Gender Ideology in Europe: Plotting Normative Types in a Multidimensional Space

Two recent articles (Grunow et al. in J Marriage Fam 80(1):42–60, 2018; Knight and Brinton in Am J Sociol 122(5):1485–1532) suggest that gender ideology is multidimensional. Such a finding is of utmost importance because, if robust, it can be used in future comparative (multilevel) research. However, these two articles present different results on which gender ideology profiles are dominant in Europe. Using the European Value Study, we replicate and extend these Latent Class analyses and address the question of generalizability and content- and criterion-related validity. We come to a five-cluster solution that not only synthesizes theoretically and empirically the results of the two articles, but also contributes to the literature by clarifying the place of these five gender ideology types in a multidimensional space. We suggest that in Europe five ‘worlds of norms’ exist that are mutually influenced by the general institutional context of welfare states.


Introduction
Two recent articles (Grunow et al., 2018;Knight & Brinton, 2017), hereafter GBB and K&B respectively) suggest that gender ideology in Europe is a multidimensional concept, rather than a continuum that varies from traditional to egalitarian. This finding is at odds with previous considerations that dominated the field. Specifically, previously, many scholars were seeing gender ideology as a continuum varying between traditionalism and egalitarianism and assumed that gender role attitudinal change was taking place due to cohort replacement and that this would lead to a cultural convergence across countries.
Thus, modernization would bring systematic and rather predictable changes in gender roles (Inglehart & Norris, 2003;Sullivan et al., 2018), directed towards more liberalism or egalitarianism as the more conservative cohorts are gradually replaced by more egalitarian ones.
The work of GBB and K&B stems from another strand of literature which questions the coherence of the concept of egalitarianism (Cotter et al., 2011;Verloo, 2007) and shows that the gender revolution has stalled (England, 2010;Goldscheider et al., 2015). According to this literature, gender inequality is reproduced not only by patriarchal norms but also by essentialist attitudes. Such essentialist attitudes imply the conviction that men and women have fundamentally different traits: men are 'made' to focus on paid work and women are 'made' to take care of children and carry out various types of unpaid work (Ridgeway & Correll, 2004). According to these attitudes, although female employment is not rejected, women are still considered to be better suited for caring for children. Evidence for this consideration is provided by the fact that Eastern-European countries do not fit into the traditionalism-egalitarianism continuum. In these countries, one can observe a clear discrepancy between the high rates of female labour market participation and the rather conservative attitudes of women when it comes to care-giving roles (Kalmijn, 2003). This implies an important 'second shift' (Hochschild and Machung 1989) for women in these societies and calls for the inclusion of a new dimension: the dimension of essentialism (see explanation below). As this dimension exists next to the dimension of 'egalitarianismtraditionalism', the analysis of gender ideologies needs to allow for multiple dimensions (for Europe: Grunow et al., 2018;Knight & Brinton, 2017; for the US: Scarborough et al., 2019).
The finding that gender ideologies are multidimensional is also important for comparative studies in the field. This implies that measures of one-dimensional egalitarianism-traditionalism need to be replaced by measures of different gender ideology types as macroindicators of the normative context. More specifically, one could think of using the average posterior probabilities of the latent class results (i.e. gender ideology types) to assess the extent of the presence of certain gender ideology type(s) in a country. Alternatively, one could think of creating clusters of countries on gender ideology, as has been done in the past for welfare regimes (Esping-Andersen, 1990, 1999. Such macro-indicators or country classification could then be used to predict micro-level associations. Despite the ground breaking nature of the research of K&B and GBB, we still lack a clear picture on gender ideology types in Europe. The reasons for this are both methodological and substantive. From a methodological point of view, although both articles use Latent Class Analysis, they reach somehow different results. Ideally, one would expect that, despite the slightly different approaches, the two studies would identify similar gender ideology profiles: the dataset used (EVS) is the same, the concept of gender ideology is roughly similar, and the theoretical considerations, based on adding the essentialism dimension to the separate/joint sphere dimensions (initially defined as the 'egalitarianismtraditionalism' dimension), are also to some extent equivalent. Yet, the two studies come to partly different gender ideology profiles: a four-class solution was found by K&B, 1 whereas GBB come to a five-class solution. 2 Substantively, the results of both articles are difficult to interpret; i.e. the gender ideology types that are found are not placed in a multidimensional space of joint/separate spheres and the degree of essentialism.
The aim of this paper is to provide a generalizable and valid result on the types of gender ideology in Europe. With 'valid' we refer to both content validity, based on the definition provided by Davis and Greenstein (2009) on gender ideology, and criterion-related validity. To achieve this, we first extend the analysis of GBB and K&B and examine the sensitivity of our results to the use of different (more) countries and items that measure gender ideology (i.e. we provide an empirical generalization and conceptual extension to K&B and GBB (Tsang & Kai-Man, 1999: p. 766)). Second, and most importantly, we provide a substantive explanation of the gender ideology types by discussing the position of these types in a three-dimensional space that is defined by (1) an axis of joint versus male 'ladder' privilege 3 in the public sphere, i.e. the extent to which women should contribute to paid work (the x-dimension); (2) joint versus female 'web' virtue in the private sphere (the y-dimension), i.e. the extent to which men should contribute to unpaid work; (3) and finally an axis of essentialism, i.e. to what extent do people belief that gender roles are based on gendered traits or are individual choices (the z-dimension).
The paper is organized as follows. In Sect. 2, we discuss the similarities and differences in the conceptualizations of gender ideologies by K&B and GBB, and we theorize which classes we expect to find ourselves. In Sect. 3, we discuss the European Values Study dataset and we report on the method and operationalization used. We also replicated both K&B and GBB's article (see Appendix for the replication study). Subsequently, we extend the analyses of K&B and GBB to be able to generalize towards the EU and synthesize their theoretical and empirical approaches into a new, integrated, classification of gender ideology types. In Sect. 4, we summarize and discuss our findings, whereas in Sect. 5 we give examples on how to apply the found gender ideology types in future research. Finally, Sect. 6 concludes this article.

