Thirty Years after the Fall of the Berlin Wall—Do East and West Germans Still Differ in Their Attitudes to Female Employment and the Division of Housework?

Previous cross-sectional studies highlight persistent East–West differences in gender ideologies after German reuniﬁcation. This study examines the extent to which gender ideologies in the East and West have converged and whether differences are still relevant for younger cohorts who experienced childhood around the time of reuniﬁcation, or after 1989. Using data from the German Family Panel pairfam (2008–2019) and differences in regime-speciﬁc socialization for three cohorts born before and after reuniﬁcation, results reveal that different dimensions of gender ideologies have only partly converged 30years after reuniﬁcation. Attitudes towards housework and female employment converged particularly, yet, in all cohorts, views on maternal employment remain substantially different between East and West. Observed convergence occurred only partly due to contrasting trends of modernization in West Germany and re-traditionalization in East Germany. Moreover, the results highlight smaller attitude changes with increasing age, particularly for the younger cohorts, contributing to further variations in East–West differences. Overall, the ﬁndings conﬁrm the existence of long-lasting ideology differences due to regime-speciﬁc socialization, and a persistently altered composition of society in East and West Germany. At the same time, they point towards slow convergence among younger cohorts due to a more similar institutional and socialization context following reuniﬁcation.


Introduction
Understanding the formation and change of gender ideologies is crucial, as traditional attitudes towards maternal employment and family work are linked to gendered inequalities in paid and unpaid work (Davis and Greenstein, 2009). Previous longitudinal studies have highlighted the importance of parents' views and attitude-related behaviour for the socialization of gender ideologies and related behaviour of their children (Cunningham, 2001b;Farré and Vella, 2013;Platt and Polavieja, 2016). However, there is little empirical research on whether the institutional context in childhood has a similarly long-lasting influence on gender ideologies or whether adults' contemporary institutional frameworks are of greater importance.
To investigate the lasting influence of the institutional context, and in search of external variations, numerous studies have exploited the case of formerly divided Germany and its reunification in 1989/1990. During the 40 years of separation, the socialist German Democratic Republic (GDR) and the liberal, democratically oriented Federal Republic of Germany (FRG) supported gender equality very differently. As East Germans, born before 1990, expressed less traditional gender ideologies even after reunification, earlier crosssectional studies suggest long-lasting regime-specific socialization (Braun, Scott and Alwin, 1994;Adler and Brayfield, 1997;Banaszak, 2006;Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012;Ebner, Kü hhirt and Lersch, 2020).
Given more similar institutional and political conditions in today's reunited Germany, the question arises as to whether younger East Germans have adapted their attitudes, and now express more traditional gender ideologies. According to cohort-replacement theory, younger cohorts with more similar socialization conditions and, thus, similar ideologies replace older cohorts with regime-specific ideologies over time (Bolzendahl and Myers, 2004;Brooks and Bolzendahl, 2004;Baxter et al., 2015). Hence, younger East Germans should express views that are more traditional compared to older East German cohorts and, thus, more similar to those of West Germans. Moreover, assuming attitudinal change over the life-course, older East Germans should have adapted their attitudes at least partially to the post-reunification context.
Empirical evidence leads us to question the assumption of cohort replacement-based convergence and diminishing East-West disparities over time, as some studies indicate a slightly increasing attitude gap after reunification (Banaszak, 2006;Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012). However, evidence stems exclusively from repeated cross-sectional studies based on the General Social Survey, which cannot account for unobserved heterogeneity, or disentangle attitude changes based on cohort-change from individual-level within-changes. 1 Moreover, previous cross-sectional studies concentrated mainly on older cohorts, as the data did not include individuals born after 1970, or did so only to a limited extent (Ebner, Kü hhirt and Lersch, 2020). Therefore, it is unclear whether and to what extent East-West differences in gender ideologies are still relevant for younger cohorts that grew up after reunification.
Focusing on Germany, this study extends the literature on the formation of and changes in gender ideologies, and investigates the role of regime-specific socialization for different dimensions of gender ideologies in East and West. It contributes to the existing literature by (i) examining whether previous crosssectional results on East-West differences in gender ideologies are reproducible with large-scale panel data; (ii) analysing East-West differences with panel data that also include sufficient subsamples of younger respondents born shortly before or after the reunification; and (iii) investigating whether the trends observed in gender ideologies are caused by slowly emerging cohort differences, or also by short-term within-changes in respondents' ideologies. With its historical regime change, the formerly divided Germany offers a unique opportunity to investigate how different political regimes, and longlasting East-West differences in the institutional context have affected the formation of gender ideologies, how persistent these regime-specific influences are, and which factors explain ideology change over time.

Theoretical Framework
Life-course research has highlighted the enormous interdependencies between personal characteristics and actions, and other individuals, as well as institutional and structural conditions, all of which shape individual life-courses. The life-course perspective considers modern life-courses as embedded in three main domains: the family and associated division of labour, the market and the institutions of the employment system, and the state's legal and normative regulations (Mayer, 2003). Accordingly, life-courses engage in a self-referential process, so that later phases are strongly influenced by earlier experiences, especially those from childhood and adolescence. However, life-courses are shaped by ageing and associated life events, and historical differences in institutional context and social conditions, leading to enormous differences in life-courses between birth cohorts, explaining social change over time (Elder, 1994).
Against this background, the following section first discusses the general theoretical framework explaining the emergence of cohort differences in ideologies, followed by a presentation of the cohort-specific socialization contexts in West and East Germany (see Supplementary Table A1).

