. More specifically, while scholars have studied the effects of custodial parenthood on political attitudes, we know less about how men’s and women’s involvement in this role transforms political attitudes (Elder and Greene Reference Elder, Greene and Whitaker2008; Tavits et al. Reference Tavits, Schleiter, Homola and Ward2023).
We ask how experiences within parenthood, including extending family parental leave or fulfilling weekly parental tasks, shape opinions about the state’s role. We move beyond studying the differences between parents and non-parents to examining how variability in parental engagement affects parents’ attitudes. We argue that fathers’ in-depth caregiving represents behaviors culturally and historically associated with women, which counter the behaviors typically expected of men (Wood and Eagly Reference Wood and Eagly2002, Reference Wood, Eagly, Olson and Zanna2012), and providing such care will illuminate the importance and challenges of caregiving and the value of policies that help families meet these needs. Consequently, we expect that fathers will view the welfare state’s assistance more positively. For mothers, intense experiences of parenthood and childcare provide experiences that affirm women’s traditional gender role, further deepening their relationship with that role. We therefore expect that mothers who take longer parental leave or engage heavily in parenting tasks will lead to more gender-traditional attitudes compared to women who do less, and we expect fewer changes among women compared to men, overall.
Our study adds to the broader literature on the effects of gender on public opinion by heeding the call of scholars to use insights from psychological theories on how the socialized concept of gender, as opposed to the biology of sex, alters political experiences (Beckwith Reference Beckwith2005; Schneider and Bos Reference Schneider and Bos2019). Social policies provide a critical backdrop for shaping gender, its associated roles, and how men and women perform them (Banducci et al. Reference Banducci, Elder, Greene and Stevens2016). Studying various paths of involvement within parenthood offers a prime example of this process. In countries without comprehensive public policies tailored to parents, fathers especially tend to lack the opportunity to be the primary caregiver for their child for an extended period of time (Hyde, Essex, and Horton Reference Hyde, Essex and Horton1993). By focusing on Sweden, where policies allow parents to take generous parental leave to be the primary caregiver for their child, we can begin to understand how state policies shape the involvement that partnered heterosexual men and women have within the social role of parenthood and how such role-affirming or — disconfirming — experiences alter their views on societal priorities.
We use the Swedish Level of Living Survey (LNU), a population-based panel study that included length of parental leave, frequency of completing parental tasks, and four political items that tapped into broad views about the role of the state, including support for a society with small income differences, with gender equality in the family, with care done in the family as opposed to the welfare state, and with private alternatives for state welfare programs. Because the focus of the study was on living conditions rather than politics, we are limited to these four items; however, surveys rarely include all of these measures with a panel design in a country like Sweden. The virtues of this panel study mean that we can use structural equation modeling to examine how parental leave influences attitudes about the state while addressing the reciprocal relationship between extended parental leave and these attitudes. Our results demonstrate that the specific parental role experience of taking leave influences mothers and fathers differently, and largely in a pattern that matches their role-relevant goals, though we do not find similar effects for parental tasks. We conclude by discussing the size of these effects and the significance of finding results in a country with high baseline levels of gender equality.
Theoretical Background
Social Roles
We situate our expectations about the effects of parental leave policies in social role theory, which offers a broader theory of how socially constructed gender roles shape behavior (Eagly and Wood Reference Eagly, Wood, van Lange, Kruglanski and Higgens2012; Eagly, Wood, and Diekman Reference Eagly, Wood, Diekman, Eckes and Trautner2000). Historically, men and women occupied different societal roles: men were providers or leaders and women were caregivers. Due to ongoing enforcement of this division of labor, men and women in the present, particularly those in heterosexual relationships, still follow similar patterns, though with different intensities and variations (Charles and Grusky Reference Charles and Grusky2005; Yavorsky, Kamp Dush, and Schoppe-Sullivan Reference Yavorsky, Dush and Schoppe-Sullivan2015). Because these roles require particular personality traits for success, social role theorists argue that men and women develop traits differently to thrive in the roles they inhabit. For example, the roles men undertake commonly require agentic personality traits (e.g., aggression, ambition), whereas those of women often rely on communal traits (e.g., kind, warm, caring) for success. Social role theorists argue that gender stereotypes flow from observing men and women in these roles and the presumption that they naturally possess these traits (Eagly, Wood, and Diekman Reference Eagly, Wood, Diekman, Eckes and Trautner2000; Wood and Eagly Reference Wood, Eagly, Olson and Zanna2012). Society’s rewards for conformity to gendered expectations create incentives for people to stay in traditional gender roles (Eagly and Koenig Reference Eagly and Koenig2021). The net result of the sex segregation of roles, and the high frequency with which men and women continue to adopt roles associated with their gender, is that men and women typically develop and strengthen different personality traits, behaviors, and goals consistent with their prescribed gender roles. These traits associated with a diffuse gender role, or a general sense of being a man or a woman, correlate with attitudes and behaviors across a wide variety of situations (Diekman and Schneider Reference Diekman and Schneider2010).