Theoretical Similarities and Differences in the Conceptualization of Gender Ideology Profiles by K&B and GBB
In this section, we elaborate on the theoretical similarities and differences of the two seminal articles on gender ideology classes in Europe. We are aware of the fact that the two articles had a different focus, with K&B emphasizing trends in gender ideologies, whereas GBB focused on the centrality of work and children in peoples' lives and therefore used a young analytical sample. Nevertheless, there are some theoretical differences in the two articles that are not related to the different foci. Table 1a, b illustrate the comparison of the theoretical basis of the conceptualization of gender ideology used by K&B and GBB. Table 1b was taken directly from the GBB article, while we derived and classified the gender ideology concept of K&B into Table 1a ourselves. Following GBB, we categorise beliefs on gender ideology in three aspects: the 'separate sphere' aspect in work and care and the 'choice/trait' (or essentialism) aspect. The separate sphere aspect refers to whether people believe that a certain activity should be performed by one of the two or by both genders. Specifically, GBB specify these activities to caring (the private sphere) and earning (the public sphere). In caring, the issue is whether people believe that it should be carried out by women only or by both women and men, while in earning the issue is whether people consider it a task for men only or for both men and women. The choice/trait aspect refers to whether people believe that gender roles are a result of gendered traits or choice.
Comparing Table 1a and b indicates the similarities and differences in the conceptualization of gender ideology between the two articles. The two opposite groups of traditionalism and egalitarianism are defined similarly by the two articles: traditionalism is defined as men having the primary role in earnings and women having the primary role in household work and this is due to a belief in trait differences between genders. Egalitarianism covers a non-essentialist and egalitarian view where both male and female partners are contributing to paid and unpaid work based on choice. In contrast, the two articles differ in the other two groups that they expect: K&B distinguish between two other 'egalitarian' types: 'egalitarian familism' and 'flexible egalitarianism', whereas GBB add the groups 'egalitarian essentialism' and 'intensive parenting'. We will first discuss the definition and the use of the term 'essentialism' in both articles, since the differences in the latent classes between the two articles are largely due to the different way in which each of the two articles combines the concept of essentialism with the concept of egalitarianism. In both articles, essentialism is defined purely as the extent to which different tasks or roles for men and women in paid work and care-taking are considered a gendered trait or preferences. In more detail, they rely on the definitions of Ridgeway and Corell (2004) and Charles and Grusky (2004). Ridgeway and Correll (2004: p. 527) describe essentialism as "Descriptive beliefs about the attributes of the 'typical' man or woman that are still largely shared and largely unchanged since the 1970s", while Charles and Grusky (2004: p. 15), in the context of occupational segregation, describe it as "Although prevailing characterizations of male and female traits are complex and multifaceted, a core feature of such characterizations is that women are presumed to excel in personal service, nurturance, and interpersonal interaction, and men are presumed to excel in interaction with things (rather than people) and in strenuous or physical labour. These stereotypes about natural male and female characteristics are disseminated and perpetuated through popular culture and media, through social interaction in which significant others […] support such interpretations, and through Liberal egalitarian micro-level cognitive processes […]." Thus, gender essentialism represents "[…] women as more competent than men in service, nurturance, and social interaction."; and that " […] Although biological differences between the sexes […] may have contributed to the initial development of these principles, they have subsequently become ideologically and institutionally entrenched and have accordingly taken on lives of their own." We read this as the importance of considering 'gendered traits' (whether biologically founded or being socialized) as an additional dimension to the 'traditional-egalitarianism' dimensions (i.e. the separate or joint spheres dimensions in the caring and earning sphere). Hence, based on this concept of essentialism, K&B define "egalitarian familism" as the combination of two dimensions: egalitarianism in the earnings sphere and essentialism in the caring sphere. This means that both spouses should contribute to household income but only women have a 'second shift' of being the caretaker at home. In addition, in their analysis, K&B expected to find a latent class that they name "flexible egalitarianism". Their interpretation of this latent class is that it includes people for whom individual choice prevails and for whom women are not obliged to take on any specific role. Female employment is supported, but not considered necessary, just as their care-taking role is not considered fundamental for their identity. According to K&B, flexible egalitarianists believe that any combination in the paid and unpaid work sphere is acceptable.
According to our understanding, for GBB the concept of "egalitarian essentialism" is close to the concept of flexible egalitarianism that is introduced by K&B. GBB state that they build upon Charles and Grusky (2004), among others, but most importantly they broaden the concept of "essentialism" following Cotter et al. (2011) and combine it with egalitarianism to form a multi-dimensional concept. Specifically, they describe egalitarian essentialism as "the deeply rooted and widely shared belief that gender differences are ideologically compatible with liberal egalitarian norms" (a quote from Charles & Grusky, 2004: p. 27) and state that, in this dimension, both the work sphere and care sphere are of equal value and that this type of people ("egalitarian essentialists" according to GBB) think beyond gendered traits and hierarchical power relations, thereby denying implications of power in whatever division of labour constellation. They mention that, for instance, the choice of being a housewife can just be as egalitarian as working for pay. Cotter et al. (2011) used for these types of people the term 'egalitarian essentialism' as "combining support for stay-at-home mothering with a continued feminist rhetoric of choice and equality (Stone (2007)", (Cotter et al, 2011, p. 261). To our knowledge, the combination of 'egalitarianism' and 'essentialism' in the concept 'egalitarian essentialism' has only been used as such, by Cotter et al (2011) and reused by GBB (and later on by Scarborough et al., 2019). Later on in their article, on p. 267, Cotter et al., 2011 refer again to 'egalitarian essentialism', and state that this new cultural frame "[…] combines elements of the previous conflicting frames of feminism and traditional familism and thus provided support for a return to traditional gender roles while denying any implications of lower status or power for women". This concept, according to us, thus, has more overlap with the flexible egalitarian concept of K&B than with that of egalitarian familism (which, in our view combines 'egalitarianism' with 'essentialism').
On top of this, GBB define one other gender ideology type, namely that of "intensive parenting". This type of people prioritizes-in the case when they have children-child care above the paid work sphere. Women are primarily responsible for children since motherhood is seen as a natural feminine trait. Moreover, children and children's development in particular are considered "sacred" and therefore fathers are also expected to step in into child care, especially when it comes to cognitive development of their children in an increasingly more competitive society (Hays, 1996). GBB do not clarify, however, how the 1 3 economic sphere should be divided between male and female spouses. In the theoretical part of their article, they state that the earnings sphere is encouraged to be separately male, but in their empirical part they find that the "intensive parenting" class respondents support the earnings sphere to be shared between male and female spouses.
Post-hoc, GBB also find a fifth class, which is the moderate traditional class. This class mainly is occurring in the Netherlands and differs from the other classes in that this type of people has low confidence in the child care taking skills of fathers.