Formation of Gender Ideologies -The Intergenerational Transmission
The intergenerational transmission of gender ideologies is based on children's direct and indirect socialization experiences (Davis and Greenstein, 2009). Symbolicinteractionist interpretations suggest that children directly observe their parents' behaviour and articulated views, and thus learn about the gendered division of roles and attitudes (Cunningham, 2001b). Parents may also pass on their role expectations by consciously involving their children in the household or care of younger siblings (Cunningham, 2001a;Cordero-Coma and Esping-Andersen, 2018). Empirical research finds statistically significant relationships between parents' and their children's gender ideologies in later adulthood, confirming the long-term influence of early attitude formation (Cunningham, 2001b;Farré and Vella, 2013;Platt and Polavieja, 2016).
The gender-specific behaviour and attitudes of the mother seem to be particularly relevant. Particularly, working mothers have less time than homemakers. In addition to their less traditional ideologies, they may involve their children more in housework, thus passing weaker ideas of gendered roles to their offspring. Indeed, studies highlight the importance of gender ideologies as a channel for the intergenerational transmission of gendered behaviour with respect to maternal employment (van Putten, Dykstra and Schippers, 2008) and housework (Cunningham, 2001b;Cordero-Coma and Esping-Andersen, 2018).
Articulated attitudes can also deviate considerably from gendered behaviour, so that parents' behaviour may be more decisive for children's ideology formation (Platt and Polavieja, 2016). However, few datasets contain information on parents' gender ideologies and their gendered behaviour, so evidence on the relative importance of behaviour modelling and intergenerational transmission of attitudes is scarce. The few available results indicate an additional, lasting and strong relative influence of parents' gender ideologies (Cunningham, 2001b;Platt and Polavieja, 2016).
A rich literature on stratification and social mobility further suggests that parents pass on their ideologies through the intergenerational transfer of resources and demographic characteristics, i.e. education, employment and religious affiliation (Cunningham, 2001a;van Putten, Dykstra and Schippers, 2008;Farré and Vella, 2013). Low educational attainment, increased religiousness and a migration background correlate with more traditional gender roles (Davis and Greenstein, 2009). Via the intergenerational transmission of the social composition, ideologies are therefore also passed on indirectly.
Although parents' important role in the intergenerational transmission of gender ideologies is undisputed, the timing and persistence of socialization effects are not yet completely understood. Parental socialization effects in mid-adolescence are considered particularly persistent (Cunningham, 2001b). According to the impressionable-years hypothesis, young adults have the least stable ideologies compared to greater attitudinal stability at higher age (aging-stability hypothesis; Alwin and Krosnick, 1991). However, some studies suggest that individuals adapt their views to individual circumstances as certain later life events, such as career entry or childbirth, activate gender ideologies they had experienced in childhood: a so-called sleeper effect (Cunningham, 2001a;Min, Silverstein and Lendon, 2012;Platt and Polavieja, 2016).

Formation of Gender Ideologies -Regime-Specific State Socialization
Welfare states are presumed to form gender ideologies in two ways: first, through state institutions and legal frameworks (Zoch and Schober, 2018) and second, through intentional or unintentional influences on the social composition of the population (Banaszak, 2006). Institutions and policies form opportunity structures and thus favour certain types of behaviour and genderspecific roles. According to exposure-based perspectives based on identity theory and social identity theory (Stets and Burke, 2000), individuals construct new images of themselves whenever they occupy specific roles. Changes in role exposure are therefore often accompanied by psychological adaption and stronger identification with the new role or social category, thus altering the self-concept (Banaszak, 2006). In this way, opportunity structures promote gender ideologies positively aligned with the individual's social behaviour.
Family policies provide an opportunity structure for reconciling family and work and thus appear crucial for forming gender ideologies. Cross-national comparisons find family policies to correlate with gendered behaviour, but few studies examine the relationship with direct family policy measures (Kangas and Rostgaard, 2007). Some recent family policy evaluation studies have revealed moderate changes towards less-traditional gender ideologies for the introduction of paternal leave in Norway (Kotsadam and Finseraas, 2011) and increased childcare capacities in Germany (Zoch and Schober, 2018). However, these studies focused on within-changes in ideologies and did not examine general trends.
As a second mechanism, welfare states are presumed to shape society's social composition, thus indirectly influencing gender ideologies (Banaszak, 2006). Additional financial resources or supportive policies strengthen the importance of particular contexts and opportunity structures, e.g. female employment, education, or religion, relevant for ideology formation. Life events, such as completing education, childbirth, and employment transitions, are associated with attitudinal changes due to psychological adjustments to altered roles (Baxter et al., 2015;Zoch and Schober, 2018). Interest and exposure-based explanations argue that people often support those ideologies closest to their behaviour, or from which they benefit. Thus, highly educated or working women are more likely to have egalitarian views because they are more common in their environment, because their behaviour is consistent with these views, and because these women benefit more from these perspectives themselves (Bolzendahl and Myers, 2004;Banaszak, 2006;Davis and Greenstein, 2009). In this way, the institutional context of welfare states affects individual gender ideologies and group differences.
Although parental socialization in early childhood and mid-adolescence is considered to be particularly relevant for ideology formation, the role of regime-specific state socialization has not yet been sufficiently investigated. Most comparative studies implicitly assume that ideology formation in relation to the overall context of parental and welfare state socialization in childhood remains stable in the further life-course. However, the question of whether the institutional context experienced during childhood has a similarly long-lasting influence on gender ideologies compared to parental socialization effects bears great theoretical importance.