Social role theory also predicts that specific roles — those roles that men and women adopt throughout their lifetime, such as an occupational or familial role — also influence their attitudes and behaviors. For example, having an occupation in fields such as education, health care, and social work (termed “redistributive occupations”) contributes to supporting social welfare programs, even when controlling for other factors (Howell and Day Reference Howell and Day2000). As someone adopts a new role, there should be subsequent goal and attitude shifts since specific roles can illuminate previously unseen goals and social issues (Diekman and Schneider Reference Diekman and Schneider2010). Participating in the labor force, for example, predicts support for increased government involvement in job creation (Iversen and Rosenbluth Reference Iversen and Rosenbluth2006). We pursue a new direction in this theory and research area by clarifying how the intensity of role experiences matters for attitudes. Because people vary in the extent to which they embrace or engage in a new role, role change alone might not be enough to explain or predict the effects of a role on attitudes (Diekman and Schneider Reference Diekman and Schneider2010). In a general sense, it is important to not just focus on the assignment of social roles per se, but to integrate variation in how intensely individuals experience their new role, how drastically their behavior and identity change, and how new interests and goals come to light.
As we define the intensity of the role, in part, as to how long the individual uses state-supported parental leave, we also contribute a new theoretical and empirical approach in the tradition of scholars who study how policy influences public opinion (Soss and Schram Reference Soss and Schram2007; Svallfors Reference Svallfors2010) by examining how government policies can create specific role experiences and affect attitudes toward the role of the state differently in women and men.
Parenthood, Parental Leave, and Role Change
Effects of Parental Engagement on Mothers
Existing research already demonstrates how the diffuse role of being a woman creates attitudes that differ from men, and inhabiting the specific role of being a mother leads to different attitudes from non-mothers and fathers. Compared to men, women are more likely to support a stronger role of the state in providing services and helping people (Banducci et al. Reference Banducci, Elder, Greene and Stevens2016; Lizotte Reference Lizotte, Bos and Schneider2017). According to social role theorists, motherhood strengthens many of the same characteristics of the diffuse role of being a woman, and therefore, the two roles seem to mutually reinforce each other in shaping the political attitudes of women. For example, mothers, compared to non-mothers, tend to be even more supportive of childcare/school funding, social services such as food stamps, welfare, health care, job services, and aid to minority groups (Elder and Greene Reference Elder, Greene and Whitaker2008, Reference Elder and Greene2012; Greenlee Reference Greenlee2014; Goodyear-Grant and Bittner Reference Goodyear-Grant, Bittner, Thomas and Bittner2017). Furthermore, mothers are more supportive than fathers of government spending on schools, childcare, and social services (Lizotte Reference Lizotte, Bos and Schneider2017).
These cross-sectional studies address the effects of becoming a mother; however, they cannot address how women’s specific experiences in the motherhood role, such as parental leave or engaging in parental tasks, shape political attitudes over time, which is the gap we fill. Motherhood is an important, defining role for women (Greenlee Reference Greenlee2014; Katz-Wise, Priess, and Hyde Reference Katz-Wise, Priess and Hyde2010), so much so that the intensive experience within parenting might not change attitudes any further because of ceiling effects in support of state funding for services focused on women and children. In one study, an appeal specifically emphasizing that current policies might threaten general parental concerns did not change the attitudes of mothers the way that it did for fathers (Klar, Madonia, and Schneider Reference Klar, Madonia and Schneider2014), which suggests that mothers have related role-specific goals internalized, whereas for fathers the salience of parenthood depends on how the context reminds them of their fatherhood role (Park and Banchefsky Reference Park and Banchefsky2018; Stalsburg Reference Stalsburg2010). In short, extended parental leave or performing more parental tasks may not be able to increase mothers’ support for state services that benefit women and children any further.
Alternatively, women’s engagement in activities associated with being the primary caregiver may serve to affirm and reinforce traditional gender roles, highlighting perceptions of women’s “natural” fit for the primary role of caregiver (Wood and Eagly Reference Wood, Eagly, Olson and Zanna2012). Consistent with this view, evidence from some contexts suggests that mothers tend to support traditional values and gender roles more than non-mothers (Banducci et al. Reference Banducci, Elder, Greene and Stevens2016; Elder and Greene Reference Elder and Greene2012; Katz-Wise, Priess, and Hyde Reference Katz-Wise, Priess and Hyde2010), and also seem to prefer a conservative view of government intervention on moral issues (Eagly et al. Reference Eagly, Diekman, Johannesen-Schmidt and Koenig2004). Based on these cross-pressures, it makes sense to expect — relative to fathers — less change among mothers. If anything, we expect women with high engagement in parenting roles to mainly move toward a gender-traditional direction compared to women with less engagement in this role because the state-provided extended parental leave for women reinforces this traditional role for women.
We formulate expectations for the two political attitude questions in our survey (described in more detail in the Data, Measures, and Analysis section) that relate to the experiences of involved parenting. It is on these attitudes — those that most directly pertain to the state’s role in providing services for parents and equality for men and women in the household — where we should observe the most change. Thus, we hypothesize:
The more intense women’s engagement in the parental role (through parental leave or parental tasks), the less likely they will be to support gender equality in the family, and the more likely they will be to support a shift in government resources to pay for care within the family rather than paying for care through the welfare state. However, the effects will likely be small or non-substantial.