Conceptualizing Gender Ideologies: Our Expectations on 'Five Worlds of Norms'
During life, people develop beliefs about the roles and expectations that are linked to each (binary) gender. 4 Leaving biological characteristics and capabilities aside, the process of gender socialization may entail reinforcement of gender roles by parents and peers on the one hand, and modelling, i.e., copying another's behaviour, on the other hand. Apart from social learning, people also have a more active role during the life course. As they grow older, individuals may develop changes in their gender role beliefs, which is guided nevertheless by the gender schema's they have internalized to start with (Stockard, 2006). Societal gender ideologies are "sets of widely taken-for-granted cultural beliefs about the essential natures and relative worth of men and women" (Chatillon, et al., 2018;p. 217). They are influenced and built into social institutions and traditions of socio-historical contexts, thereby supporting the division of labour between heterosexual spouses. In this article, we will take a 'snapshot' of such socio-historical European contexts (the year 2008), and even though we may not be able to assess time trends in gender ideologies, this snapshot will enable us to give a more accurate and generalizable picture of the EU welfare states in the last one and a half decade.
We expect to find five types of gender ideologies: 1. Traditionalism; 2. Transitionalism; 3. Sameness feminism; 4. Difference feminism; 5. Third wave feminism. To understand how we arrived at these 'five worlds of norms' that we claim to exist in the EU, we need to look into the historical developments of the different European contexts. Most importantly, many scholars point to a (in some contexts 'near complete') development of the 'first half' of the gender revolution, i.e. women's accepted participation in the public sphere. Women's (economic) lives have changed quite dramatically the last half a century (England, 2010;Goldscheider et al., 2015). More and more EU societies are confronted with the (slow) disappearance of the male breadwinner model and an emergence of the dual earner (or oneand-a-half-earner) model. To avoid cognitive dissonance in their behaviour, people will adjust their attitudes accordingly -here we follow the structural approach of Goldscheider et al., 2015 rather than for instance the Theory of the Second Demographic Transition or related post-modernism theories, such as that of Inglehart (1997) and we assume that structure is mostly affecting culture rather than the other way around, although of course we may expect mutually reinforcing influences (Pfau-Effinger, 2005). This adjustment of attitudes may be a 'true' reflection of peoples changing beliefs about gender roles or a weaker version of that, namely socially desirable answers on questions about gender equality, especially in more gender equal contexts. Whichever the case, we expect to find a weakened traditional class compared to the one we used to know. The traditional class had people fully supporting the separate sphere division of labour (men participating in the public sphere and women in the private sphere), based on gender essentialist beliefs that men are fundamentally different than women and 'made' for a different participation in society with respect to gender roles. Instead, we expect to find a new class, 'the transitional' class, that includes people who provide less extremely conservative answers in questions that imply a separate sphere for men and women, in particular when it comes to women's participation in the earnings' sphere. In fact, this class will be difficult to be distinguished anymore from the 'egalitarian' classes when it comes to economic aspects (women's economic independence and their equal contribution to the household income), but we hypothesize that other aspects (and more specifically on the combination of work and care for women) will identify 'transitional' membership. We expect that this class will especially be 'transitional' in the role for women in society. As said, people from this class would acknowledge women's newly gained position in the work sphere, but they would also see the negative consequences of this for the family as expressed in the costs of women who are working for children. Even though this class does not find it an absolute necessity to have children for women's identity, once women have children, this class identifies work-family life balance problems. Finally, we refer to this class as 'transitional' because we could see this class as suffering from confusion of what is the norm in a changing society on the road between traditional towards more egalitarian elements. As Esping-Andersen and Billari (2015) point out in their multiple equilibrium framework: "A once-stable family equilibrium gave way to a prolonged period of uncertainty and normative confusion as to what constitutes proper gender roles and identities in family life". This type of people also realises the difficulty in combining work and care when policies and other institutions (and men) do not step in in a phase of unstable equilibria (or normative confusion). Esping-Andersen and Billari refer to changing women's roles while institutions and partnerships have not yet adapted, as an important reason to explain lower childbearing rates in Europe. Hence the label 'transitional'.
Furthermore, in line with K&B, we expect to find three types of 'egalitarian' classes. Based on their adherence to different meanings of gender equality and the overlap with the 'type of feminism' that one is supporting, we label these three types differently than K&B. First of all, what does 'egalitarian' or 'gender equality' mean? According to Verloo (2007), people can have three different visions of gender equality: (1) gender equality can be seen as a problem of sameness (i.e. women do not have equal opportunities); (2) gender equality can be seen as affirming differences from the male norm (women and men are different, each have their values, and what is female should not be subordinate to the male norm); and (3) gender equality can be seen as transforming all established norms and standards of what is/should be female and male (the strategy of gender mainstreaming, or going beyond gender categories). One can link each of these three visions to the discourse between different streams and waves of feminism. Note that our discussion will remain restricted to the European and US (being highly influential for Europe) discourse of feminism as this articles' focus is on the EU. Note also that individuals do not have to identify themselves as being an 'active feminist', they may even reject being named a 'feminist' when asked. However, in line with Aronson (2003), we expect to find many individuals who passively support feminism and who would, if asked, not express antifeminist sentiments. 'Feminism' as a term has been used in many ways and received at some point a negative connotation for many people. This is reflected in a popular stereotype being 'I do not want to hate men', 'feminism is equal to lesbianism', or 'feminism has gone too far' (Aronson, 2003). However, we want to stress here that, even though people (including men) have not thought about their position towards 'feminism' or more specifically their ideology towards gender equality, we expect that most individuals do respond to certain items or questions about gender equality in a way that might be coherent with one of the three classes of 'feminism' that we point out here. We thus use an inclusive definition of feminism, following Sylvia Walby (2011): "feminism encompasses those individuals that advance women's interests but do not claim an explicitly feminist affiliation." This definition, in other words, thus includes all visions about increasing gender equality or women's interests' advancement, including those of men and women who might not self-identify as feminist. After all, many feminist ideas have become part of the common sense of European culture, even though many of those ideas may not be recognised (anymore) as being feminist essentially (Whelehan, 1995). Of course, there is even more diversity in the field of feminism than the three approaches that we will lay out here. But we aim to get at core differences between types of people, thereby trying to capture three main classes that are based on mass-ideas and actions at the expense of blurring within-class or hybrid-class heterogeneity. In our discussion of feminism, we will focus first on the sameness / difference debate in second wave feminism and then we will discuss an alternative approach that is used by third wave feminists. In the sameness / difference debate, one could say that the first perspective is endorsed by 'sameness feminists' (or by K&B a.k.a 'liberal egalitarianists')-stemming from the original liberal feminism thoughts of white middle-class heterosexual womenwho emphasize the similarities between men and women, and therefore women should be treated just like men. Building upon male-oriented liberalism, male primacy is what should be fought against according to people who believe in this perspective (Whelehan, 1995). Hence, these types of people would endorse a gender role ideology where tasks are equally shared between men and women, both in the public and the private sphere. The 'difference' viewpoint, on the other hand, argues against these 'sameness' arguments in that these people believe that women and men are fundamentally different. Moreover, they believe that women have a structural disadvantage in living up to male norms. For instance, women may experience structural disadvantages in the workplace and might need 'extra help' or 'special treatment' to perform as 'ideal workers' (e.g., by means of maternity benefits and public child care). Women may make different choices in their work-life balance, simply because they are not the same as men are. 'Difference feminists' or (by K&B a.k.a. 'egalitarian familists') are aware of the male bias of our social institutions and the difficulties a woman has in being both a mother and an 'ideal worker'. Hence, these feminists are aware of differences in male and female norms, but they note that the male (e.g. hierarchical, rational, independence, and self-control) norm is the standard, whereas female norms are considered subordinate. Most importantly, they would strive for remaining differences between men and women, but changing the superiority of the male norm to an equal norm compared to that of the female norm (Capps, 1996;Williams, 1991). These differences between men and women would culminate in an individual belief that women would only identify themselves as women in their mother role, even though they are equally expected to play a role in the work sphere. In other words, children (and housework) are key elements in women's lives and create/confirm/reinforce women's identity. However, in the view of difference feminists, accentuating unpaid work for women would not mean a subordinate position for women as these are just different norms (not male, but female norms) that each have an equal stand. These types of feminists, thus are egalitarian with respect to the work sphere (joint participation in earnings), but they are quite essentialist (rather than considering individual choice) when it comes to the private sphere, believing in gendered traits, i.e. women and men are fundamentally different from one another and this corresponds to separate tasks in the private sphere. Finally, 'third wave feminists' (a.k.a. 'flexible egalitarianists' by K&B) would consider gender equality as a problem of going beyond the current existing norms and developing gender neutral norms. Note that, even though the term 'third wave' may suggest that feminists adhering to this strand of feminism position themselves 'after' those of 'second wave feminism', we support the view of Mann and Huffman (2005) and position the 'third wave feminists' 'against' the thinkers of the 'second wave', rather than chronologically afterwards. Driven by an intersectionality perspective initially, 'third wave feminists' rebelled against the 'second wave feminist' mantra that the personal is political (including that personal lifestyle choices should not undermine feminist politics) of the white middle class who still had to fight for equal acceptance in the labour market. 'Third wave feminist believers' came to rise in the early nineties, unitedly disqualifying several generalized, be it sometimes misleading, connotations of 'second wave feminism', such as viewing women as victims and their white middle-class view on women as all belonging to the same category (i.e. 'sisterhood') (Mann & Huffman, 2005;Whelehan, 1995). 'Third wave feminists' strive to deconstruct all group categories of women that are made up of people with heterogeneous experiences. Rather than seeing affirmations of identity as liberating, 'third wave feminists' observe them as disciplinary, restrictive, and regulatory. Identities are multiple, fluid, and unstable instead and there is aversion towards all forms of categorization and binary thinking. This 'new generation' embraces 'hybridity' and 'messiness'. As Heywood and Drake (1997: 8) state it: "The lived messiness characteristic of the third wave is what defines it: girls who want to be boys, boys who want to be girls, boys and girls who insist they are both, whites who want to be black, blacks who want to or refuse to be white, people who are white and black, gay and straight, masculine and feminine, or who are finding ways to be and name none of the above". Whether one supports the 'new empowerment' of this 'power feminism' or not, one thing the 'third wave feminists' have in common, which is the emphasis on 'individualism' or individual choice. As Karp and Stoller (1999: 310-311) in their popular chronicle of girl culture write: "We've entered an era of DIY feminism-sistah, do-it-yourself-and we have all kinds of names for ourselves: lipstick lesbians, do-me feminists, even postfeminism. … No matter what the flava is, we're still feminists. Your feminism is what you want it to be and what you make of it. Define your agenda." This reverses the second wave's mantra that the personal is political as the political becomes totally personal. Hence, 'third wave feminists' wouldn't consider the consequences of their behaviour whenever it is done by individual choice because they look beyond male primacy, binary identity categories, and gendered traits. Whelehan (1995: 241) critically observes in her conclusion that the idea of the 'new generation feminists' has been legitimized by a conservative backlash, announcing "a return to the sanctity of the 'private', and emphasis upon the family and childcare". 'Anything goes', as long as one feels empowered individually (even though being in a 'subordinateposition'). Thus, we derive that for 'third wave feminists' any division of labour between men and women would be supported; it is just the attached male and female (subordinating) norms that go together with it which define that one decision would be superior or subordinate towards another decision. In contrast to 'sameness feminists', 'third wave feminists' rebel towards a collective category of women fighting to be just like men by living their personal lives in a way that consider political consequences of their gender role behaviour, thereby considering internalized rather than external oppression. In contrast to 'difference feminists', 'third wave feminists' undermine inequality by rejecting both hegemonic masculinity and hegemonic femininity categories and in this way being 'in power' or 'emancipated'. They reject that men possess power and women are subject to and/or lack power and instead conceive power as relational, dynamic, having multiple tactics and strategies, and being available to subordinate groups and not just the possession of dominant groups (Schippers & Sapp, 2012). Note that here, again, we simplified 'within-wave' heterogeneity and included visions that some refer to as 'post feminism' rather than as 'third wave feminism', for instance when we emphasize femininity as a matter of individual choice and a set of practices from which women can pick and choose. Post feminism states that power is based on individual feelings of empowerment, thereby ignoring structural power relations (Schippers & Sapp, 2012); women are seen to be in power if they say they are, no matter their individual choices.