Regime-Specific State Socialization in the FRG and GDR
The conservative welfare-regime of the FRG actively supported the marriage-based male breadwinner family with free health insurance via the working husband and marriage-related benefits and tax allowances (Leitner, Ostner and Schmitt, 2008). With no public childcare for under-threes and hardly any institutional, or afterschool care for older children, female employment rates remained comparatively low. If they worked at all, most mothers worked part-time after long familyrelated employment interruptions (Rosenfeld, Trappe and Gornick, 2004).
Gender equality was also limited before the law. As maternal employment was considered harmful, until the late 1970s, husbands could terminate their wife's employment contract for neglect of household duties. Unable to identify their child's father, single mothers were automatically granted an official guardian and lost comprehensive parental custody (formally until 1998). Abortion without medical necessity was a criminal offence until early 1976, and -although not prosecuted under certain conditions -has remained prohibited until today. The debate on the abortion law mobilized the women's movement in the FRG particularly strongly, so that the question of general equality was taken up during the late 1970s (Leitner, Ostner and Schmitt, 2008). The FRG thus followed the pattern of other Western societies in which gender roles became somewhat less traditional (Davis and Greenstein, 2009).
In the GDR, full-time employment was expected for both men and women to compensate for the greater need for manual labour, which emerged due to the often outdated, limited technical support and increased emigration of the skilled labour force. Supportive workfamily policies, gender equality before the law, and propagandized norms of gender equality in education and employment therefore aimed at shaping less traditional perspectives on maternal employment and nonparental childcare. A comprehensive public childcare system provided care for half of the under-3-year-olds, for almost all children between 3 and 6, and -also for older children -after-school care and children's camps during school holidays. Children born before the mid-1970s attended childcare frequently, starting at a very early age, as parental leave was not introduced until 1974, and initially was only for second and higher-order births. It was only extended to all mothers in 1986. As a result, female employment rates (89 per cent, 1989) were highest in the world and comparable to men's (92 per cent) (Rosenfeld, Trappe and Gornick, 2004). Thus, it was usual practice for both parents to work in the GDR and to use public childcare. The working mother's image was even included in textbooks and children's songs like 'Wenn Mutti frü h zur Arbeit geht' [When mommy goes to work early].
Likewise, gender equality was established before the law, and marriage did not restrict women's rights. Despite low alimony obligations, the divorce rate was high and parental rights not restricted for illegitimate births (Leitner, Ostner and Schmitt, 2008). Since 1972, abortions were possible up to 12 weeks without preconditions, and contraceptives were available free of charge. However, it is important to stress that these measures were primarily intended to promote the GDR's economic development.
From a life-course perspective, the socialist indoctrination of propagandized norms with a strong employment orientation began early, since children came into close contact with working life early in school. From grade 7 onwards, children visited factories or farms regularly to perform unpaid, simple manual labour tasks in the production process. 2 Children were members of GDR mass youth movements (Thä lmann Pioneers: age 6-14; Free German Youth: from age 14) organized by the Socialist Party to bring up young work-oriented socialist citizens. They actively participated in mass parades and demonstrations of workers on public holidays, in the organized collection of waste materials to increase their pocket money, or in other, usually unpaid, work assignments, so-called 'subbotniks'.
As mothers usually worked full time, children also had to help to a greater extent in the household. However, although dual employment led to a moderately more egalitarian division of labour between parents, the division of domestic work and childcare still remained strongly gendered (Rosenfeld, Trappe and Gornick, 2004). Since 1952, working mothers and married women were granted a monthly 'household day '. 3 This day off was only granted to men under certain conditions, institutionalizing gendered inequalities in unpaid work. Similarly, additional days off and reduced working hours applied mostly to mothers and not to fathers.
Lastly, the GDR also permanently changed the composition of the East German population. Supportive work-family policies and equal access to higher education contributed to a higher female education and employment level. Conversely, religiosity and church membership were less supported or even actively punished, drastically reducing religiosity. Due to closed borders and the separate accommodation of guest workers, natives and foreigners generally remained separate and to this day, the proportion of people with a migration background is significantly lower in East than in West Germany. Previous research has found more traditional gender ideologies among homemakers, religious persons, and persons with a migrant background (Davis and Greenstein, 2009). Therefore, the different composition of the population partly explains East-West differences in gender ideologies (Banaszak, 2006;Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012).

Socialization in the Post-Reunification Period
Reunification created one German state with the same laws and institutions practically overnight: in the long run, a convergence of ideologies was expected. However, many studies remain vague as to whether convergence was also expected for older cohorts born in the GDR, or only for younger post-reunification cohorts. It is doubtful that parental socialization in East and West became the same overnight. With a different population composition and, in part, still different institutions, life in the East continued to differ from life in the West.
After reunification, due to increased unemployment rates, but also the support offered by extended parental leave, East German mothers began to have more frequent, and longer employment interruptions and were more likely to work part time (Leitner, Ostner and Schmitt, 2008). With greater childcare offers (mostly full time) East German mothers were nevertheless still more likely to work, usually doing more hours, and taking shorter employment breaks than West German mothers (Rosenfeld, Trappe and Gornick, 2004). However, after reunification, East German mothers took on higher shares of household labour than previously, partly because of increased time availability. The gender-specific division of gainful employment and housework became thus more traditional than in the GDR, but remained more egalitarian than in the FRG.
In West Germany, the possibility of gainful employment continued to be limited, and female part-time shares remained high. Only in 1996 was a legal entitlement to a childcare place from age four introduced, and public childcare increased (Leitner, Ostner and Schmitt, 2008). In line with theoretical assumptions on cohortreplacement induced modernization over time, previous studies suggested a continuous decline in traditional gender ideologies (Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012;Ebner, Kü hhirt and Lersch, 2020). The debate on abortion law in the course of reunification and the strong GDR women's movement partly supported this development (Leitner, Ostner and Schmitt, 2008). Nevertheless, as women's employment participation and childcare take-up continued to be low, the socialization of West German children born after reunification remained more traditional than in East Germany.

