Lingering Male Breadwinner Norms as Predictors of Family Satisfaction and Marital Instability

: Scholars have assumed that as gender revolutions are completed and societies achieve advanced levels of gender egalitarianism, married persons become happier, and marriages become stable. This study investigates how the norms about gender roles are associated with marital instability. The analysis is based on two propositions: (1) marital dissolution is an outcome of two rather distinct processes, deterioration of marital quality and formation of a decision to leave a marriage, and (2) the antithesis of advanced gender egalitarianism is a set of lingering male breadwinner norms, not gender inequality often manifested by working women performing second shifts. The data are from 68 national surveys conducted in 2002 and 2012 through ISSP coordination, and the sample of person-level analysis is restricted to ages 30–49, supposedly in the life cycle stages of family formation and expansion. The norms of gender roles are classiﬁed into four types: traditional norm, prescribing gendered division of labor; lingering male breadwinner norm, emphasizing men as the primary breadwinners while allowing ﬂexibility of women’s roles; super woman norm, prescribing women to perform double roles; and egalitarian norm, emphasizing equal sharing of roles. At the country level, aggregate variables were constructed by calculating the percentage of adults who held each type of norm. The results strongly support the prediction that the male breadwinner norm at the societal level is detrimental to marital quality, while persons holding the egalitarian norm are most satisﬁed with their family lives.


Introduction
While poor marital quality is routinely considered the underlying cause of marital breakups (Rogers and Amato 2000;Schwartz and Han 2014;Teachman 2010), the literature on marital dissolution tends to focus on the process of dissolving per se (i.e., the decision to leave a marriage), often exploring who initiates the process (Kalmijn and Poortman 2006;Sayer et al. 2011). Two dominant theoretical perspectives contributed to analyzing such a process. The sociological exchange theory, which shares common propositions with the new home economics theory in economics, postulates that marital dissolution is an outcome of married persons' cost and benefit assessments. For example, in families where the wives are employed outside their homes, the gains provided by marriage are smaller for both spouses but particularly for women. This model contrasts with families where work and family roles are divided by gender, and the spouses benefit from each other's production (Becker 1991;Oppenheimer 1997). According to this theory, other things being equal, marital instability would be greater when wives were employed (Schwartz and Han 2014;Teachman 2010). The modernization theory postulates that rising individualism, accompanied with material affluence in a society, motivates people to pursue their needs of self-actualization and helps liberate them from the institutional constraints of marriage. According to this theory, marital instability will be higher in societies where individualism is highly valued and among people with higher education (Lesthaeghe 1995;Thornton and Young-DeMarco 2001;van de Kaa 2002). In both theories, marital quality 2 of 17 is not an integral part of the marital dissolution process; rather, it is treated as an external factor that requires its own theories and explanations.
One problem with not explicitly incorporating marital quality in the research of marital dissolution is the potential to make misleading conclusions. Factors affecting the costs and benefits of leaving a marriage may have very different dynamics affecting marital quality. For example, liberal attitudes in general or liberal gender role attitudes may increase the propensity of leaving a marriage (Lesthaeghe 1995) but at the same time may safeguard marital quality by allowing flexibility in dealing with family problems (Kaufman 2000;Zuo and Tang 2000). The independence hypothesis, postulating women's economic independence will increase marital instability, poses an interpretation puzzle of whether women's economic resources hurt marital quality or simply facilitate women's departure from marriages of poor quality (see Rogers 2004). If not conceptually distinguishing between the issues of marital quality and the dynamics of dissolving a marriage, any research predicting marital dissolution is prone to erroneous understandings. Methodologically, it would be a challenge to combine the two rather distinctive processes into a single theoretical framework. However, this study assumes that research predicting marital instability must address both processes at the same time and clarify the potentially conflicting effects of any predictors.

Liberal Gender Role Norms and Marital Instability
According to the modernization theory, individualism and economic independence help overcome institutional constraints to achieve personal freedom. An expansion of higher education has indeed promoted the ideals of individual freedom, including approval of divorce as a personal choice. It appears that there is no return in the historical trajectory of the values appreciating individualism and personal choices. However, over time, it has become a theoretical puzzle whether the force of individualism or liberalism, once tied with self-actualization, might be the underlying reason for family erosion or family decline in ever more affluent contemporary societies (Lewis 2001; Thornton and Young-DeMarco 2001). Researchers came to question whether such social trends as more infrequent marriage, more cohabitation and single-person households, lower fertility and higher divorce were the products of expanded personal freedom or a manifestation of certain structural social problems.
A layer of complexity is added by the trend of what is called the marriage (and divorce) divide. People of a lower socioeconomic status are increasingly less likely to marry, and once they marry, they are more likely to divorce than people with college educations and higher earnings (Martin 2006;Wilcox and Wang 2017). Some theorists blame particular social groups for their irresponsible individualistic attitudes, while others blame poverty or limited economic resources in the context of rising inequality that makes it hard for families to sustain themselves (Murray 2012;Wilcox and Wang 2017). In recent years, the value of individualism or liberalism has been hailed again as the force that might help revive the family institution. This time, it is a particular form of liberalism: gender egalitarianism. The argument is that the trend of family erosion is suspended or being reversed most visibly in those societies with the strongest gender egalitarianism, notably seen in Scandinavia (Esping-Anderson 2016; Goldscheider et al. 2015). According to the argument, societies lagging in attitudinal transitions to advanced levels of gender egalitarianism also lag in the trend toward family revival.