We have only two additional items in our survey related to politics; these items tap into support for a society with income differences and with private alternatives to welfare state provision. Because these items capture issues that are only indirectly connected to the goals or experiences of the parenting role, we have no strong theoretical reason to predict in which way parents’ views on these items will be altered. We include these items to explore the possibility that parents’ experiences with state-supported leave may lead them to support the state more broadly, or that parents may be making connections between parental leave experiences and other aspects of the state. Since these items roughly capture a left-right dimension, we can observe whether respondents move uniformly to the left or the right on all issues and, therefore, these additional attitude questions serve as a useful comparison point to our main items. Finally, the inclusion of these items also allows us to make the most use of our limited data. The first of these additional items measures supporting a society with small income differences and the second measures support for a society with private alternatives to the welfare state sector. In the Swedish context, this latter item refers to the public or private delivery of services; a society with more private alternatives means that the state would give more money to (non-state) for-profit or non-profit social service providers who implement welfare state programs or related services (Svallfors and Tyllström Reference Svallfors and Tyllström2019), potentially with some aspect of public regulation (Rothstein Reference Rothstein1998). This includes, for instance, private health care provision, private elderly care, or privatization of state monopolies such as Apoteket (the state pharmacy), which lost its monopoly in 2009 (Burström Reference Burström2015).
Effects of Parental Leave and Parental Tasks on Fathers
Extant research examines the effects of transitioning into the role of father for men. Fatherhood connotes the traditional diffuse role of being a man through traits and goals associated with agency, including seeking individual autonomy, status, and success (Dowd Reference Dowd2000), as well as an association with the role of protector and provider (Lamb Reference Lamb2000). Thus, men who have children support different policies compared to non-fathers, being particularly less supportive of government services, childcare, and school spending (Elder and Greene Reference Elder and Greene2012; Goodyear-Grant and Bittner Reference Goodyear-Grant, Bittner, Thomas and Bittner2017) and more supportive of traditional values (Elder and Greene Reference Elder and Greene2012) and gender roles (Katz-Wise, Priess, and Hyde Reference Katz-Wise, Priess and Hyde2010). Experimental evidence also supports these conclusions: when fathers received a prime suggesting that new policies could threaten their concerns as a parent, they became more supportive of national security and cuts to social services to reduce the deficit, while also becoming more opposed to reducing prison sentences (Klar, Madonia, and Schneider Reference Klar, Madonia and Schneider2014).
Our research focuses on how fathers’ attitudes toward the role of the state are affected when they fully inhabit a caregiving role; a role typical for women, but counter-stereotypical for men. Social role theory predicts that engagement in behaviors that conflict with traditional experiences for men will highlight new goals and motivations relevant to childcare, such as recognizing the hard work of caring for a child and the importance of state assistance in childcare both for themselves and other families, as well as, more generally, giving men “other-oriented” experiences or practice attuning to the needs of another person. Consistent with this perspective, Tavits et al. (Reference Tavits, Schleiter, Homola and Ward2023) have provided important causal evidence that fathers’ leave reduced sexist attitudes. Elder and Greene (Reference Elder, Greene and Whitaker2008) found that parental involvement (for mothers and fathers) predicted favorability toward spending on government services, health care, and childcare. Studying descriptive evidence of the policy preferences of Swedish legislators, Stensöta (Reference Stensöta2020) suggested that after being on parental leave, men expressed more interest in social policy than before taking leave, while women expressed high levels of interest regardless. In Norwegian data, fathers, who have opportunities for parental leave similar to Sweden, reported greater support for anti-poverty redistributive policies in the old-age pension system compared to non-fathers (Grødem and Kitterød Reference Grødem and Kitterød2023). Taken together, these results suggest that a shift in the specific role and intensive engagement within that role changes the political attitudes of men by making them more aware of gender inequality and policy benefits concerning the state’s role in helping others, especially parents. We thus form hypotheses on the two most relevant attitudes:
The more intense men’s engagement in the parental role (through parental leave or parental tasks), the more likely they will be to support gender equality in the family and to oppose a shift in government resources to pay for care within the family, rather than paying for care through the welfare state.
As with mothers, we do not have strong predictions in either direction on how engagement will affect fathers’ attitudes about societal income differences or preferences regarding private versus public service delivery; our analysis of these relationships is exploratory on these two items. As we argued above, while they do not directly relate to extended caregiving experiences, we still explore the possibility that parental leave shapes the left-right dimension more generally.
Data, Measures, and Analysis Plan
We employ the LNU for 2000 and 2010, which is unique in its panel design and measures. The LNU is a nationally representative sample of the adult population in Sweden, administered face-to-face in respondents’ homes, where 61.5% of the pooled sample of respondents answered both waves and can, therefore, be used as a panel. The LNU queried these respondents about their family relations, living conditions, health, working conditions, parental leave time, and parental responsibility in these two waves.
The eligible sample includes the mothers and fathers who answered the four attitude measures in both 2000 and 2010 (described below) and for whom we have data on parental leave or parental tasks (N = 1,175), 583 mothers and 592 fathers. The majority of the respondents were in dual-parent households when responding to the survey (75% in 2000 and 87% in 2010) and in heterosexual relationships (0.6% of the analyzed sample has a same-sex partner) for which our theorizing should best apply. We note that our respondents include those who had their last child before 2000 as well as those who had their first child or additional children between 2000 and 2010, although we explore different sample specifications in the Robustness Checks section below.
Sweden as a case provides an opportunity because many fathers become the primary caregiver. Sweden has a lengthy parental leave policy that allows 480 days of leave, with 60 days reserved for the father and mother respectively in our study period. The father also gets 10 days together with the mother and the baby directly after birth. A 2012 report showed that Swedish fathers used 23% of all paid days in 2010; this percentage was 11% in 2000 when our measures started (Swedish Social Insurance Agency 2012, 11). During the period 2001–10, 88% of fathers used some parental leave, while 48% of the fathers used more than the allotted quota (Marynissen et al. Reference Marynissen, Mussino, Wood and Duvander2019).