Hypotheses
To investigate East-West differences in gender ideologies, I consider three dimensions of gender ideologies measured with three attitudinal items: attitudes to (i) female employment, (ii) maternal employment, and to (iii) the division of housework. Based on previous research, I assume that these attitudes are affected differently by socialization experience in East and West, and between cohorts born before and after reunification, leading to the following assumptions: Given long-lasting differences in parental and regime-typical socialization between the FRG and the GDR, and the hypothesis of relatively stable ideologies in adulthood (Alwin and Krosnick, 1991), I expect that the less traditional ideologies of East Germans will continue to differ markedly from those of West Germans long after reunification. Older East Germans born before reunification should express the least traditional attitudes. As younger East Germans spent their formative years in reunited Germany, they are presumed to have somewhat adapted their ideologies to the more traditional institutional framework and West Germans' more conservative ideologies. Thus, I expect the postreunification cohort of East Germans to express more traditional attitudes towards the employment of women or mothers (H1a) and the division of housework (H1b) than older East Germans born before reunification (Retraditionalization-Hypothesis). Conversely, following the general trend towards modernization in all Western countries, I expect the post-reunification cohort of West Germans to express less traditional attitudes towards the employment of women or mothers (H2a) and the division of housework (H2b) than those born before reunification (Modernization-Hypothesis).
With these contrasting trends in East and West, the overall East-West gap in gender ideologies is assumed to have narrowed for younger cohorts. However, I do not expect complete convergence, as East German parents did not immediately adapt their attitudes and behaviour patterns, but partly passed them on to their children. With the institutional differences that still exist, I assume younger East Germans to still have somewhat less traditional attitudes towards the employment of women and mothers than their West German peers. However, the gap in employment-related attitudes should be less pronounced in the post-reunification cohorts than those born before reunification (H3a, Convergence-Hypothesis). Moreover, given a strongly gendered division of housework in both parts of Germany, the attitudinal differences concerning housework should be less pronounced than attitudes towards employment. I therefore assume that East-West differences in attitudes towards the division of housework have completely converged for the post-reunification cohort (H3b).

Data and Estimation Strategy
The German Family Panel The study uses data from the German Family Panel (pairfam), release 11.0 (2008-2019) (Brü derl et al., 2020). Starting in 2008, the annual survey offers detailed information on approximately 12,000 randomly selected respondents of three cohorts (1991-1993, 1981-1983, and 1971-1973) (Huinink et al., 2011). Pairfam is currently the only panel dataset for Germany that regularly surveys items on gender ideologies.
The To distinguish between a socialization experience in East and West Germany, I focused on individuals who were born and live in the respective part of the country. Hence, I excluded respondents who moved from East to West or vice versa during the period of observation (N ¼ 221), and who may have adapted their ideologies to the other context. Homosexual respondents, who tend to have particularly egalitarian attitudes, were also excluded (N ¼ 245). Based on all restrictions, the sample consisted of 5,870 West Germans (28,088 observations) and 2,322 East Germans (11,984 observations). Respondents of the 1990s-cohort were observed between the ages of 14 and 28, followed by the 1980s-cohort between 24 and 38 years old, and the 1970-cohort between 34 and 48 years old, so that the cohorts overlapped only slightly in terms of observed age ( Figure 1). Respondents participated in five waves on average, with 50 per cent of the sample being surveyed in at least five of the seven waves.

Estimation Strategy
The empirical analyses followed a two-step procedure. First, I estimated random-effects growth curve models (REGCM) with clustered standard errors to examine average East-West differences in attitudes for each cohort separately. Baseline models included only an interaction effect of an East-dummy and respondent's age (linear and quadratic term). 4 Further variables were then included to investigate whether the differences observed were related to the sample compositions in East and West.
Second, I applied a 2-fold Kitagawa-Oaxaca-Blinder (KOB) counterfactual decomposition (Oaxaca, 1973;Jann, 2008) to further investigate East-West differences in mean attitudes. This method divides the attitudinal differential into a first part that is explained by compositional differences of the two groups (explained part/ endowment or composition effect). In other words, it investigates the proportion of the part of the East-West gap that can be attributed to East-West differences in characteristics. A second part subsumes the observed differences due to unobserved predictors (unexplained part). This part attributes the remaining proportion of the gap to differences in attitude formation, i.e. how their formation is linked to observed and unobserved characteristics. Table 1 presents the attitudinal items and their wording in the questionnaire used to operationalize three dimensions of gender ideologies: attitudes to (i) female employment (referred to as: FemEmpl), (ii) maternal employment (MatEmpl), and to (iii) the division of housework (Housework). The response scale ranged from 1 strongly disagree to 5 strongly agree. I recoded the egalitarian phrased item on housework, so higher values represented more traditional attitudes. Although an explanatory factor analysis suggested one factor, Cronbach's alpha was rather low for all subsamples (0.47 to 0.54). As agreement on individual dimensions does not have to coincide, and attitudes can develop in different directions, i.e. less traditional attitudes in one dimension are not automatically accompanied by less traditional in another dimension, an index would blur possible differences and diverging trends in ideologies. Hence, I analysed all items separately, which falls in line with recent discussions on conceptually distinct dimensions of gender ideologies (Knight and Brinton, 2017). Accordingly, the three items reflect the gendered division of unpaid work in the private sphere, of paid work, and the intersection of the two spheres in the form of maternal employment.