Validating the Gender Ideology Profiles of K&B and GBB: An Empirical Extension and Synthesis of the Two Articles
In this section, we touch upon the question of external and internal validity with respect to gender ideology types. Are the results of K&B and GBB robust to a slightly different context and different operationalization of the same theoretical construct? After having replicated their studies (see Appendix) to exclude secondary observer effects-which occurs when the researchers themselves are the source of variability in results (Breznau, 2016), we first empirically extend their LC analyses and then we interpret our results and come to a theoretical synthesis of the two articles, as well as a positioning of the gender ideology types in a three-dimensional space.

Data
In our empirical extension, we use the EVS for the year 2008. The EVS is a large-scale cross-national comparative survey on values and attitudes. For each country, a representative stratified random sample was drawn from the adult population. In wave 4, the net sample size for each country is 1500 respondents. 5 The EVS is the dataset that offers the most extensive measurement of gender ideology. Since it covers almost all EU countries, it is the most evident candidate for use by multi-level studies that investigate the impact of the normative context on micro-level outcomes. 6 Using data for 2008 only, while we disregard trends over the different EVS waves, we can analyse 24 countries. 7 Hence, we include seven additional countries that were not analysed by K&B, and 16 countries that were not analysed by GBB. The countries included in K&B and GBB are mentioned in the Appendix, Table A1, in which the discrepancy in country selection between the two studies emerges as well as the availability of countries in the EVS2008. By including more countries than K&B and GBB, we improve the external validity of the analysis as we have a larger and more representative sample of European countries. Note that the percentage of Eastern-European countries in our sample (42%) is larger than that of K&B (35%). More importantly, we included countries that are often overlooked in many studies due to historical contingency and path dependency problems (Ebbinghaus, 2005). 8 Because we want to generalize towards the population of the European Union, we use population size weights (see Appendix, Table A2). Using such weights further increases external validity of the analyses compared to K&B-who used equal-size weights-(but note that we also performed a sensitivity analysis using equal-size weights, which lead to similar findings as using population-size weights). 9 We select respondents aged 20+.
Missing values were between 3 and 9% of the sample (depending on the item). We treat these missing values with full information maximum likelihood estimation assuming implicitly that they are Missing At Random (MAR).

Operationalization
In the literature, the seminal article of Davis and Greenstein (2009: p. 89) defines gender ideology (equivalent to gender attitudes or beliefs about gender) as "[…] an individual's level of support for a division of paid work and family responsibilities that is based on the notion of separate spheres". The authors distinguish six different categories that capture beliefs about relationships between men and women: 1. Primary breadwinner role 2. Belief in gendered separate spheres 3. Working women and relationship quality 4. Motherhood and the feminine self 5. Household utility 6. Acceptance of male privilege The EVS covers only the first five categories with 14 items (see Table 2). 10 This means that the concept of gender ideology was never measured to its full extent in the European Values Study (EVS) and that certain aspects of it were not taken into consideration. Table 2 presents the availability of the various items that measure gender ideology according to the definition of Davis and Greenstein (2009) in EVS that were actually used by K&B and GBB. K&B and GBB selected seven of these items-albeit somehow different -covering three or four of the dimensions of gender ideology defined by Davis and Greenstein (2009). In both articles, some theoretical discussion about the concept of gender ideology is present, but none of the two discusses exactly why they discarded some of the items. 11