Dependent Variables
The LNU survey includes four questions that broadly measure various aspects of attitudes toward the role of the state (see Supplementary Information [SI], Table A1 for the Swedish wording). Respondents received statements about different types of societies in which Sweden might invest and were asked to judge each of them on a scale from “very bad idea (1)” to “very good idea (5).” The first society for consideration was “where income differences are small.” We call this Small Income Differences; the question measures respondents’ support for investing in a society with less inequality, which correlated closely with left-right ideology in Sweden at the time (Oscarsson and Holmberg Reference Oscarsson and Holmberg2016, 233).
The second question asked respondents “what do you think of the idea of investing in a society where men take as much responsibility for children and the household as women do?” We call this variable Gender Equality in the Family. In Sweden, support for a society with gender equality is high. In 2010, 83% thought an investment in gender equality was a good idea, a proportion that had not changed much from 2000 (Martinsson and Andersson Reference Martinsson and Andersson2021). Although women are 10 percentage points higher than men in their support, the gender gap has closed slightly over time (Naurin and Öhberg Reference Naurin, Öhberg, Andersson, Rönnerstrand, Öhberg and Bergström2019).
The third question relates to preferences for how the state should fund solutions for child and elder care. Discussion of such preferences has a long history in Sweden; the Christian Democrats, following a campaign pledge, first implemented policies between 1991 and 1994 that provided direct monetary benefits to parents who perform care themselves when the child is a toddler. In this system, “Vårdnadsbidraget,” which directly translates to “custody allowance,” follows the child so that families can choose whether to give taxpayer funds to the parent, often the mother, to stay home, or to give funds to a childcare facility or school; we note that this funding is entirely separate from parental leave. At the time of the LNU survey, the public debated the policy because the center-right government revived the idea between 2006 and 2014. Opinions toward the custody allowance policy correlate closely with the left-right ideological dimension and indicate whether someone supports state involvement in providing childcare versus leaving the childcare for the family to provide privately.
The specific LNU question about this policy asks respondents “what do you think of the idea of investing in a society where care for children and the elderly occurs to a greater extent within the family?” This item, which we call Shift to More Care in the Family, measures support for this policy. Because this policy is often contrasted with support for welfare state solutions, we interpret disagreement with the question as expressing the willingness to have the public welfare state assist with taking care of children and the elderly. Thus, the question helps us understand whether the respondent prefers welfare state care versus more family-oriented solutions.
The fourth attitudinal item asks respondents, “what do you think of the idea of investing in a society with more private alternatives within school, nursing and care?” We call this variable Private Alternatives. In Sweden, for-profit private companies often receive state funds to perform tasks and services available to everyone under the welfare state. Sweden’s welfare sector, for example, health care and education, was operated and owned by the public sector beginning in the mid-1900s. In the 1990s, privatization reforms changed the delivery of these services, aimed at improving consumers’ options and using market forces to improve service quality by introducing competition to the state model services (Rothstein Reference Rothstein1998). Notably, there has been an increasing political and popular opposition toward profit-taking by these private companies that, by the 2010s, provided over 20% of care and education services (Svallfors and Tyllström Reference Svallfors and Tyllström2019). Thus, this item asks respondents about the organization of the state and whether to allow more private options in welfare state provision. It is not an item that taps attitudes toward the size of the state directly or whether respondents like welfare state funding. Other studies have found that it captures the left-right dimension of politics, though they have not found gender differences on this question (Naurin and Öhberg Reference Naurin, Öhberg, Andersson, Rönnerstrand, Öhberg and Bergström2019, 403).
We do not combine these variables because of our separate theoretical expectations and because correlation coefficients among our dependent variables are low (about 0.2; see Table A2 and A3 in the SI for more details). The general tendency to agree or disagree with all four items is not strong enough to justify analyzing the data as a single latent construct (Cronbach’s α 2000 = 0.31, α 2010 = 0.27).
Independent Variables
Using the LNU data, we can operationalize engagement within the parental role in two ways.
Parental Leave
The data contains registered information on how many weeks of parental leave the respondent took with each child that is matched with information from the Swedish authorities. We created a continuous variable representing the total number of weeks of parental leave the respondent had with the child. Our sample includes respondents who took parental leave for their children who were 18 years or younger during the 2010 survey (i.e., children born after 1992). In practice, the 2000 data point measures the parental leave taken between 1992 and 2000, while the 2010 data measures the leave taken between 2000 and 2010.
Parental Tasks
The LNU includes three items about how frequently respondents complete specific parental tasks: how many times during a normal week (from 0 to 7) respondents “pick up children,” “drop off children,” and “put children to bed.” For ease of reporting, we add the “picking up” and “dropping off” into one variable (α = 0.72), which has a maximum value of 10 times/week, since this activity is mostly done on weekdays. We do not find significant differences when leaving them separate (not shown).