Independent Variables
Full models included the following individual-level control variables (see Supplementary Table A2): age (centred for each cohort), gender, education, migration background, religiosity, and family status. Householdlevel controls accounted for differences in household income (log), number of children, youngest child's age, and current pregnancy. All models controlled for potential period effects by including the unemployment and childcare rate on the federal level directly. To account for other unobserved influences, changing uniformly over time, all models included period dummies, grouping several years to avoid perfect multicollinearity with respondent's age. Lastly, all models included interviewer gender to account for potential interviewer effects.

Descriptives
East Germans showed lower values for all three attitude measures, i.e. less traditional attitudes, than West Germans (Table 1). East-West differences were statistically significant (t-test, P ¼ 0.05) and, as expected, stronger for the post-reunification cohorts and for employment-related attitudes. In line with the Modernisation-Hypothesis, for West Germany, the youngest cohort stated the least traditional attitudes towards female employment (FemEmpl) and the division of housework (Housework). However, they held more traditional attitudes towards maternal employment (MatEmpl) than older pre-reunification cohorts. For East Germany, the youngest cohort reported the most traditional attitudes of all cohorts only for MatEmpl, while for the other two items, differences between the youngest and oldest cohort were small. Furthermore, East and West Germans differed considerably in their characteristics (Supplementary Table  A2), with particularly striking East-West differences for the two pre-reunification cohorts, and considering religion, migration background, education, and employment. Most East Germans had no denomination and more than 60 per cent never attended church, so that the proportion of non-religious people was twice as high as in West Germany. Similarly, about a quarter of all West Germans had a migration background, compared to only 7-12 per cent of East Germans. Educational attainment and employment participation differed only slightly between the youngest cohorts. In the 1970s-and 1980s-cohorts, however, more East Germans had completed vocational training (medium), while more West Germans had a university degree (high). In contrast, part-time employment was higher in the oldest West Germans cohorts, while more East Germans were in full-time employment, reflecting the different employment participation of women. Older East Germans were cohabiting more often and had more and older children than West Germans, indicating that family formation started earlier. For the post-reunification cohort, many had not yet started a family. Finally, higher unemployment and childcare rates in the East reveal persistent differences in the current institutional context. Overall, the descriptive results suggested that observed East-West differences in gender ideologies may be partly due to differences in composition, especially among the older cohorts, and due to age-specific life events such as employment and childbirth. Table 2 presents East-West differences in attitudes observed at the mean-centred age from cohort-specific REGCM without further independent variables (see M1 in Table 2, full models in Supplementary Tables A3-A5). Across all cohorts, East Germans had lower mean attitude scores at average age 21 (1990s-cohort), 31 (1980s-cohort), and 41 (1970s-cohort). Thus, on average, East Germans expressed less traditional ideologies over the entire observation period. For all cohorts, East-West differences were statistically significant for all three items, except for the 1980s-cohort's attitudes to Housework. Effects sizes corresponded to 7 (Housework) to 64 per cent (MatEmpl) of the respective standard deviation. As expected, East-West gaps were more pronounced for attitudes to the employment of women (FemEmpl, À0.13 to À0.23) and mothers (MatEmpl, À0.44 to À0.72) than housework (À0.06 to À0.15). In line with the Convergence-Hypothesis, East-West differences were smaller for the 1990s-cohort  1971-1973 1981-1983 1991-1993 1971-1973 1981-1983 1991-1993 Gender (À0.13 to À0.44) than those of the two pre-reunification cohorts (À0.15 to À0.72).