Table 2
Discrepancy in item selection between the two studies of K&B and GBB and the availability of items in the European Values Study EVS2008; Numbers in the first column refer to the dimensions that were defined by Davis and Greenstein (2009) In most questions, answers involved a scale varying from agree strongly (1) to disagree strongly (4). Two items were coded differently: for the item "When jobs are scarce, men have more right to a job than women" respondents could score "agree", "disagree", or "neither", while for the item "Do you think that a woman has to have children in order to be fulfilled or is this not necessary", the answers were "needs children" or "not necessary".
Also in our extension, we acknowledge that we let data availability prevail over contentvalidity. We do this deliberately because we want to enable scholars to use our results in future research and we therefore build upon already highly cross-national comparable available data. Note that we couldn't use the most recent wave of EVS (the 2017 wave) because certain important discriminating items were not asked anymore in that wave. The eight items that we initially used cover both men's and women's perspectives towards the division of labour within the household (see the seven bold items in Table 2 plus "responsibility"). We considered using the entire range of 14 items. However, due to the large number of dependencies between these items, we decided to focus on eight of them. Latent class analysis is not well suited for a data-driven selection of items when there are items that are not highly correlated with the rest because it generates additional local dependencies between the highly correlated items. We therefore chose the eight items that are asked in the same question of the EVS questionnaire (q48) that were intended to measure gender role values, have similar answer categories, and are not influenced by a possible "priming" effect of other items. Note that some items measure a similar concept identified by Davis and Greenstein (2009). Having more items measuring the same concept makes the measurement more reliable (Carmines & Zeller, 1979), but items measuring the same concept should not be too highly correlated (see the discussion on Bivariate Residuals below). Note also that we did not include items that were capturing categories 7 and 8 that we added ourselves to the Davis and Greenstein (2009) conceptualization (and we consider to be capturing less gender role divisions in couples), nor did we consider the item of the importance of sharing chores in marriages which is capturing household utility. We did this because these questions were asked in different parts of the questionnaire and contained different answering categories. In more detail, the item "Do you think a woman has to have children in order to be fulfilled?" is asked before the question battery on gender ideology (q48), but after having asked "what makes a successful marriage", where having children is one possible answer (q42) and whether one agrees or disagrees with the statement about "a child needs a home with both a father and a mother to grow up happily" (q43). These two questions might "prime" respondents to answer about fulfilment by having children either more often with agreement because respondents may take into account the fact that children may be a necessity for marriage to work or more often with disagreement because respondents may take into account their opinion about both a father and a mother need to be present for a child to grow up happily. Hence, having a child may not be wanted against any price. Moreover, the item "When jobs are scarce, men have more right to a job than women" is not a part of the question battery on gender ideology and has a dichotomous answer (agree/disagree). What is more, it follows the question "When jobs are scarce, employers should give priority to [NATIONALITY] people over immigrants", which may implicitly raise discrimination concerns among respondents.
Although we prefer to theoretically deduce why we need to include certain items and exclude others, we did perform a sensitivity analysis with all the 14 items that we initially considered. One can see that the solution is rather complex (assumptions of local independence are violated for the five-class solution, see Appendix Table A3) and the model fit is worse than our final solution. We also analysed 11 items covering five Davis and Greenstein (2009) dimensions, and also this analysis didn't improve the model fit, nor did it give a simpler interpretable solution with few local dependencies to be added to the five-cluster solution (see Appendix Table A4).

Method
Both K&B and GBB, as well as our extension, use Latent Class Analysis (LCA) to estimate gender ideology profiles. LCA is a highly appropriate method for this enterprise as it uses several categorical indicators of the phenomenon of interest as input to estimate the prevalence of certain (latent) groups in the population as output. The aim of LCA is to determine the smallest number of classes that is sufficient to account for the associations observed between the indicators (items). The assumption of local independence is very important in this aspect and entails that the items are mutually independent of one another conditional upon latent class membership. One can distinguish between homogeneous models (models without any covariates) and heterogenous models, in which the impact of covariates such as country and sex are allowed. Adding intercepts of country/sex/any other covariate in the homogeneous model gives a partially homogeneous model. See Fig. A1 in the Appendix for the path diagrams of (partially) homogeneous and heterogeneous models.
The corresponding log-linear specification for the fully heterogeneous model is the following (see Kankaras et al., 2010): where A-G are the M indicators, X are the latent classes, and Z is the group (country and sex) variable. For the partially homogeneous model (i.e. measurement invariance), the following restrictions apply: In other words, in such a partially homogeneous model, there are no interactions between the group variable Z and the latent variable X. Therefore, the relationships between the latent variable X and the indicators A-G are identical across groups Z. This makes it possible to compare the latent class memberships across the groups Z.
Both K&B and GBB estimated several forms of models and their final latent class solution concerns a partially homogeneous model. In K&B, this is a model with country-wavespecific intercepts, while in GBB a model with sex and country specific intercepts.
In line with K&B and GBB, we estimated a range of heterogeneous models (both country-and sex-intercepts and -slopes may differ), partially homogeneous models (only country-and sex-specific intercepts), and fully homogeneous models. We do this because we believe that there are important country differences in class membership and between men and women (as GBB have shown). The fact that the model fit of the partially homogeneous model is close to that of the fully heterogeneous model makes that we can compare the class memberships between men and women of different countries. The results of the BIC's of the different models of these analyses with 24 countries and eight items are presented in Table 3. The homogeneous as well as heterogeneous and partially homogeneous model point to a five-class solution. Based on the BIC (comparing with a four-class and six-class partially homogeneous model, models available upon request)), as well as on meaningful distinctions between the classes, we choose the five-class solution. A necessary next step, however, is to inspect our solution for local dependencies between the items in our latent class solution, which is something that neither K&B nor GBB reported. We do so by means of the Bivariate Residuals (BVR's). The size of the BVR is an indication of how well the observed association between two items is reproduced. Each BVR is a Pearson Chi-square statistic (divided by the degrees of freedom) in which the observed frequencies are compared to the expected frequencies in a two-bytwo table under the corresponding LC model (Oberski et al., 2013;Vermunt & Magidson, 2004). Α BVR larger than 1 indicates a strong association between the two relevant items that violates the local independence assumption. The BVR's of the eight-items model (see Table A5) suggest that further steps need to be taken to reduce local dependencies.
Several strategies for reducing local dependencies have been proposed (Vermunt & Magidson, 2004). First, we can directly model such a dependency in the LC model. Second, we can delete one or more items involved in these high local dependencies. Third, we can increase the number of latent variables in the model by introducing a multi-factor model. In Table 3 we show the BIC results for all of these possible adjustments to the five-class model. The first part of this Table presents the LC models that allow for bivariate relationships. These models show that allowing for such relationships is not a viable way to go since reducing the BVR's requires the inclusion of too many bivariate relationships, which could create a problem of identifiability of the LC model. Including one bivariate relationship at the time, moreover, doesn't lead to a better model fit, according to the BIC value. The second option is to delete one or more items. When looking at the items that introduce high dependencies, we see that one of them is the item "responsibility". We decided to delete this indicator because it is highly correlated with fatherssuited" and, more importantly, because we think that this item could be interpreted from both a time and an economic perspective; men taking responsibility in economic terms measures a more traditional rather than an egalitarian approach, implying a male breadwinner household. Hence, it is unclear whether this item captures more traditional or more egalitarian values. The seven-item model leads to a five-class solution as well, but with substantially lower BVR's. We also tried several multi-factor models, but their model fit (see Table 3) was worse than the five-class model with one latent variable. Moreover, these models are more difficult to interpret and less parsimonious.
Hence, we finally decided to continue with the five-class partially homogeneous model (with country-and sex-intercepts) based on seven items. We rechecked the five-class decision by comparing this model to the four-class and six-and seven-class solutions (see Fig. A2) and are confident that a five-class solution is the most optimal. The BVR's of this model are presented in Table A5, in the Appendix.
Note that we used five items that were similar to K&B and six that were similar to GBB (see Table 2). In addition to K&B, we included one of the male perspective items on how well fathers are suited to look after children and the well-known and frequently-used item on a preschool child suffers when mother works. In contrast to GBB, we excluded the male perspective item on "responsibility" for home and children by men because the correlation with the 'fatherssuited' item is very high. In contrast, we included the item on "independent", for which we did not see any reason to exclude it. One could argue that this item is a bit vague on what would be the 'best way' to be an independent person. However, we believe that in the current rather still traditional and essentialist context, the majority of people will interpret this question as compared to having a male breadwinner in their family as a competing way to be an independent person. Moreover, this question is asked in the battery of questions on gender roles, right after the "fulfilling housewife" item.