Analytical Strategy
Parental Leave
We analyze the effects of parental leave as a two-wave panel using multigroup structural equation modeling (SEM). SEM relies on the same underlying logic as OLS regression; however, it allows us to create a more complex structure around our data, which is particularly appropriate for panel data. First, SEM allows us to include autoregressive paths from variables at time one to the same variables at time two; such a model simultaneously captures how prior parental leave affects welfare state attitudes knowing that prior welfare state attitudes also likely have a substantial influence on later welfare state attitudes. This is facilitated by the fact that the parental leave information stems from official registry data and that it precedes measures of our dependent variables (Kline Reference Kline2016, 124–5). Second, SEM allows us to address selection effects or the idea that parents who choose parental leave are also those who have particular attitudes. As such, in our SEM, we model weeks of parental leave and attitudes as a function of income (in addition to other variables such as age, years of education, age of the youngest child, whether the respondent is working or not, having a partner or not, and number of children in the household in the Robustness Checks section) in 2000 and use the same variables plus attitudes from 2000 to predict weeks of leave in 2010. This specification helps us rule out the possibility of confounding variables that could affect our core hypothesized relationships.Footnote 1 Third, SEM can control for time-invariant unmeasured/latent variables that could bias observed relationships. Since omitted variables are likely because of the plethora of potential influences on the decision to take parental leave and for how long, it is an important benefit to be able to reduce the likelihood of omitted variable bias by using SEM. Fourth, SEM permits testing whether the residual variances of our dependent variables differ between men and women, in order to rule out the possibility that differing variances between the groups is the cause of different results observed between mothers and fathers. Finally, SEM regression coefficients account for measurement error in both independent and dependent variables (Byrne Reference Byrne2013), unlike in traditional multivariate regression analysis.
In short, using SEM allows us to model the relationship between parental leave and attitudes in 2010 (path δ in Figure 1 and in equation 4 below) while taking the corresponding relationship in 2000 into account (path β 1 and in equation 4), and as a function of the reverse effect of attitudes in 2000 on the parental leave taken between 2000 and 2010 (path γ 1, equation 3) to account for reverse causality. The extent to which past values predict current ones is captured in γ 2 (attitude autocorrelation) and β 2 (parental leave autocorrelation). The coefficient we are particularly interested in is δ, which is the effect of parental leave on attitudes in 2010.
Figure 1. Hypothesized relationship.
In the models, we also include the household income since this is a factor often posited as a reason why fathers take less parental leave than mothers, and is also often correlated with political attitudes. We enter income as a confounder in both the parental leave and attitude equations in 2000 and 2010 to account for a spurious relationship between these measures. The Swedish registry data measures this as fixed monthly income divided into income categories, with higher values indicating higher levels of income (16 categories, starting from 0, then 1–4,999 SEK/month, with 5,000 SEK increments up to incomes over 70,000 SEK/month). Missing data is replaced by multiple imputations (Arbuckle, Marcoulides, and Schumacker Reference Arbuckle, Marcoulides and Schumacker1996). Two hundred and fifty-eight observations in 2000 (16%) and 98 observations in 2010 (10%) needed replacement.
Thus, the four equations that we estimate are:
$$ \mathrm{Parental}\ \mathrm{leav}{\mathrm{e}}_{\mathrm{i},2000}={\unicode{x03B1}}_1+{\unicode{x03B6}}_1\mathrm{Cntrl}{\mathrm{s}}_{\mathrm{i},2000}+{\unicode{x025B}}_{1\mathrm{i}} $$
$$ Attitud{e}_{i,2000}={\alpha}_2+{\beta}_1 Parentalleav{e}_{i,2000}+{\zeta}_2 Cntrl{s}_{i,2000}+{\varepsilon}_{2i} $$
$$ Parental\ leav{e}_{i,2010}={\alpha}_3+{\beta}_2 Parental\ leav{e}_{i,2000}+{\gamma}_1 Attitud{e}_{i,2000}+{\zeta}_3 Cntrl{s}_{i,2010}+{\varepsilon}_{3i} $$
$$ Attitud{e}_{i,2010}={\alpha}_4+\delta Parentalleav{e}_{i,2010}+{\gamma}_2 Attitud{e}_{i,2000}+{\zeta}_4 Cntrl{s}_{i,2010}+{\varepsilon}_{4i} $$
We first had to establish that the variance of the dependent variable is similar in the two groups (measurement invariance) to ensure that comparisons can be made between men and women. Therefore, we ran one structural equation model for each dependent variable (the same structure as the one shown in Figure 1) and tested whether the residual variance of the dependent variables differed between women and men. No test was significant at the 0.05 level, which means coefficient comparisons between the groups are more likely to be valid; see also the observed standard deviations in Table 1).
Table 1. Descriptive statistics
Notes: The table report shows standard deviations in parentheses. The minimum and maximum values are the following: The four dependent variables all range from 1–5; Total parental leave: Mothers (2000): 0–418, (2010): 0–1037, Fathers (2000): 0–314, (2010): 0–314; Times/week picking up/dropping off child: 0–10; Days/week putting children to bed: 0–7; Income (in 1,000 SEKs): Mothers (2000): 0–55, (2010): 0–95, Fathers (2000): 0–62, (2010): 0–130. See more details in SI Table A4.
Parental Tasks
In contrast to the parental leave data, the parental task measures are based on survey responses collected at the same time as the dependent variable. We therefore report the correlations between parental tasks and attitudes in 2000 and 2010. These correlations indicate whether there is a relationship between various parental tasks and political attitudes, and whether they differ between women and men, but do not provide direct causal evidence.
Results
Table 1 gives the descriptive statistics of the variables by respondent gender and year of survey and shows that on our attitudinal dependent variables, Swedish respondents are generally favorable toward a strong social welfare state and gender equality. The most support can be found on Gender Equality in the Family for both men and women, indicating that we are facing high ceiling effects to move such positions; while Shift to More Care in the Family has the least support overall. While mothers and fathers have similar means and standard deviations for both 2000 and 2010, as in other countries, women tend to be slightly more supportive of the state than men in Sweden, particularly when it comes to Small Income Differences. Mothers’ tendency to be more positive about Small Income Differences compared to fathers constitutes the largest gender difference of our four outcome measures, which is also the only gender gap that is statistically significantly different (statistics not reported). However, women are slightly more supportive of Private Alternatives in the welfare state compared to men. Shift to More Care in the Family is the least supported aspect of the welfare state and it shows few gender differences.