The apparently conflicting associations between liberal norms and marital instability may be resolved by conceptually distinguishing between the process of shifting marital quality and the process of dissolving a marriage. When not carefully disentangling these two processes, researchers may blame the individualistic attitudes of those who struggle with contradictions in gender role prescriptions and experiences of poor marital quality (see Murray 2012). In addition, poverty or even rising inequality may be associated with higher marital instability and poor quality, not because spouses lack access to material resources, but ultimately because they have not solved the contradictions between their expectations and the realities of performing their prescribed roles (see Lowrey 2014; Wilcox and Wang 2017).
The literature discussed above suggests that across many societies, attitudes toward gender roles may shift progressively from traditional to transitional to egalitarian phases, and in many societies, historically observed marital instability may have increased for a while and then decreased somewhat to eventually reach an equilibrium point (as marital conflicts would exist regardless of gender role arrangements) (Esping-Anderson 2016; Pessin 2018). However, the rate of marital dissolution should also be recognized as being determined by multiple other contextual factors, including public welfare programs, which would affect the costs and benefits of being unmarried or a single parent as opposed to being married.
As described above, theories have addressed the associations between liberal norms and marital instability by focusing on the process of dissolving a marriage. This study assumes that the quality of marital relationships is an integral part of marital instability and aims to theorize the associations between marital quality and gender role norms at both the societal and individual levels. After discussing theoretical backgrounds of how gender role norms may affect marital quality, we will empirically examine the associations of gender role norms with the attitudes toward both divorce and marital quality. We hypothesize that strong gender egalitarianism will improve marital quality but, at the same time, will increase approval of divorce.

Dimensions of Gender Roles and Typology of Gender Role Norms
Social surveys usually have multiple questions on attitudes toward gender roles, with each question asking respondents whether they agree or disagree with a specific role arrangement and comparing their views with a scale representing traditional and liberal attitudes. A common practice is to sum the scores from multiple questions and characterize the respondents along a linear scale as either traditional or liberal (Brinton and Lee 2016;Brooks and Bolzendahl 2004;Rogers and Amato 2000). However, going deeper into the survey questions, they address different aspects of role arrangements in such a way that people who express liberal attitudes in one question are not necessarily equally liberal in terms of other questions. For example, people who are against gendered division of labor (a man being the achiever outside the home and a woman taking care of her home and family) do not necessarily believe that children and family members will not suffer when their mother (or wife) is employed outside the home (Cotter et al. 2011;Cunningham 2008;Davis and Greenstein 2009).
In principle, gender roles consist of four dimensions: work (i.e., income earning) versus family (i.e., housework and childcare) roles for each of the two male and female gender groups. The distinction is necessary, because people's attitudes are not consistently symmetrical between the roles of women and men and between the areas of these roles, as widely observed across many societies and often referred to as a stalled gender revolution (England 2010;Knight and Brinton 2017;Lee 2016;Raymo et al. 2015;Takeuchi and Tsutsui 2016). For example, strong support for women's gainful employment outside the home may accompany strong support for men's family roles but not necessarily pair with strong support for men being exempted from provider roles. Studies arguing for men's adaptation to changing women's roles advocate men to share homemaking responsibilities with women but rarely advocate men becoming secondary earners or full-time homemakers.
Apparently, two archetypical attitudes are distinctive: most traditional and most liberal. The former endorses strict division of labor by gender between work and family roles, and the latter rejects gendered division of labor in any form while endorsing equal sharing of the roles in both areas. The most liberal type is referred to as an advanced level of gender egalitarianism and, at the societal level, is exemplified by Scandinavian societies. Marital stability has gained momentum in these ultra-egalitarian societies in recent years (Esping-Anderson 2016; Esping-Andersen and Billari 2015). Yet, associations between gender role norms and marital instability are considerably more complex than what can be summarized in a few rules. First, as is well known, marital instability is relatively low in societies with strong traditional norms (OECD 2019). As such, during the processes of urbanization and economic growth, the trend of family erosion appeared to be persistent until the signs of reversal emerged in recent decades.
These observations lead to a speculation that confusions and contradictions in gender roles arise and expand during social transitions while paralleled by the transition from most traditional norms to non-traditional norms. Rising marital instability seems to be an outcome of massive economic, social and attitudinal transformations which create conflicts in role expectations among different social groups and across individuals' relations. The literature often assumes that the antithesis of gender egalitarianism is gender inequality, which is often translated to mean the conditions where men's attitudes and behaviors reflect the legacy of patriarchal social arrangements. For example, men's housework time does not increase proportionally to women's labor market activities (Breen and Cooke 2005;Hochschild and Machung 2003;Thebaud 2010). Yet, the true antithesis of gender egalitarianism may be normative refusal to equally share roles on the part of both genders. The male breadwinner norms are oppressive of those men who perceive they are unable to fulfill the expected roles (Lee 2020). Even when the neoliberal economy is being restructured in such a way that men's positions in the labor markets deteriorate as opposed to those of women, lingering in many societies is the norm that men should be the primary or sole breadwinners.