Standard errors in brackets
This table presents the results of the LC-model with five classes, seven items and 24 countries. The second row presents the estimated size of each latent class, while the rest of the rows present the estimated probability of agreeing with each item conditional on class membership. Blue shading is high score on "'traditionalism' in public and private sphere and essentialism", red is high score on "egalitarianism and non-essentialism" Fig. 1 The location of five classes in a three-dimensional gender ideology space

Five Worlds of Norms: A Synthesis of K&B and GBB
The latent class profile of our solution is presented in Table 4.
In interpreting these results, we place the estimated latent classes in a multi-dimensional space (see Fig. 1) that is defined by the recent debate in the literature (amongst which K&B and GBB) and that resembles a coordinate system with the following three dimensions: (1) on the x-axis: the well-known separate/joint sphere dimension when it comes to paid work (the public sphere). This dimension distinguishes individual beliefs according to the extent to which partners should distribute the tasks performing paid employment (both partners or only the male partner participates in paid employment); (2) on the y-axis: the separate/ joint sphere dimension with respect to unpaid work (i.e. domestic work and childcare-the private sphere). This dimension distinguishes individual beliefs according to the extent to which partners should both participate in doing domestic work or whether only women should focus on this; (3) in the colour of the latent classes: the dimension of essentialism that describes to what extent people believe in prescribed separate roles for men and women or whether in individual choice with respect to the division of labour. Although this dimension is actually a continuum (from essentialist to individual choice), for simplicity, we only depict it with three distinct values: essentialist/indifferent/individual choice. 12 The five different gender ideology types that we found in Table 4 are placed in this multidimensional space. The first class that we estimated combines a high score on the joint paid work sphere dimension, a high score on the joint domestic work dimension, and a very low score on the essentialism dimension-i.e. full preferences for individual choice (see the red bubble in the upper right corner of Fig. 1). More specifically, the answering pattern in this latent class is that people agree with both men and women performing in the economic sphere and both contributing to domestic work, thereby assuming that people support this out of individual choice rather than believing in 'pre-set' gender differences. In contrast, items such as "fulfilling housewife", "womenwantkids", and "childsuffers" are generally disagreed with. In line with K&B and GBB, we could have labelled this latent class 'liberal egalitarianism'. In this article, however, we opt for a different label, based on our theoretical expectations laid out in Sect. 2.2. Because egalitarianism (or feminism) does not fully capture the extent of individual choice (as there might be different forms of egalitarianism, see e.g. K&B and GBB), we refer to this class as sameness feminism, i.e. those that support the views of the original second wave liberal feminism where fighting the male privilege is at the core of advancing women's interests in society.
The second class is labelled difference feminism and resembles the liberal egalitarian class with respect to the joint sphere approach for paid work. However, from the perspective of women, a "second shift" is required by combining fulltime paid work in the public sphere with doing the lion's share of unpaid work at home (see the blue bubble in the lower right corner of Fig. 1). This is due to a belief in gendered traits when it comes to caring for children and housework. Thus, "difference feminism" people generally agree with the items "fulfilling housewife" and "womenwantkids". Moreover, they moderately agree with the belief of "childsuffers". Note that in this class although joint paid work could be considered an individual choice we believe this is unlikely. The reason is that, as many values studies point out (e.g. Kalmijn, 2003), during the post-communist period, the housewife ideal is more often valued rather than the ideal of women contributing to the earnings sphere (i.e. it provides families with status if the wife doesn't need to work). Hence, partners jointly contribute to income, most likely not because of individual choice, but more often because of economic necessity. 13 We label the third class as the traditional class. This class combines a strong belief in separate spheres of paid and unpaid work with an equally strong belief in the gendered trait perspective (see the blue bubble in the lower left corner of Fig. 1). People belonging to this class respond in a traditional way to all items. They even mostly disagree with the item "warmrelationship" and moderately disagree with "fatherssuited", which distinguishes it from all other latent classes.
Fourthly, we distinguish the transitionalists from the traditionalists by those who respond more positively to the economic items "contribute" and "independent". They also answer less extreme than traditionalists on the items "warmrelationship" and "fatherssuited". We label this class "transitionalism" because, in contemporary Europe, a distinction between 'egalitarianists' and this class can less rarely be made on the basis of the economic dimension. The lion's share of people supports at least equal paid-work opportunities for men and women-i.e. the first half of the gender revolution is accepted to be completed (Goldscheider et al., 2015). Since this type of people do not have extreme answers considering the separate/joint sphere dimension, nor on the choice/trait dimension, this class is depicted in the middle of the coordinate system of Fig. 1.
Finally, we label the last class to be third wave feminism. This class generally gives less extreme joint sphere responses with respect to paid work compared to'sameness feminists' and 'difference feminists'. In the case of third wave feminists, they have an unconquered belief in individual choice, no matter what the outcome of division of labour for couples would be. Besides providing an "egalitarian answer" on most of the items, one can distinguish them by their agreement on the item "fulfilling housewife" and moderate agreement with the items "contribute" and "independent". K&B (building upon Cotter et al. (2011)) consider this class 'flexible egalitarianists' and they also stated this to be based on the feminist backlash in the 1990s (reacting against the second wave feminists), supported by the third wave feminists. In line with the interpretation of GBB, this class goes beyond gendered traits and hierarchical power relations as these individuals deny any implication for power of whatever division of labour constellation. For instance, an agreement with the item "fulfilling housewife" is considered to be an individual choice and does not imply any kind of traditionalism. Since any combination of paid and unpaid work for men and women is acceptable in this group, this class is depicted as a red 'wall' covering the whole of the coordinate system of Fig. 1 where only individual choice prevails. With respect to the division of labour in the work and caring sphere, both men and women can participate, or either one of the sexes. 'Anything goes'. 14 Table 5 Differences and similarities of gender ideology types found by the five worlds of norms solution of this article, K&B, and GBB (description and percentages belong- Our five-class solution includes four latent classes that are in line with K&Bliberal egalitarianism, flexible egalitarianism, egalitarian familism, traditionalism-and one class that comes close to the "Intensive parenting" class by GBB (transitionalism). In Table 5, we sum up the similarities and differences between our five worlds of norms solution and the solutions of K&B and GBB. Flexible egalitarianism (by K&B), which more or less covers the egalitarian essentialism class of GBB, in our view, does not capture essentialism, but a more flexible way of looking at gender roles (i.e. anti-essentialism), emphasizing individual choice. We therefore specifically reject the label of egalitarian essentialism of GBB and chose the label third wave feminism. The class of traditionalism seems to be corresponding to both K&B and GBB's solutions. The size of the traditional class is, similar to GBB, rather small. K&B had a larger class with traditionalists, but this probably has to do with the fact that, in their solution, the class of the transitionalists is merged with the traditionalist class. This becomes clearer in their Appendix where they discuss a five-class solution. 15 Also, a difference between K&B and our study may be that we could include the "fatherssuited" item that is the only item concerned with father's role in the household. Also, the size of our sameness feminist class is smaller than that of the (liberal) egalitarian class of K&B and GBB. Both the flexible egalitarianism (third wave feminists) and egalitarian familism (difference feminists) classes are smaller in our solution than in the K&B solution. Again, partly these classes are now covered by the transitionalist class, which take a kind of middle position in the coordinate system. This transitionalist class seems to contain men and women that are supporting a separate sphere division of labour when it comes to caring (and women are considered the ones that are better in this than men), although men are expected to contribute to care-and housework as well. However, with respect to the economic field (paid work in the public sphere) both men and women are clearly expected to contribute their share, although this type scores lower on the economic items than the sameness feminist class. This is the reason why we do not label this class "intensive parenting" (that is a concept which mainly refers to the white middle-class and not the entire population), but label it "transitionalists". Another label for this class might have been 'ambivalent', thereby following Scarborough et al. (2019) who, also with a LC analysis on USA data, distinguish between a pro-public anti-private ambivalent type and a pro-private anti-public ambivalent type. These authors state that the traditionalists of earlier periods and generations are the 'ambivalents' of today. However, after careful interpretation of their findings, we decided to also refrain from using this label. Using the term 'ambivalent' would combine both pro-public and pro-private ambivalent classes, and we here find only one class; one that is actually closer to the pro-public class, which could also be considered, in its extreme form, to be difference feminism. By using the term 'ambivalent', we would thus increase more confusion rather than clarity to what this gender ideology type entails.