East-West Differences in Gender Ideologies
To ensure that the differences, so far observed only at the cohort-specific mean-centred age, did not change over time, and thus level out differences, Figures 2-4 presents the results from Table 2 (M1) graphically by age. Predicted attitudes are plotted in profile plots (left side), whereas the conditional effect plots (right side) illustrate how the estimated East-West differences change with age, and thus, over the observation period. Among the oldest cohort, the conditional effect plots confirm striking and significant East-West differences for MatEmpl and smaller but statistically significant differences for FemEmpl, compared to mostly small gaps for Housework. Conversely, the post-reunification cohort showed the smallest East-West differences, providing partial support for the Convergence-Hypothesis.
Additionally, the graphical display illustrates small within-changes in attitudes, slightly altering cohort's East-West differences over time. With increasing age, Female Employment (FemEmpl) East German À0.23 *** (0.04) À0.14 (0.11) À0.23 *** (0.04) 0.01 (0.10) À0.13 ** (0.04) À0.00 (0.10) Age À0.02 *** (0.00) À0.03 *** (0.01) À0.01 (0.00) À0.02 * (0.01) À0.02 *** (0.00) À0.03 ** (0.01) Age-squared À0.00 * (0.00) 0.00 (0.00) À0.00 *** (0.00) À0.00 (0.00) À0.00 (0.00) À0.00 (   N Cohort-1970 ¼ 14,161;N Cohort-1980 ¼ 12,480;N Cohort-1990    N Cohort-1970 ¼ 14,161;N Cohort-1980 ¼ 12,480;N Cohort-1990  East and West German respondents in all cohorts indicated less traditional attitudes to FemEmpl and MatEmpl. However, for the two older cohorts, West Germans experienced more substantial changes, hence, overall East-West differences in the employment-related attitudes decreased. For the youngest 1990s-cohort, indicating the most traditional attitudes to MatEmpl of all cohorts in the first wave, within-changes were particularly pronounced. Although the gap narrowed considerably over time, yet, East-West differences for MatEmpl remained substantial in size and statistically significant for all cohorts. While employment-related attitudes became more egalitarian, the 1970s-and 1980s-cohorts reported somewhat more traditional attitudes towards Housework over time. With stronger changes in East Germany, attitudes converged. Conversely, the youngest East and West Germans reported less traditional attitudes also to Housework. Though change was particularly strong in West Germany and East-West differences even reversed, the gap remained small and statistically insignificant. Including the full set of independent variables reduced the average East-West differences for all cohorts and items (Table 2 (M2), full models in Supplementary Tables A3-A5 and Figure A1). 5 Effect sizes of the controls were small, and in line with theoretical considerations and previous studies. East-West differences of the youngest cohort were particularly reduced and became statistically insignificant for all items. For the two older cohorts, the coefficient for FemEmpl and Housework ceased to be statistically significant. Overall, compositional effects explained large parts of the East-West differences, yet for MatEmpl the gaps remained substantial for all cohorts and statistically significant for the two oldest cohorts.
Altogether, these first results revealed that even 20-30 years after reunification, East Germans still report less traditional gender ideologies than West Germans. Compositional differences between East and West explained most of the gap, particularly for the youngest cohort. The comparatively small East-West differences in attitudes to housework correspond to Hypothesis H3b, assuming a complete convergence after reunification. Conversely, attitudes towards the employment of women and mothers converged somewhat, but continue to differ with regard to maternal employment as suggested in H3a -even among the post-reunification cohort. The more pronounced and persistent gaps for the older cohorts point towards further unobserved differences linked to different regime-specific socialization experiences. Table 3 presents results for cohort-specific counterfactual KOB 2-fold decompositions of the East-West differences in attitudes (see Supplementary Table A6 for detailed decomposition results). The decomposition is reported from the viewpoint of East Germans. Hence, the explained part illustrates how East Germans' attitudes would look if they had the same characteristics that are relevant for attitudes as West Germans, e.g. education or migration background. The unexplained part attributes remaining disparities to differences in attitude formation, i.e. how their formation is linked to observed and unobserved characteristics. It therefore illustrates how East Germans' attitudes would change if they were formed and linked to characteristics in the way attitudes are formed among West Germans.

Decomposing East-West Differences in Gender Ideologies
The cohort-specific mean attitudes mostly confirmed the REGCMs findings, with less traditional ideologies for East Germany across all items and cohorts (M1 to M9). Except for housework (M4), East-West differences were again more pronounced and statistically significant for the oldest cohort (M1 and M7). Although attitude gaps were mostly smaller for the youngest cohort, the calculated cohort-specific means showed that West Germans did not follow a steady modernization trend. Instead, the youngest cohort in West Germany reported the most traditional attitudes concerning Housework (M6) and MatEmpl (M9). Only for FemEmpl (M1 to M3) did the results confirm the expected continuous modernization, with the least traditional attitudes among the post-reunification cohort (M3). Similarly, younger East Germans did not always report more traditional attitudes than their elders: The 1990s-cohort reported attitudes to FemEmpl (M3) and Housework (M6) that were again less traditional than those of the mid-1980s-cohort (M2 and M5) and, thus, also more similar to those of the 1970s-cohort (M1 and M4). Hence, the results provide only tentative evidence for a continuous and overall trend towards a retraditionalization in East Germany (H1) and a modernization in West Germany (H2).
The decomposition further revealed that all East German cohorts would have more traditional attitudes if they had the same distribution of determinants as West Germans (explained part). However, the results showed that compositional differences in the determinants only partly explained the observed gaps (explained part in per cent). Most importantly, results suggested the important role of differences in the formation of attitudes based on similar observed and unobserved characteristics (unexplained part in per cent). 6 For most items, these behavioural factors explained larger shares of the attitudinal gap (45 to 99 per cent).

Decomposing East-West Differences in Gender Ideologies across Cohorts of Similar Age
The different cohorts overlapped only partially in terms of age so that compositional differences between the three cohorts could result from age-related differences in life events. To reduce age-specific differences, I compared the cohorts again in pairs when observed at similar age. The oldest 1970s-cohort (observed 2008-2012) was compared with the 1980s-cohort (2015-2019) when individuals were 34-38 years old. Similarly, the 1990s-cohort (2015-2019) was compared with the 1980s-cohort (2008-2012) at ages 24-28.
Again, the overall results confirmed mostly small and statistically insignificant East-West differences for Housework and more pronounced gaps for FemEmpl and MatEmpl (see Table 4; Supplementary Table A7 for  detailed composition results). Similarly, the decomposition results for the comparison of the three cohorts of similar age confirmed the presented decomposition results for the full sample. 7 Moreover, the findings confirmed neither a strict modernization trend in West Germany nor a re-traditionalization in East Germany.
Comparing the mean-attitude scores of the oldest 1970s-and 1980s-cohort at ages 34-38 (M1 and M2) showed, as expected, a modernization of West Germans and a re-traditionalization of East Germans for FemEmpl and Matempl. West Germans showed particularly marked changes towards less traditional attitudes (FemEmpl: 2.67-2.84 ¼ À0.17; MatEmpl: 2.55-2.98 ¼ À0.43). For East Germans, the increase in traditional attitudes was of smaller and similar magnitude for both items (þ0.08/þ0.13). As a result, East-West differences decreased and remained statistically significant again only for MatEmpl. Contrary to theoretical expectations, attitudes towards Housework became more traditional not only in the East (þ0.28), but also in the West (þ0.16) and the gap even reversed.
Source: Pairfam 11.0 (2008-2019), own calculation. attitudes became not only less traditional in West Germany, but surprisingly also in East Germany. Despite similar trends, but more substantial modernization in the West (À0.34 and À0.21) than in the East (À0.22 and À0.14), East-West disparities nevertheless decreased, yet remained substantial and statistically significant for MatEmpl (0.42). Moreover, attitudes towards Housework became slightly more traditional in West Germany (þ0.04) and less traditional in East Germany (À0.06), so that the East-West differences for the 1990s-cohort widened again and were substantial in magnitude and statistically significant. Lastly, comparing mean differences for the 1980s-cohort at ages 24-27 and 34-37 provided further evidence for attitudinal changes over the life-course, altering East-West differences over time. Especially West Germans expressed less traditional attitudes towards FemEmpl and MatEmpl with age. Surprisingly, their East German counterparts also reported less traditional attitudes for MatEmpl at older ages. East-West differences decreased but remained substantial and statistically significant for MatEmpl. Considering Housework, both groups developed slightly more traditional attitudes with age. Overall, within-cohort changes were rather small, however, the magnitude was similar to some of the comparisons of two different cohort.