Many women as well as men hold the view that men are the primary breadwinners, even when they allow much flexibility in women's roles. It is only logical that men will not be able to fully share the responsibilities of family roles unless their burdens as the primary breadwinners are lessened and they are liberated from provider anxiety. Phrased differently, under the male breadwinner norms, few men can afford to take away substantial time and energy from their work roles. Whether wives' employment and earnings will pose psychological threats or provide a sense of relief and support for the husbands depends on the normative structure of the society (Zuo and Tang 2000). The former is more likely in the environment of male breadwinner norms. To summarize, in the contemporary neoliberal economies with expanding service sectors, strict male breadwinner norms will press both men and women to experience disappointments with their family lives. This study will demonstrate data supporting this hypothesis.
Beside Esping-Andersen's studies, other recent studies also theorize and document curvilinear associations between gender role attitudes and marital instability (Goldscheider et al. 2015;Pessin 2018). Using aggregate measurements of attitudes by region and year in the United States, Pessin (2018) found an inverted U-shape relationship between the evolution of egalitarian gender role attitudes and marital instability, with the highest instability observed at the turn of the 21st century. Rogers and Amato (2000) similarly reported deterioration of marital quality between 1981 and 1997, finding that greater work-family conflicts explained this change. Using the typology of gender role norms at both the societal and individual levels, this study sheds light on how transitions in gender role norms may be responsible for the curvilinear patterns of marital instability over time.

Research Questions and Design of Study
This study examines the effects of gender role norms on each of the two components of marital instability. The major research questions to be addressed are as follows. How are the types of gender role norms, measured either at the individual level or at the societal level, associated with marital quality? In particular, does the male breadwinner norm negatively (and the advanced gender egalitarian norm positively) affect marital quality? How do relative incomes between spouses affect marital quality? How are gender role norms associated with attitudes toward divorce? Do all these associations differ by gender? Lastly, at the societal level, do the divorce rates vary by the composition of different types of gender role norms among adults?
Unlike most other empirical research, which intends to fully explain the dependent variable, this study adopts an analytic approach, focusing on the key independent variablegender role norms-and its associations with a few different dependent variables. As such, in each equation analyzing a dependent variable, most of the important proximate determinants are missing. In other words, investigating the exact mechanisms by which gender role norms are associated with the outcome variables, including marital quality, attitude toward divorce and crude divorce rate, is beyond the scope of this study. The key question in the analysis is whether the norms of gender roles at the societal or individual level matter for marital instability, either for change in marital quality or for propensity to leave a marriage. Once the associations are documented, the detailed mechanisms will warrant future studies.

Data and Methods
Data are from the 2002 and 2012 Gender and Family Surveys coordinated by the International Social Survey Programme (ISSP). Every year since 1985, ISSP has helped administer the equivalent of a General Social Survey in the United States across dozens of countries with a particular yearly theme. There are 10 different themes, and each is repeated every 10 years. The Gender and Family module was used in 2002 and 2012. The analysis includes data from all the participating countries (the terms "society" and "country" are used interchangeably) except the Philippines, where divorce is officially not allowed, and Spain in 2012, where the measurement scales are not comparable to those in the standard questionnaire. In all, data are from 68 surveys, where 9, 11 and 24 distinct countries participated in 2002, in 2012 and in both years, respectively, totaling 44 countries. Refer to ISSP (2021) for the list of participating countries.
As this study investigates gender role norms, the analysis focuses on people from the ages of 30-49 who are in the life cycle stages of family formation and expansion and expect to bear and raise children. It is presumed that couples in this age group are pressed to fulfill their family obligations and would feel role conflicts most acutely with multiple responsibilities at work and at home. For the same reasons, marital conflicts or dissolution would be most consequential for the couples and their children. The sample includes more than 32,000 women and men for the 2 years combined, but the number of cases varies somewhat across the equations due to the design of the survey excluding particular respondents for different questions. All the person-level analysis is performed separately for women and men.

Person-Level Variables
Types of gender role norms: Based on the discussions above, we distinguished four distinct types of gender role norms: traditional for both genders (referred to as traditional norm); traditional for men but liberal or flexible for women (referred to as male breadwinner norm or MBN); liberal or flexible for men but traditional for women (referred to as super woman norm or SWN); and liberal or flexible for both genders (referred to as egalitarian norm). The classification was based on two survey questions: "A man's job is to earn money and a woman's job is to look after the home and family" (dol) and "All in all, family life suffers when the woman has a full-time job" (fsfr). Figure 1 illustrates the classification. The responses for the division of labor question were grouped into two sets: combining "strongly agree", "agree" and "neither agree nor disagree" (comprising the first two types) versus "disagree" and "strongly disagree" (comprising the last two types). For the latter variable, the category of "neither agree nor disagree" was grouped with "disagree" and "strongly disagree" on the side of first two types, but on the side of the last two types, "neither agree nor disagree" was grouped with "agree" and "strongly agree". In this way, the first and last types unequivocally presented most traditional and most egalitarian attitudes, respectively. The numbers are the percentages of the cell frequencies in the sample. The four norm types comprised 26, 21, 24 and 30 percent of the respondents, respectively. frequencies in the sample. The four norm types comprised 26, 21, 24 and 30 percent of the respondents, respectively.