Mothers and fathers clearly differ on parental leave and parental tasks, our main independent variables. In 2000, mothers in this sample had taken, on average, 51 weeks of parental leave, while fathers took only six weeks. The number of weeks mothers and fathers had taken between 2000 and 2010 was 57 and 13, respectively. This means that the proportion of all parental leave weeks taken by fathers increased in the sample to a similar extent as in the population, from 10% to 19%, but mothers also show a slight increase in parental leave weeks in this time frame, reflecting the overall higher number of leave weeks available to parents in 2010 compared to 2000.Footnote 2 Mothers also engaged in parental tasks more often than fathers in 2010, even though the gap was smaller than in 2000. For example, mothers dropped off or picked up children at school or daycare on average 3.0 days a week (same in 2000), while fathers dropped off and picked up 2.3 days a week (1.8 in 2000). The average age of the respondents’ youngest child was around eight years in 2010, which means that many children likely walk to school without their parents, which is common in Sweden. Furthermore, mothers put children to bed 3.9 days a week in 2010 (5.0 in 2000), while fathers put children to bed 3.4 days a week (3.7 in 2000). Table 1 also shows the means of our confounding variables, such as age and income.
We first examine the bivariate correlations between parental tasks and our dependent variables (SI Tables A2 and A3). For mothers, dropping children off relates negatively to Gender Equality in the Family (r = −0.11 in 2000, r = −0.06 in 2010), and the same is true for putting children to bed (r = −0.07 in 2000, r = −0.05 in 2010). For fathers, putting children to bed was positively related to Gender Equality in the Family in 2000 (r = 0.12), but non-existent in 2010. We note that the significant correlations align broadly with the hypotheses, however, the small bivariate relationships strongly suggest that performing parental tasks does not significantly change attitudes toward the state.
We next examine the correlations between parental leave and political attitudes to provide an initial test of our hypotheses and to confirm the use of multigroup SEMs for the parental leave data (Jöreskog Reference Jöreskog1971). We observe relatively moderate or weak correlations between parental leave with the studied attitudes; however, some are slightly stronger. For example, support for Shift to More Care in the Family and Private Alternatives, correlates positively with parental leave for mothers in both measurement periods. For fathers, the largest correlations indicate a positive relationship between parental leave and Gender Equality in the Family and a negative relationship between leave with Shift to More Care in the Family and Private Alternatives. We thus display preliminary evidence of our hypotheses that fathers become more favorable toward the state, while mothers less so, and, crucially, we confirm that the main relationships differ for mothers compared to fathers, suggesting that separate SEMs for mothers and fathers are warranted.
Next, we test the expected relationships from Figure 1 using SEM for the parental leave variable. The top third of Table 2 displays the unstandardized coefficients and standard errors from four SEMs for mothers (one for each dependent variable), while the middle third lists the results for fathers. The most relevant coefficient is the δ path, which examines the relationship between leave and the appropriate dependent variable, controlling for respondents’ income in 2000 and 2010. This path is shown in Figure 1 and is depicted in the shaded rows for mothers and fathers in Table 2. In the lowest portion of the table, we report overall model statistics and global fit indices (see more details in SI Section B). In Figure 2, we visualize the predicted relationship between parental leave and the four political attitudes based on the structural equation models.
Table 2. Relationship between parental leave and attitudes toward the state
Notes: Unstandardized coefficients. The two rows where we report our central relationship is shaded light gray. Parental leave is measured in 100s of weeks. The LR test is a test of the difference between the model vs. the saturated model. CFI = comparative fit index; RMSEA = root-mean-square error of approximation; CI = confidence interval; SRMR = standardized root mean square residual. CFI > .95, RMSEA < .06, SRMR < .08 are considered good fit.
+p < 0.1, * p < 0.05, ** p < 0.01.
Figure 2. Predicted relation between parental leave (number of weeks taken) and political attitudes in 2010.
First, for mothers, we find no significant effects of taking parental leave on views of the state. Table 2 shows that the models for mothers have small coefficients and none of them are statistically significant. Only one of the coefficients approaches significance (β = −0.10, p = 0.08), Gender Equality in the Family. The coefficient’s direction supports our hypothesis that, as their parental leave becomes longer, mothers will become slightly more supportive of traditional gender roles and therefore less supportive of Gender Equality in the Family. Also not significant and very small, but supportive of mothers tending more toward traditional attitudes, mothers are more in favor of a Shift to More Care in the Family (β = 0.04, p = 0.54). The direction of the coefficients for the final two items suggests that mothers with more parental leave are (insignificantly) less in favor of Small Income Differences (β = −0.03, p = 0.68) and against Private Alternatives (β = −0.09, p = 0.20).