Sensitivity Checks
To ensure that selective panel attrition would not drive the results, I re-estimated all models for respondents participating in at least six waves. However, REGCMs (Supplementary Figure A2, Tables A8-A10) and KOB decompositions (Supplementary Tables A18 and A19) showed similar patterns, reinforcing the presented findings.
Furthermore, I re-estimated REGCMs separately for men and women (Supplementary Figure A3). Men of all cohorts revealed larger East-West differences in attitudes to MatEmpl. Additional models examined gender disparities within East and West (Supplementary Figure  A4), revealing less traditional attitudes to MatEmpl and Housework for women than men in both regions. In West Germany, gender differences were particularly pronounced for MatEmpl and even increased over time.
As gender ideologies and employment are closely related (Davis and Greenstein, 2009;Zoch and Schober, 2018), I examined the role of respondents' employment status (unemployed (ref.), part time, full time) in the REGCMs. To avoid reverse causality, I used information from the previous wave, reducing the number of observations. Results for FemEmpl and Housework showed similar patterns, and therefore confirmed the presented findings (see Supplementary Figure A5 and  Tables A11-A13). For MatEmpl, coefficients ceased to be significant for the pre-reunification cohorts, yet, effects remained substantial (À0.17 and À0.20). A substantial and statistically significant employment effect was only observed for full-time employment, and only for the 1970s-cohort (negative coefficient for FemEmpl and Housework) and the 1990s-cohort (positive coefficient for FemEmpl). Hence, the results revealed that employment differences help to better explain the remaining East-West gap in gender ideologies. As fulltime employment was not associated with less traditional, but rather more traditional attitudes in the youngest cohort, the results provide an empirical example of the formation of attitudes based on similar characteristics changing across cohorts.
Lastly, I examined the role of articulated attitudes and behaviour of respondents' mothers. Unfortunately, their current attitude towards the examined attitude measure, and the information on how many years they had been employed before the respondent's sixth birthday was only available for 50 per cent of respondents' mothers. Mothers' current attitudes and previous employment confirmed expected cohort-specific patterns, with increased maternal employment in West Germany but less employment in East Germany. Hence, East-West differences in mothers' employment and current attitudes were smallest for the youngest cohort (Supplementary Table A14). The additional variables explained the remaining East-West differences in attitudes towards MatEmpl, particularly for the prereunification cohorts (Supplementary Figures A6 and A7  and Tables A15-A17). The results therefore highlight the importance of regime-specific differences in behaviour modelling and the intergenerational transmission of gender ideologies.