A man's job is to earn money, and a woman's job is to look after the home and family (  To rephrase, the traditional norm endorses the division of labor between the two genders, with men devoted to paid work outside the home and women to housework (26). The male breadwinner norm (or MBN) is flexible about women's employment (disagreeing that family life will suffer if women have full-time jobs) and yet endorses gendered division of labor, emphasizing the men's breadwinner role (21). The superwoman norm (or SWN) rejects gendered division of labor and yet shows inflexible attitudes toward women's roles (agreeing that children and family life will suffer if women are employed) (24). Finally, the egalitarian norm rejects a division of labor by gender and is flexible about women's employment (30). The traditional and egalitarian norms form two theoretically opposite ends, and MBN and SWN are likely to be observed during the social transitions from traditional to egalitarian types. This scheme was supported by the results from latent class analysis (LCA) in preliminary analysis. The analysis used five questions from the 2012 Gender and Family Survey, including the following three with the above two: "Both the man and woman should contribute to the household income" (reversed) (win); "A job is all right, but what most women really want is a home and children" (wom); "A pre-school child is likely to suffer if his or her mother works" (mw). Responses consisted of 5 categories from "strongly agree" to "strongly disagree", and the values were coded in such a way that a lower score represented a traditional attitude and a higher score a liberal attitude, ranging from 1 to 5. The analysis distinguished four different latent classes or normative types, and Figure  2 shows the mean predicted scores of the five attitude questions among the four groups of latent classes. The figure clearly corroborates the above classification of four normative types created with the two questions (dol and fsfr). Figure 2 also suggests that all four latent classes (or normative types) strongly support women's contribution to the household income, which appears to not necessarily mean women's full-time employment. The four normative types also do not have substantial distinctive mean values regarding what most women really want. In all, the scheme used in this study seems to be more advantageous than the latent class analysis approach because it is easily replicable and conceptually more transparent. To rephrase, the traditional norm endorses the division of labor between the two genders, with men devoted to paid work outside the home and women to housework (26). The male breadwinner norm (or MBN) is flexible about women's employment (disagreeing that family life will suffer if women have full-time jobs) and yet endorses gendered division of labor, emphasizing the men's breadwinner role (21). The superwoman norm (or SWN) rejects gendered division of labor and yet shows inflexible attitudes toward women's roles (agreeing that children and family life will suffer if women are employed) (24). Finally, the egalitarian norm rejects a division of labor by gender and is flexible about women's employment (30). The traditional and egalitarian norms form two theoretically opposite ends, and MBN and SWN are likely to be observed during the social transitions from traditional to egalitarian types. This scheme was supported by the results from latent class analysis (LCA) in preliminary analysis. The analysis used five questions from the 2012 Gender and Family Survey, including the following three with the above two: "Both the man and woman should contribute to the household income" (reversed) (win); "A job is all right, but what most women really want is a home and children" (wom); "A pre-school child is likely to suffer if his or her mother works" (mw). Responses consisted of 5 categories from "strongly agree" to "strongly disagree", and the values were coded in such a way that a lower score represented a traditional attitude and a higher score a liberal attitude, ranging from 1 to 5. The analysis distinguished four different latent classes or normative types, and Figure 2 shows the mean predicted scores of the five attitude questions among the four groups of latent classes. The figure clearly corroborates the above classification of four normative types created with the two questions (dol and fsfr). Figure 2 also suggests that all four latent classes (or normative types) strongly support women's contribution to the household income, which appears to not necessarily mean women's full-time employment. The four normative types also do not have substantial distinctive mean values regarding what most women really want. In all, the scheme used in this study seems to be more advantageous than the latent class analysis approach because it is easily replicable and conceptually more transparent.
Dependent variables: This study assumes that marital dissolution is an outcome of two conceptually distinguished processes: deterioration of marital quality and formation of a decision to leave the marriage. It intends to examine the effects of gender role norms on these two processes separately. In ISSP, marital status has a separate category of cohabitation, and the cohabiting and married respondents were grouped together in the analysis. To refer to them, the terms "partnered" and "married" are used interchangeably throughout the paper. As a proxy measure of marital quality, we use satisfaction with family life among partnered respondents. As a proxy measure representing the propensity to dissolve one's marriage, we explore attitudes toward divorce among both partnered and unpartnered respondents. Both these measures have shortcomings for fully representing the two components of marital instability. First, satisfaction with family life would reflect not only the marital quality but also the relationships with other family members and the other living conditions of the family. However, in this scale of cross-cultural comparisons among dozens of countries during two time periods, more sophisticated measures are hardly available, and family life satisfaction is an acceptable measure with practical implications. Most of all, across the 44 countries, an extended family living arrangement was rare, with the mean household size range being two to three, and "family" was likely to refer to spouses and young children for the married sample of ages 30-49. Additionally, it is known that marital happiness is highly correlated with satisfaction with other relationships and family conditions (Dush and Taylor 2012;Glass et al. 2016). More importantly, the attitude toward divorce reflects only the general propensity to approve of divorce and does not reflect people's real-life circumstances. This study aims to demonstrate how any predictors of marital instability may have conflicting effects on the two component processes of worsening marital quality and leaving one's marriage.