For fathers, our models show two significant effects. With more parental leave, fathers turn against a society that favors Shift to More Care in the Family (β = −0.57, p = 0.00) (see the gray-shaded row). The coefficient for Gender Equality in the Family is small and not significant, but goes in the direction of the hypothesis — that fathers who take out more parental leave would be more in favor of Gender Equality in the Family (β = 0.12, p = 0.46). The other significant effect shows that with more leave, fathers become less supportive of Private Alternatives (β = −0.50, p = 0.01). This is the largest and most stable coefficient across various ways of modeling the relationship. This issue largely captures a left-right ideological phenomenon and implies that fathers who take out more parental leave become more oriented toward the left by rejecting private alternatives for the welfare state. A potential interpretation that fathers with parental leave are moving overall to the left is, however, not supported by their views on our other left-right issue, Small Income Differences: fathers are more negative about redistribution at higher levels of parental leave (β = −0.16, p = 0.38).
We note that the effects are overall small. To take the largest coefficient (i.e., fathers changed attitudes toward Shift to More Care in the Family) as an example: the parental leave variable is scaled in terms of 100s of weeks; the model thus suggests that one standard deviation (22 weeks) increase in parental leave results in a −0.13 scale point shift on a five-point scale (−0.12 SDs).Footnote 3 However, we also note that Figure 2 clearly shows that the coefficients in the models of mothers are overall smaller than the corresponding coefficients for fathers. This supports the idea that stepping into counter-stereotypical experiences within a new role — for fathers, caregiving is not consistent with the prescribed role of being a man — matters more to attitudes than experiences confirming an already prescribed role — for mothers, caregiving is consistent with the diffuse role of being a woman. We also note that the reverse paths, where attitudes in 2000 predict 2010 parental leave levels (γ 1), are small and non-significant. While this does not rule out endogeneity problems, it supports the notion that reverse causality is not biasing the main estimates.
Robustness Checks
We ran a number of different robustness checks and sensitivity tests in order to ensure that our findings and results hold true across different possible critiques. None of these analyses, which are described briefly below and in more detail in the SI, diverge substantially from our main analyses.
To mitigate the possibility that people who take more parental leave differ in political attitudes from those who take less, we created models using additional possible confounding variables: age, years of education, age of the youngest child, whether the respondent is working or not, having a partner or not, and number of children in the household (which might include non-biological children). For example, previous analyses of selection into parental leave in Sweden have shown that education level and employment situation are important predictors (Sundström and Duvander Reference Sundström and Duvander2002). Those analyses support the results presented here, with the same or slightly stronger estimates, but poorer goodness-of-fit (see SI Sections C and D). While we argue that income is a relevant control, we also reproduce the results using models that do not control for income levels in 2010 (see SI Tables C1 and C2). A related problem is that we are unable to account for all confounding variables, even though SEM models the potential feedback loop between parental leave and attitudes. For this reason, we ran change score regressions that focus solely on within-person changes to rule out time-invariant factors we could not capture in the SEM models (see SI Section H). These results are very similar to our main specifications.
For some parents, there was a lag time between parental leave or their last child and the data collection. Therefore, in SI Section J, we explore the interaction between time since the last parental leave experience and length of parental leave. The direction of the interaction is in the expected direction for fathers in the case of Shift to More Care in the Family and Private Alternatives. That is, change is greater when fathers are closer to their most recent leave experience, combined with longer parental leave, though it is only significant on Private Alternatives. We could interpret this to mean that parental leave effects are strongest when recently experienced and there is a diminishing effect over time, however, because we have low power in these tests, they should only be seen as suggestive findings. A related concern is respondents who already had their last child before 2000 would bias the sample; however additional robustness checks excluding respondents who had their last child prior to 2000 demonstrate that the results are similar (see SI Table G1). Furthermore, while our main comparison is within the parental sample because we are interested in different levels of engagement with the parental role, there is still a concern that we are picking up broader shifts in attitudes in the population. In SI Section I, we analyze change score differences between parents and a matched non-parental sample and find that the same two significant estimates for fathers are also reproduced under this specification. Finally, to account for different possible measurements of parental leave, we ran alternative models where the parental leave in 2010 sums up the leave taken between 1992 and 2010. These show that the direction is consistent, but a sign that the link between parental leave and attitudes is weaker as time passes (see SI Section C). Using different cutoff points or treating the variable as nonlinear did not change our conclusions (see SI Sections E and F).
Concluding Discussion
Our results show that, in Sweden, some attitudes of fathers are affected by how much parental leave they take, while mothers’ attitudes are not shaped by parental leave in the same way. The differential findings on mothers and fathers are consistent with the expectations of Social Role Theory: diffuse and specific roles matter, in addition to experiences within those roles. Experiences that affirm existing gender roles (as in the case of mothers) will have less of an overall effect than experiences that disconfirm existing gender roles (as in the case of fathers). For women, motherhood is a role that overlaps with women’s diffuse gender role and it makes sense that even these particular experiences of parental leave did not alter mothers’ attitudes much. Moreover, women in Sweden are also working and many are navigating male-stereotypical roles in addition to their roles as mothers, suggesting that the small effects might result from this role conflict and its cross-pressures.
For men, full engagement in parenting requires activities that differ from the diffuse role of being a man, experiences that illuminate new goals surrounding the importance of caring for children, and having state support to do so. More visible change for fathers than mothers supports the idea of how gendered experiences matter. The effects of parental leave among fathers are not large, however, they are substantively interesting — especially considering that these fathers are likely not performing a female-typical role throughout a given day. Even if fathers take leave and care for the children most of the time, they still live in a world where other male-typical roles are available when they leave the home or for them to observe. The consequences of counter-stereotypical experiences are therefore likely balanced by the confirmation of male stereotypical roles when they spend time away from the child. It remains for future work to think further about how such dueling or shifting roles and experiences might influence political attitudes.