Sensitivity Analysis: The Importance of Countries and Items
To check the robustness of this latent class solution, we re-estimated the model leaving out one country at a time. The class sizes and profiles vary little over the different country samples. We also compare our 24-country sample with the 17-country sample of K&B to see if a four-class solution would come out. The profile of the five-class solution of the 17-country sample hardly differs from that of the 24-country sample (see Table A7 in the Appendix), hence our solution is robust to the country sample.
However, what also did differ between our extension and the analyses of K&B and GBB were the items that we selected. We selected two different items than the ones that were used by K&B and one item that was different from the ones that were used by GBB (see Table 2, we used the bold items). Hence, our selection of items is closer to GBB but our country selection approximates K&B. Adding one item ("responsibility") to eight items changed the results to a more difficult to interpret and more complex solution with many added direct effects between several items. For theoretical reasons we did not include more items into our models because this would change the scope of the concept of gender ideology to a broader concept that captures also need for a family or work-life balance, dimensions that were not mentioned by Davis and Greenstein (2009). However, we did sensitivity analyses using 11 items covering five Davis and Greenstein (2009) dimensions and using all possible 14 items, which also include additional dimensions to Davis and Greenstein (2009). Neither of these analyses come to a better model fit, let alone to a more parsimonious conclusion (see Appendix Tables A3 and A4).
As an additional measure of validity (criterion-related validity), we relate our five-class solution to seven distal outcomes (the seven items that we did not include in our gender ideology concept) and examine the associations between those seven distal outcomes and our gender typology. The results, which are presented in Table A8 in the Appendix, show that, controlling for socio-demographic characteristics, associations are more or less as expected, confirming the characteristics of each latent class and the findings by K&B with respect to their distal outcome analysis. More specifically, we find a confirming distinction between the three 'feminist' classes with respect to their answering patterns on the distal outcomes. The 'sameness feminists' and 'difference feminists' of the second wave feminist ideals are the most positive towards sharing chores equally in the household, whereas the 'third wave feminists' do not find this necessary at all (i.e., individual choice), and take a middle position compared to the 'transitionalists' and the 'traditionalists'. Further, the 'difference feminists' distinguish themselves from the other two 'feminist' classes in that they highly support family friendliness in a job (this refers to their support for 'special treatment' for women compared to men to be able to be on a par with men), and they have very high support for the item that women need children in order to be fulfilled. Having children is clearly considered a necessity for women's identity for this type of people, whereas for the other two 'feminist' classes femininity doesn't depend so much on that, or the concept of femininity is even ignored or deemed irrelevant altogether ('third wave feminists').

Socio-Economic and Demographic Correlates of Class Membership
As a last exercise, we relate different socio-economic and demographic characteristics to our final latent class outcomes to see if we find some expected and meaningful relationships between them. In Table 6, the logit coefficients of a multinomial logistic regression on the classified latent class scores are presented. This is a three-step approach (Bakk et al., 2013) in which we first estimated the five latent classes. Subsequently, respondents are classified into one of these classes based on their posterior probabilities (using the highest probability of those five classes to assign them to a class). In the last (third) step, we conducted a multinomial logistic regression in which we predict classification to each of these classes (corrected for uncertainty of this classification) by cohort, marital status, education, being religious or not, having children or not, income, and having full time employment or not. As can be seen from Table 6, older cohorts (especially 66+) are more likely to be traditional or transitional and less likely to be any of the feminist types. Note that we would have expected to find a particularly high score on the 'third wave feminism' class amongst the people up till age 50 since this strand of feminism originated amongst those born in the 1960s and 1970s. Against our expectations, we see somewhat higher scores as well amongst older cohorts, suggesting a distribution of third wave feminist ideas in older generations too. Compared to the divorced, singles are more likely to be third wave feminists, whereas the married/cohabitants are more often belonging to the traditional or transitional gender ideology type. The widowed are relatively more likely to be any of the types believing in gendered traits (traditionalists or difference feminists). In line with previous research considering a continuous traditional-egalitarian dimension (e.g. Cunningham, 2008), we find that higher educated people are more often (sameness/third wave) feminist rather than traditional, but they are also quite likely to have a difference feminist answering pattern (which is probably caused by many highly educated women in Eastern Europe). Religious people are the most likely to be traditional or transitional, and the least likely to be (sameness/third wave) feminists. Those who are having children are mostly belonging to the third wave/difference feminist classes (and not to the transitionalist class, which would have justified the intensive parenting label), whereas the childless are more often in favour of either sameness feminist or traditional or transitional answers. Having a higher income is related to being third wave feminist and part-time working as well. The full-time working people are most likely to be of any feminist type.