Soc. Sci. 2022, 11, x FOR PEER REVIEW 7 of 17 Figure 2. Predicted means of five attitude questions among four types of gender role norms: latent class analysis. dol: a man's job is to earn money, and a woman's job is to look after the home and family (5 = strongly disagree); win: both the man and woman should contribute to the household income (5 = strongly agree); wom: a job is all right, but what most women really want is a home and children (5 = strongly disagree); mw: a pre-school child is likely to suffer if his or her mother works (5 = strongly disagree); fsfr: all in all, family life suffers when the woman has a full-time job (5 = strongly disagree).
Dependent variables: This study assumes that marital dissolution is an outcome of two conceptually distinguished processes: deterioration of marital quality and formation of a decision to leave the marriage. It intends to examine the effects of gender role norms on these two processes separately. In ISSP, marital status has a separate category of cohabitation, and the cohabiting and married respondents were grouped together in the analysis. To refer to them, the terms "partnered" and "married" are used interchangeably throughout the paper. As a proxy measure of marital quality, we use satisfaction with family life among partnered respondents. As a proxy measure representing the propensity to dissolve one's marriage, we explore attitudes toward divorce among both partnered and unpartnered respondents. Both these measures have shortcomings for fully representing the two components of marital instability. First, satisfaction with family life would reflect not only the marital quality but also the relationships with other family members and the other living conditions of the family. However, in this scale of crosscultural comparisons among dozens of countries during two time periods, more sophisti- dol: a man's job is to earn money, and a woman's job is to look after the home and family (5 = strongly disagree); win: both the man and woman should contribute to the household income (5 = strongly agree); wom: a job is all right, but what most women really want is a home and children (5 = strongly disagree); mw: a pre-school child is likely to suffer if his or her mother works (5 = strongly disagree); fsfr: all in all, family life suffers when the woman has a full-time job (5 = strongly disagree).
The variable of satisfaction with family life is based on the question "All things considered, how satisfied are you with your family life?" The responses had seven categories: (1) completely satisfied, (2) very satisfied, (3) fairly satisfied, (4) neither satisfied nor dissatisfied, (5) fairly dissatisfied, (6) very dissatisfied and (7) completely dissatisfied. The respondents saying "can't choose" or not answering comprised less than one quarter of one percent and were excluded from the analysis. In the analysis, the score was reversed so that a higher score represented higher satisfaction. The attitude toward divorce was measured by one of several attitudinal questions: "To what extent do you agree or disagree?" The statement given was "Divorce is usually the best solution when a couple can't seem to work out their marriage problems". The answer had five categories from "strongly agree" to "strongly disagree", and again, the scale was reversed in the analysis for more intuitive interpretation. A higher score represents stronger approval of divorce as the best solution.
Relative incomes: The ISSP surveys in both years asked the question "Considering all sources of income, between you and your spouse/partner, who has the higher income?" The seven response categories were (1) my spouse or partner has no income; (2) I have a much higher income; (3) I have a higher income; (4) we have about the same income; (5) my spouse or partner has a higher income; (6) My spouse or partner has a much higher income; and (7) I have no income. Only in the 2012 Taiwan survey was there an additional category of "both have no income", which was recoded into value 4. While the actual income amounts may not be easily comparable across countries due to wide differences in their economic structures, this relative income is more comparable and informative. The scale was grouped into three categories: self has a higher income, about the same income, and spouse has a higher income (omitted category). Among married persons in the United States, an increase in wives' incomes relative to the husbands' increased their happiness but decreased their husbands' well-being (Rogers and DeBoer 2001), suggesting the presence of men's expectations of being the breadwinner in the family.
Control variables: For the analysis, 30 was subtracted from respondents' actual age to make the intercept in the regression equation look reasonable. As the measure of educational attainment, we used years of schooling in the belief that it was the most comparable indicator across the countries due to the varied systems dividing the levels of schools. Table 1 presents descriptive statistics of the individual level variables by year and gender. For the analysis, we created a weight variable which equalized the sample sizes of the countries. The year was another control variable, but one needs to be cautious in comparing the two years because the participating countries for each are not directly comparable. A brief interpretation of Table 1 will follow the introduction of the countrylevel variables.