Engagement in the parental role influenced fathers’ attitudes toward the role of the state in areas related to their prolonged experiences with childcare. Fathers who took parental leave in our study became less positive toward care being done in the family instead of by the welfare state. Further work should investigate the mechanisms of this change. We suspect that fathers realize the challenges of childcare and desire state assistance without wanting to provide more assistance themselves, considering there was no effect on the gender equality item. Thus, the state-driven experience of parental leave seems to inspire fathers’ openness to and desire for more state-driven solutions in childcare. While we do not measure this directly, the increased desire for state solutions is also confirmed by one other noticeable effect of parental leave, which was that fathers became more opposed to private alternatives to government programs, shaping them to become more left-oriented on a left-right dimension of state influence. We suspect that fathers who used state-provided parental leave learned that state regulation of care works, and thus, state-provided experiences surpass private options. Interestingly, fathers do not change their views on gender equality; though they become more positive, the effect was not significant. Of course, our Swedish respondents, both mothers and fathers, were highly in favor of gender equality from the start of the study, suggesting ceiling effects. Future research should test the robustness of this null effect and probe it further. Overall, our findings confirm that parental leave strengthens fathers’ openness to the role of the state in policymaking, yet future research should further explore whether and how support for selected policy aspects is shaped for fathers (and differently for mothers) when they experience the benefits of state policies.
There is an effect of taking parental leave on attitudes, but no effect of performing occasional parental tasks, such as picking up and dropping off children at school or putting children to bed. This means that fully inhabiting the role of a primary caregiver through parental leave is associated with a change in attitudes toward the role of the state; performing parental tasks does not induce that kind of change. However, an ideal data collection would also include more detailed measures of how parents fulfill various tasks to better understand how fully they inhabit their roles.
By studying Sweden, we are taking advantage of a context where fathers and mothers are able to take on a primary caretaker role for extended periods of time. We are also likely testing a difficult case (Naurin et al. Reference Naurin, Stolle and Markstedt2023). Since men are already socialized into taking responsibility for the day-to-day care of children in Sweden more than in other countries even at the time of our study, the effect of parental leave or engaging in parental tasks should be smaller. Put differently, in a circumstance where gender equality norms are strong, fully inhabiting a gendered role should have less effect. Still, even so, the effects of changing attitudes of fathers toward the role of the state are robust. Whether or not the enactment of parental leave for fathers is politically feasible and viable in other countries and institutions beyond that of Scandinavia is, of course, a different question. However, our research suggests that such leave policies do shape fathers’ societal and political views and have the potential to do so in a less egalitarian context.
Furthermore, we note that our study follows the parents over 10 years, and many of them had the experience of parental leave quite some time before they answered our focal dependent variables in 2010. It is possible — and indeed we show in additional materials — that our findings would have been stronger had our measurement points been closer to each other. Finally, our data speaks best to the experiences of heterosexual couples in partnered relationships. Research has found that same-sex couples are more liberal on gender role attitudes and are more equal in dividing domestic labor (Shechory and Ziv Reference Shechory and Ziv2007), though the data are less clear on whether single parents also espouse liberal views of gender equality (Kågesten et al. Reference Kågesten, Gibbs, Blum, Moreau, Chandra-Mouli, Herbert and Amin2016). Again, we encourage future research to tease out the ways that these different types of gender roles intersect, while noting that, though small, the presence of same-sex couples in our dataset who already likely have gender-equal role experiences makes this a harder test of our hypotheses.
The data we use does not include many, nor very detailed, measures of political attitudes. The focus of the LNU is on family relations, living conditions, health, and working conditions rather than on political attitudes or vote choice. The fact that fathers show changes for Shift to More Care in the Family suggests that if the survey had included more attitudes that were closely connected to the experiences of caring for a child, we would have detected more effects. Such attitudes worth considering for future studies could include support for or opposition to policies that allow the parent to combine work and leisure with parenthood, including better support and compensation for parents who take leave, investment in daycare, and longer or more flexible leave even as the child ages.
Future work should study how different experiences with the parenting role have different effects on views of the state. For example, when a child has disabilities, the parent likely experiences the state and its policies differently than parents of neurotypical children, particularly in the realm of education, health care, and home and community-based services for long-term care. It remains for empirical work and gender scholars thinking about Social Role Theory to delve into the details of the many different types of experiences with the parental role, how mothers’ and fathers’ existing social roles as men and women might condition their responses, and the specific mechanisms through which priorities and goals shift and attitudes change. Since our results indicated a change in attitudes closely related to role experiences, as well as in an attitude we thought would be further from the role, it is worth investigating how respondents see these policies as connected to their experiences, or not.
Lastly, our research provides insight into the psychological underpinnings of the feedback loop that policies create. In particular, our study draws attention to the effects of roles and the experiences within them that can be created by the state and also cause relevant attitude change. By providing their citizens with the option of being full-time parents with state support, our evidence suggests that the Swedish state affects attitudes toward the role of the state among fathers, and this result has the potential to hold in other less gender-equal contexts.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1743923X25000054.
Acknowledgments
We are grateful to SOFI at Stockholm University for providing access to the data. We would like to thank Caroline Roethlisberger for her valuable research assistance and Amanda Bittner and Elizabeth Gidengil for their insightful comments. This research was supported by the Swedish Research Council (grants 2015-01546 and 2022-06355) and the Knut and Alice Wallenberg Foundation (grants KAW2017.0245 and KAW2022.0186).
Open access