Summary and Discussion
Focusing on the context of East and West Germany, this study provides evidence of long-lasting ideology differences due to regime-specific socialization and compositional differences that are closely related to the former division of Germany. By exploiting large-scale panel data and differences in regime-specific socialization for three cohorts born in the early 1970s, 1980s, and 1990s in East and West Germany, the study shows that different dimensions of gender ideologies have only partly converged 30 years after reunification. Attitudes towards the employment of women and the division of housework converged particularly, yet East-West differences in the assessment of maternal employment and the consequences for children's well-being remain substantial. 8 Especially for the oldest cohorts, with the strongest and longest experienced regime differences, East-West disparities in gender ideologies were still particularly pronounced and related to mother's behavioural and attitude differences. However, attitude differences were also evident even in the youngest post-reunification cohort although these gaps were the smallest and mostly explained by compositional differences. These smaller differences for younger respondents point to more similar socialization experiences and living conditions in the formative years following reunification. From a theoretical perspective, the findings provide evidence for persistent cohortspecific socialization experiences in the East and West before and after reunification but also contradict earlier cross-sectional findings on increased East-West differences in gender ideologies (Banaszak, 2006;Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012).
Most importantly, the results provide only partial evidence that convergence occurs exclusively due to contrasting modernization trends in West Germany and retraditionalization effects in East Germany. For West Germany, the results mostly confirmed a modernization trend across all three cohorts and particularly for the employment-related attitudes. However, for East Germany, the expected re-traditionalization only becomes apparent when comparing the 1970s-and 1980s-cohort. The less traditional attitudes for the youngest cohort in East Germany contradicted the theoretical assumption. Further convergence is therefore not the result of opposing trends, but somewhat greater modernization in West Germany.
Beyond this, the panel data confirmed small withinchanges in gender ideologies, contributing to altered East-West disparities for the period from 2008 to 2019. In line with the impressionable-years hypothesis (Alwin and Krosnick, 1991), younger cohorts changed their attitudes more markedly with age than older respondents. Moreover, West Germans in particular developed less traditional attitudes towards the employment of women and mothers, which is most likely related to recent expansion of public childcare provision for under-threes, and associated changes in work-care norms in West Germany (Zoch and Schober, 2018). These significant institutional changes could explain the modernization trend in West Germany in contrast to the stagnation of attitudinal change, or even re-traditionalization of gender ideologies observed in other countries (e.g., Brooks and Bolzendahl, 2004). Lastly, the older cohorts in both regions developed slightly more traditional views to housework with age, confirming previous findings on re-traditionalization processes after the transition to parenthood (Nitsche and Grunow, 2016). Overall, the results therefore expand upon previous cross-sectional studies that were unable to acknowledge both cohort-based changes and short-term within-changes, contributing to overall changes in gender ideologies (Banaszak, 2006;Lee, Alwin and Tufis, 2007;Bauernschuster and Rainer, 2012).
Overall, East Germans of all cohorts would have had more traditional gender ideologies if they had the same characteristics as West Germans. Together with the descriptive findings, the results show that the composition of society in the two regions has somewhat narrowed. However, the observed attitude differences are only partly related to a different sample composition of East and West Germans. Moreover, large proportions of the attitudinal gaps seemed to be linked to differences in attitude formation, based on observed and unobserved characteristics. These characteristics seem to have changed in their influence on the formation of attitudes, so that, for example, for younger respondents, maternal employment today is not as closely linked to egalitarian attitudes as it is for older generations.
A major limitation of the study is the small overlap of the three cohorts in terms of observed age, which did not allow age and cohort effect to be precisely distinguished. As unobserved characteristics are likely to correlate with cohort-specific socialization and gender ideologies, the risk of biased estimates remains. 9 Although estimates are not causal and should therefore be interpreted with caution, they provide a more robust picture than previous cross-sectional findings. Some recent studies have argued that observed East-West differences are not necessarily related to the socialist legacy, but to differences present before the German separation (Becker, Mergele and Woessmann, 2020). Without data to test these assumptions, this debate is unfortunately beyond this paper's focus but might be clarified in further studies. Nevertheless, the presented results confirm ideology changes, with different trends in East and West. As pairfam currently represent the best longitudinal database for gender ideologies in Germany, the present study supplements previous cross-sectional studies with a longitudinal view, and for the first time expands the focus to sufficient subsamples of younger cohorts born shortly before and after reunification.
Overall, this study illustrates that with their very different policies and social norms, the GDR and FRG altered their contexts and the compositions of the populations in East and West Germany so profoundly that even 30 years after reunification, the ideologies of both older and younger cohorts have not completely converged. Small, but persistent ideology differences show that the former division of Germany continues to shape younger generations' ideologies, either via persistent differences in the intergenerational transmission of ideologies as part of parental socialization or/and the intergenerational transfer of resources and demographic characteristics. In any way, the post-reunification cohort apparently has grown up in two contexts that still result in different gender ideologies. As the youngest cohort is observed at comparatively young age, ideologies are likely to change somewhat, particularly in the context of parenthood and associated changes in maternal employment. Given the still different institutional support for reconciling work and family life in East and West, it is unclear to what extent the youngest generation will continue to grow together or perhaps grow apart again. Further convergence of gender ideologies would be based on more similar contexts and living conditions. With the expansion of public childcare -particularly in West Germany -some of the most important institutional differences for gender ideologies are currently the subject of political reforms. These changes provide promising avenues for future research to thoroughly understand how ideology change and possible causes are related. Besides, future research needs to investigate the formation of attitudes and ideologies at early age in different contexts to completely understand the underlying mechanisms of the intergenerational transmission of inequalities.

Supplementary Data
Supplementary data are available at ESR online.
Notes 1 Within-change is usually explained with major life events, i.e. childbirth or related employment transitions (Davis and Greenstein, 2009), but also with political reforms, such as the expansion of public childcare (Zoch and Schober, 2018). 2 The subject was called Productive Work (PA) and, together with the subjects 'Introduction to Socialist Production' (ESP) and 'Technical Drawing' (TC), was part of the practical component of the training at the most widespread 10-year tracks at the polytechnic schools. 3 In 1965, the 'Haushaltstag' was extended first to unmarried mothers with a child under 18 years of age, and in 1977 to unmarried and childless women over 40 years of age. 4 Brü derl and Ludwig (2019) explain the challenges of estimating robust group-specific growth curve models in an otherwise preferred framework with fixed-effects. 5 Surprisingly, for the 1980s-cohort, the small attitudinal difference in terms of Housework even reversed and became statistically significant. 6 As the sum of both decomposition parts needs to equal the estimated gap, negative values for the unexplained part result from large explained parts exceeding the very small and insignificant overall gaps. These large (un)explained parts of the small gap result in relative decompositions parts exceeding 100 per cent. 7 The negative contribution to the explained part implies that the East-West differences would even be larger if characteristics would be the same in East and West. 8 In both East and West Germany, the attitude level, the attitudinal changes, and thus, the East-West differences varied for the examined dimensions. From a broader perspective, the results are therefore in line with previous arguments on the competing ideals regarding gender roles for the market place and family life (Knight and Brinton, 2017). 9 Similarly, estimating cohort differences in attitude trajectories using a joint fixed-effects model for all cohorts would result in group differences that are not interpretable due to the restriction of an equal starting attitude value at age zero. Accordingly, cohort-specific fixed-effects regressions do not provide a well-defined alternative.