Country-Level Variables
Types of gender role norms: For each country in each year, we constructed four aggregate variables by calculating the percentages of adults ages 25-54 who held each type of the four gender role norms. Contrary to the gender role norms among individuals, where each respondent belongs to one of the four types, these country-level norms are continuous variables showing the percentages as their scores. For the calculation, we included a somewhat broader range of ages than the sample used for multivariate analysis, assuming people of this broader age spectrum would influence the normative environments in their society. The percentage distributions turned out to be very similar when using either age range. Obviously, the values of the four aggregate variables will be collinear, as they should add up to 100. In terms of zero-order correlation among 68 cases, the traditional norm and the egalitarian norm have the strongest correlation, which is negative, and each of the traditional and male breadwinner norms is negatively correlated with the superwoman or egalitarian norm. However, these correlations alone cannot determine which of them should be included in the multivariate analysis. It is likely that the coefficient for each variable can change depending on other variables included in the equation. Thus, our preliminary analysis tried practically all different combinations (single variable separately, two variables each time and three variables each time) to analyze the dependent variables using the person-level data. We decided to include the percentage scores of MBN and SWN as two variables in the final models. The coefficients were most stable among diverse combinations. In particular, the variable of MBN stood out in its consistently strong effects on the dependent variables.
Crude divorce rate: Data were gathered from various sources, including the United Nations, OCED statistics and national statistics offices of individual countries (OECD 2019;UN 2003UN -2013. In some cases, the crude divorce rates were calculated based on data on the number of divorces and the population, which were collected separately. Summary statistics are included in Table 1.

Descriptive Statistics
In Table 1, the first panel shows the mean scores for the four aggregate normative types, where the first column shows the values for all countries included in the survey. Among the 24 countries that participated in both surveys, the mean percentages were 25, 16, 28 and 31 in 2002 and 23, 21, 22 and 34 in 2012 for the four types of norms, respectively. As expected, the traditional norm decreased somewhat, and the egalitarian norm increased by three percentage points. On the other hand, the male breadwinner norm increased while the superwoman norm decreased. It appears that attitudinal transitions are being made in the direction from traditional to male breadwinner norms, while the superwoman norm is being transformed into the egalitarian norm. MBN represents the lingering male bread winner model in the context of more flexible expectations about women's roles, and its increase suggests that contradictions in expectations about men's roles have increased on average. A decrease in the superwoman norm suggests that contradictions in expectations about women's roles have decreased. The average crude divorce rate in the sample countries was slightly above 2.0 in both years.
The bottom panel of Table 1 presents person-level statistics. The sample sizes show persons of all marital statuses from all participating countries. Ranging from 30 to 49, the mean ages were around 40 for all groups. The mean years of schooling were slightly higher in 2012 than in 2002 for both genders. The distributions of self-reported relative incomes between the spouses changed very little over the two time periods for both genders. In approximately two thirds of the couples, the husband's income was higher than that of their wife. The figures for the two genders were roughly symmetrical and consistent, except that each gender reported a slightly higher percentage for their own income being higher than that of their spouse compared with the reports from the other gender. This may not necessarily indicate a reporting bias because, due to the age differences between the spouses, the couples' average ages, life cycle stages or life experiences as cohorts may have been different for the two gender samples. The proportions of the four types of gender role norms were roughly consistent with the statistics based on country-level data.

Predicting Crude Divorce Rate
The analysis of associations between gender role norms and marital instability starts with country-level data. Based on all 68 cases, the results from the linear regression of crude divorce rate (CDR) on the percentages of different gender role types were rather inconsistent, with the superwoman norm score having a negative coefficient in some models. Then, the R-square value more than doubled to reach 18 percent with 61 cases after excluding four countries-Russia, Latvia, Lithuania and the Czech Republic-which showed outlier values (i.e., relatively high CDRs and traditional gender role attitudes). With 61 cases, the traditional norm score had a strong negative effect and the egalitarian norm score a strong positive effect on the CDR (with a magnitude of 0.03) when they were put in the equation separately (table not presented). The results partially support the prediction that gender egalitarianism will be positively associated with tolerance for divorce. However, as discussed earlier, divorce incidence is an outcome of multiple factors, and the interpretation is tentative.

Satisfaction with Family Life among Married or Partnered Persons
Satisfaction with family life is the indicator of marital quality in this study. Table 2 presents the results from the linear regression models with robust standard errors. When taking into account the correlations among the cases within each survey, robust standard errors tended to be larger than those from the ordinary regressions, while the regression coefficients remained the same. We used STATA to conduct analysis and used the cluster option to estimate the robust standard errors (STATA 2015). Analysis was conducted separately for women and men, and the table shows four models for each gender. Model 1 includes only the personal characteristics, age, educational attainment and type of gender role norms. The next model adds more characteristics, including relative income, and country-level variables of gender role norms. The third model includes the interaction terms between the person-level gender role norms and the country-level male breadwinner norm, with the interaction term contrasting only between the individuals of the egalitarian norm and the rest of the respondents holding any of the three non-egalitarian norms. Our preliminary analysis showed very similar coefficients for the three interaction terms but with somewhat varied standard errors and significance levels; combining them produced stable and significant results. The last model adds the crude divorce rate (CDR). Note that a higher score in the dependent variable means higher satisfaction with family life. With a large number of cases and the proximate determinants of marital quality missing from the equations, the R-square values were not large, being 0.05 in the full model with all the independent variables. Table 2. Robust regression of satisfaction with family life on gender role norms and explanatory variables: married persons.  These findings show important commonalities and some differences between women and men. First, for both gender groups, persons holding egalitarian norms were significantly more satisfied with family life than their counterparts with any non-egalitarian norms. This was true even after controlling for societal gender role norms in the longer models. Second, also common to both genders was the finding that the male breadwinner norm (MBN) at the societal level decreased satisfaction with family life, although it had statistical significance only among men. One unit increase in the MBN percentages would decrease satisfaction by approximately 0.01 point. Third, looking into the interaction effects between societal and individual level gender norms, similarities between genders were more obvious. For both genders, the dampening effect of the MBN at the societal level was true only among persons holding non-egalitarian norms, while relatively high family life satisfaction among individuals holding the egalitarian norm was solid regardless of the normative context at the societal level. It was beyond the scope of this study to examine the exact mechanism of why high scores for the MBN at the societal level decreased family life satisfaction. This study hypothesized that the MBN heightened provider anxiety among many men, and their anxiety would hurt marital relationships. For example, anxious men would be less likely to devote their time to household work and may instead engage in relationship-deteriorating activities to try to restore their masculine self-identity. Among women, it was not their personal MBN but the combination of a personal non-egalitarian norm and high scores of the societal MBN that lowered family life satisfaction, and it is possible that they were affected by their husbands' provider anxiety.