Summary and Discussion
In this article, we tried to bridge the theoretical and empirical differences in the gender ideology profiles that emerged in two seminal articles in the field (Grunow et al., 2018;Knight & Brinton, 2017). First, we replicated their analyses to exclude the possibility that differences between them could have been caused by errors. Replication is an exercise that thus far has received little attention in sociology, even though it has substantial value. "Substantive replication is required by science in order to help ensure objectivity. Different researchers should arrive at the same results" (Galak et al. 2012, p. 943 in Braun 2016. Although, we could replicate K&B and GGB almost perfectly, we came across certain limitations. Obviously, we recommend researchers to include as much information as possible to be able to replicate their article. Damian et al. (2019) suggest 12 issues that should be discussed in cross-national comparative social science articles in order to check for replicability transparency. Among these issues are information about the dataset used, but also aspects such as the number and treatment of missing values, weighting, the exact operationalization of variables, and as the most complete information, the syntaxes used. Both K&B and GBB discuss almost all of these issues and thus made it rather easy for us to replicate their studies. However, we would like to suggest to scholars to clearly motivate the selection of countries and operationalization of their concepts. In this way, replication of articles becomes even easier and stricter, even though questions of validity can still be raised. By reflecting on the choice of items to operationalize gender ideology, we addressed the validity of the analysis of K&B and GBB. Our replication thus indicates that it is very difficult to replicate without validate.
Second, we extended the research of K&B and GBB by including more countries and choosing partly different items capturing gender ideologies. In this way, we address issues of generalizability and validity of both articles. After carefully selecting seven items that cover four dimensions of the concept of gender ideology defined by Davis and Greenstein (2009) for 24 countries, we come to a synthesis of the two replicated publications on latent classes of gender ideologies in 2008 and our extended analyses yields five latent classes (sameness feminism, difference feminism, third wave feminism, transitionalism, and traditionalism).
Third, we come to this new overarching interpretation of latent types of gender ideologies by placing the five types in a three-dimensional space. K&B and GBB distinguished between the classes of traditionalism and (liberal) egalitarianism and additionally conceptualise and measure some other classes that have some egalitarian aspects. Regarding the overlap of these 'egalitarian' classes, our view is that the concept of difference feminism is the closest to the essentialist view. This means that stereotypical views of men and women in the household prevail even if, in the economic dimension, people believe husband and wife should be on a par. Thus, this type of people (still) believes in gendered traits rather than individual choice as the driving force of the division of labour in couples. Moreover, we argued that the concepts of flexible egalitarianism (K&B) and egalitarian essentialism (GBB) are similar. In our view, these refer to a flexible way of thinking about gender roles; i.e. these individuals are not considering hierarchical power relations related to gender roles (anymore) and believe that individual choice of both men and women prevails to gendered traits. The clearest example in which this cluster distinguishes itself from other clusters is the response to the item "being a housewife is just as fulfilling as working for pay". The choice of being a housewife is seen as an individual choice that is not an indication of a subordination of women towards men (anymore). We labelled this type of people to be third wave feminists because they oppose to the second-wave feminists who claim equality of genders in both the paid work and caring sphere. Finally, the class of intensive parenting by GBB is closest to our transitionalist class. We do not use the same label, however, because the intensive parenting literature is not clear on how this class should score on the economic items "contribute" and "independent". What is clear is that most of the evidence on the adherence of intensive parenting stems from highly educated mothers that were employed in white collar jobs and were in two-parent above-average income families (Wall 2010), which does not cover the entire sample that we have investigated in our study. We therefore used a broader label than intensive parenting.
The reason we found five classes and not four (like K&B) is likely due to the fact that we focused on a different set of items, we only analysed data from 2008 rather than a trend over the period of 1990-2008, and we included next to a country-specific intercept also a sex-specific intercept in the partially homogeneous models. We deliberately chose to leave out items that were not part of the battery that was intended to measure gender ideology in the EVS. Those other items could be affected by "priming" (Lomazzi, 2017) or are capturing something else (e.g. the need for being a mother or father for one's identity) than gender ideology. Therefore, we chose to leave such items out of our models. Nevertheless, we did some sensitivity analyses and found no improvement of model fit when including more items. Generalization of the context to a larger country sample (24 countries) did not lead to a different result than the one that we found based on the 17-country sample.
Of course, our study has a few limitations, of which the most important ones in our view are the limited generalizability to European countries, thereby discarding any other global gender ideologies that may exist; overlooking within-country heterogeneity with respect to gender, race or ethnicity, and class, something that is beyond the aim of this paper and that we leave for future research; inability to look at time trends as we chose to prioritise better generalizability towards the European context; and finally the use of less recent data (2008) because of lack of adequate gender role items in the 2017 EVS wave.

Application of Gender Ideology Types in Future Research
The robust finding of our five-class solution (with the corresponding conditional response probabilities) is important and makes us confident to suggest that the estimated posterior probabilities for each country can be used to form scales measuring the normative context of a country. Table 7 can be used for this purpose. One can construct continuous variables that capture norms on the country level using five classes/measures or less, depending on what one is interested in. Salient are the high "ranking" scores on sameness feminism in the Nordic countries. These welfare states are well known for having both more gender equal policies as well as informal (cultural) gender equality institutions. However, despite all of the Nordic countries being classified as socialdemocratic gender equality focused countries, there are still important differences in the extent to which their citizens culturally support a gender equal division of labour between partners, a point already made a while ago by Pfau-Effinger (2005). Furthermore, notice the scores in the "ranking" for Central and Eastern European countries in the difference feminist measure, which are in line with our expectations. Post-communist systems are known for having a high female labour force participation, while women are simultaneously doing the largest 'second shift' at home, compared to other European countries (EUROSTAT, 2010). Another interesting finding is the "ranking" of typical "part-time work" countries (EUROSTAT, 2011) on the third wave feminist measure. It seems as if in those part-time work countries, women (who are the majority of the part-time workers) do not mind a 'more traditional' division of labour and they do not consider the larger consequences of working part-time for the entire society, and women's (economic power) position in particular. The transitionalist class seems more difficult to recognise in contexts. Typical transitionalist countries are Italy, Latvia, Portugal, Germany, and Switzerland. One could also create a gender norms typology based on the ranking of countries on these five classes. An example is given in the Appendix (Table A9). Note that seven countries have such polarized societies with respect to gender role values that they are difficult to classify into one of the four types and thus are considered "hybrid" (we did not distinguish "pure" traditional countries because the percentage of traditionalist in each of the countries is relatively low compared to other class memberships and the extent of traditionalism in a country will thus hardly differentiate between countries).

Conclusion
In this paper, we replicated and extended two important studies on the multidimensionality of gender ideologies in Europe (Grunow et al., 2018;Knight & Brinton, 2017). We almost perfectly replicated their analyses using the European Values Study and critically reflected upon the multidimensional concepts of egalitarian familism, flexible egalitarianism, egalitarian essentialism, intensive parenting, moderate traditional, and sameness, difference, and third wave feminism, thereby referring to the conceptualization of essentialism in previous literature (Charles & Grusky, 2004;Ridgeway & Correll, 2004). Extending the analyses of K&B and GBB, we found a five-class solution that somehow synthesizes the classes found by the two studies. We label these five classes as sameness feminist, difference feminist, third wave feminist, traditional, and transitional and we place each of these classes in the three-dimensional space of a combination of two separate/joint sphere division of labour dimensions (that in earnings and in caring) and a choice/trait dimension.
In any case, whatever the exact classes/types present in the EU, we believe that we have provided appealing evidence that multidimensionality exists and that use of the gender ideology items in a unidimensional measure to capture gender ideologies is outdated. Instead, scholars should at least specify on which gender ideology class/type they are comparing countries and include a ranking, score, or a normative typology accordingly. To assist scholars in this, we have provided normative context scores on each of the five gender ideology types in Table 7. And for those who might want to use a clustering of countries ("five worlds of norms"), we have suggested one clustering based on the ranking of countries on each of the five classes (Table A9).
We have built these five classes on items available in the European Values Study. Note, that we have considered to also analyse the International Social Survey Program (ISSP) and Generations and Gender Survey (GGS), but these surveys both did not ask for items from the male perspective. Hence, different results are to be expected based on an inferior operationalization of the gender ideology concept. Further discussion on how to operationalize gender ideologies is needed (Maineri et al., 2021;Walter, 2018) and with this study, we hope to have initiated that discussion.