The table also shows important differences between women and men. As said above, women holding MBN at the personal level were no less satisfied with family life than those holding the traditional or superwoman norms, which was contrary to the male case. Among women, the crude divorce rate was significant even after controlling for norm scores, suggesting that crude divorce rates better reflect women's family life satisfaction after controlling for the effects of normative forces.
Not surprisingly, the relative incomes between spouses had opposite but parallel effects on marital satisfaction between women and men. Men's family life satisfaction was higher when they had higher incomes than their spouses, but the association showed the opposite sign. Women who had higher incomes than their spouses showed lower family life satisfaction than their counterparts who had higher-earning husbands. For the husbands, having a higher-earning wife was also a source of dissatisfaction. Furthermore, the effect was no different across the scores of the societal norms. There were few significant interaction effects between the gender role norms and relative incomes (results not shown). Marital satisfaction decreased with age regardless of the controls, possibly due to growing responsibilities with age toward family and work roles. More years of schooling tended to increase family life satisfaction in the shorter models (results not presented), but the significance disappeared after controlling for gender role norms for both gender groups. The level of family life satisfaction did not change over the period.
In conclusion, as predicted, high scores for the male breadwinner norm at the societal level lowered marital quality as assessed by both genders, with an exception for persons with egalitarian attitudes. Although those men personally holding the male breadwinner norm experienced more marital dissatisfaction than other men, it was ultimately the male breadwinner norm at the societal level that deteriorated the marital quality. This suggests that provider anxiety among men can be a root cause of poor relationship quality.

Attitudes toward Divorce as the Best Solution
The results from the linear regression of attitude toward divorce with robust standard errors are presented in Table 3. The findings affirm the relevance of gender role norms, but the effects of gender role norms at the societal level on attitudes toward divorce were generally much weaker than the analogous effects on family life satisfaction. In addition, the directions of the effects toward marital instability were contradictory with each other; that is, for both genders, persons with the egalitarian norm were more approving of divorce as the best solution, although they were significantly more satisfied with family life than their counterparts. Conversely, women and men holding the male breadwinner norm were significantly less approving of divorce than their counterparts with egalitarian attitudes. Gender role norms at the societal level were associated with attitudes toward divorce only indirectly through their interactions with personal norms. For both genders, higher scores of the MBN at the societal level had diverging effects between persons with and without egalitarian attitudes. There was decreasing approval of divorce among those with the egalitarian norm but increasing approval of divorce among those with non-egalitarian attitudes, but neither of these effects were statistically significant by themselves. Such divergence may simply reflect the fact that those two groups had a wider gap in their approval of divorce when the social MBN score was low, and the gap narrowed as the societal MBN score increased. Interestingly, when not controlling for gender-role norms at the societal level in countries with higher crude divorce rates, both women and men were less approving of divorce (results not shown). After controlling for gender role norms, the association disappeared, suggesting that it was not the divorce rate that caused attitudes toward divorce. Without clarifying the mechanisms behind these associations, it is rather difficult to speculate on any causal relationships. Table 3 also shows that divorced people were most strongly approving of divorce compared with people of any other marital status, which could be a consequence of their own experiences completing divorce processes. Older people of both genders were more approving of divorce than their younger counterparts, regardless of the controls. Life experiences may have taught them that divorce can be the best solution after all, or this trend could be due to a higher proportion of remarriages among older people whose attitudes approve of divorce. Education did not have a strong association with attitudes toward divorce, but generally more educated persons, and especially more educated men, were less approving of divorce as the best solution.
One may expect attitudes toward divorce to liberalize over time, but this was not confirmed over this time interval. The coefficients in the models showed the opposite direction, but this was entirely an artifact of the changes in the participating countries. When limiting the analysis to only the countries that participated in both surveys, the coefficient for the year was not significant (results not shown). Lastly, we should note the wording of the question, which questioned divorce as the "best solution". Approval of divorce is possible even when people do not agree that it is the best solution should they approve of people who make the decisions for themselves. Approval may be much higher without the wording of "best solution".
In summary, two findings seem to be conclusive. One is the finding that persons holding the male breadwinner norm were less approving of divorce than the counterparts with the egalitarian norm, and this association was robust regardless of the normative contexts of the society. Second, and on a related note, gender role norms could have conflicting effects on marital instability through the two components of marital quality and propensity to leave a marriage. Both these findings have important implications. Simply speculating without further data, if men holding the MBN personally experience contradictions between their expected and actual role performances, and if their wives feel the contradictions or distress faced by their husbands, then the male attitudes of disapproving of divorce may augment their own frustrations. The implications of the second finding are more theoretical. Any study examining the effects of gender role norms on marital instability should consider these conflicting pathways in order to not erroneously attribute one to the other.

Summary and Conclusions
In recent years, some societies with relatively high divorce rates began to reverse the trend and show signs of the family becoming more stable. Some scholars argue that such a trend is related to the advancement of gender egalitarianism. They suggest that while gender revolutions were stalled (i.e., women became more active in the labor market but men lagged behind in sharing the family roles of housework and childcare), family instability was high. As the gender revolution progressed and individuals and institutions adjusted to women's employment outside their homes, married persons became happier, and marriages became stable. This study investigated how the norms about gender roles are associated with marital instability. The analysis was based on two propositions. The first was a theoretical one based on logical reasoning, stating that marital dissolution is an outcome of two rather distinct processes: deterioration of marital quality and formation of a decision to leave a marriage. The former is considered the underlying cause, and the latter is a procedural issue related to an action carrying out the transition. Second, the antithesis of advanced gender egalitarianism was the lingering male breadwinner norm, which prescribed men to be the sole or primary breadwinners while giving women more flexibility in their roles. In other words, the antithesis of advanced gender egalitarianism is not gender inequality, represented by working women performing second shifts after coming home from work. To test the propositions, we classified the gender role norms into four types and examined their effects separately for marital quality and for propensity to leave a marriage.
The data were from ISSP Gender and Family Surveys conducted in 2002 and 2012, including 33 and 35 countries, respectively, with 24 of them included in both years. The sample of person-level analysis was restricted to ages 30-49, who were supposedly in the life cycle stages of family formation and expansion and thus would feel strong pressures to fulfill both family and work role obligations. Gender roles were classified into four types: traditional norm, lingering male breadwinner norm (referred to as MBN), super woman norm (referred to as SWN) and advanced gender egalitarian norm.
Overall, the findings were unequivocal. The MBN at the societal level had a strong association with marital dissatisfaction, particularly among men. Additionally, men personally holding the MBN showed the lowest marital satisfaction among the four gender role groups, even though the difference was significant only compared with men holding the egalitarian norm. Furthermore, the effects of the MBN on dampening satisfaction at the societal level were visible only among persons with non-egalitarian norms, while persons with the egalitarian norm were not affected by the normative contexts. These results demonstrate the importance of both person-level gender role norms and societal normative contexts. The exact mechanisms of how the norms are associated with marital instability are beyond the scope of this study, but the findings were consistent with the prediction that lingering breadwinner norms would press many men to experience provider anxiety in economies with expanded service sectors. The wives' employment may pose psychological threats to their husbands in the context of the MBN, whereas wives' employment and earnings could bring about a sense of relief and support to their husbands in different environments with more gender egalitarian norms.
Although women holding the superwoman norm showed lower marital satisfaction than their counterparts with the egalitarian norm, high scores for the SWN at the societal level did not lead to any lower marital satisfaction. Instead, the scores for the MBN decreased women's marital satisfaction if they also held non-egalitarian norms. Being expected to perform double shifts may cause women to feel different kinds of pressures than what male breadwinner norms cause for men. To be a super woman or not may be a different issue than the test faced by men to be a real provider for the family or fail to be one. For both women and men, the egalitarian norm at the person level was the strongest and most consistent predictor of marital satisfaction, supporting the thesis of an inverted U-shaped pattern of marital instability over time.
However, persons with the egalitarian norm were also more approving of divorce compared with their counterparts with the male breadwinner norm. This result indicates that gender role norms can have conflicting effects on the two components of marital instability (i.e., marital quality versus propensity to leave a marriage). Studies examining marital instability should consider these conflicting pathways to avoid making misleading conclusions. Some of the associations discussed in this paper are likely reflections of some macro-level institutional arrangements, such as the welfare systems in Scandinavian countries. A supportive welfare system may boost divorce rates and promote gender egalitarian norms and social arrangements, even when persons do not necessarily agree that divorce is the best solution.
The data from 68 national surveys provided valuable information for examining gender role norms both at the societal and personal levels. A broad implication is that during a society's process of norm transitions, lingering male breadwinner norms in the society can be detrimental to marital stability. However, empirical linkages that can explain the associations are broadly missing in this study. Most critically, this study assumed an underlying historical transition from traditional to egalitarian norms, even though no temporal order was assumed between the two transitional types of the MBN and SWN. Future research should more carefully examine the institutional settings of the countries studied to elaborate this study's propositions.