1 Introduction

It is not just the case that gender-specific competence patterns and educational decision making frequently differ between girls and boys as well as between women and men. They also seem to become increasingly divergent over the life course. Even at the age when girls and boys start their preschool education and go on to compulsory schooling, gender-specific educational achievements, interests, choices, and motivations seem to already exist. These small differences in early educational trajectories then seem to become more pronounced during secondary school and grow even further in vocational training and tertiary education. Gender patterns are then most evident in terms of occupational segregation when women and men opt for different jobs and pursue different careers in various sectors of the economy.

Life course research has hypothesized that events and states during earlier educational stages often have important consequences for later educational trajectories and their outcomes in the labor market (Mayer & Tuma, 1990). Dannefer (1987) introduced the so-called Matthew effect hypothesis into the life course literature and used it to describe this phenomenon. In the context of our research project conducted within the DFG Priority Programme 1646 “Education as a Lifelong Process. Analyzing Data of the National Educational Study (NEPS),” we studied whether or not small initial gender differences and competence inequalities between girls and boys at preschool age actually do gradually increase over the school career and the later job trajectory. Thus, the aim of our empirical analyses in this DFG project was to investigate (with longitudinal data and the cohort sequence design of the National Educational Panel Study [NEPS]) the emergence of gender differences in early stages of education and to trace how these differences evolve over the later educational career and job trajectories.

This chapter summarizes the main findings of our project. Our research focused on competence development and educational decision making at the most crucial phases in educational careers. We studied the development and dynamic interaction of gender-specific competencies, self-concepts, and learning efforts, as well as the educational and job decision making that not only families engage in for their daughters and sons but also women and men engage in for themselves. The cohort sequential design of NEPS data allowed us to follow up a stepwise life course perspective. We stringed together different starting cohorts with clear connections between successive phases of the educational career. We started by analyzing the early gender patterns and disparities using NEPS Starting Cohort Newborns and NEPS Starting Cohort Kindergarten. This allowed us to study gender-specific performance up to school entrance. We then analyzed gender-specific performance in elementary school and the transition from elementary to secondary school with the help of NEPS Starting Cohort Grade 5. Then we drew on NEPS Starting Cohort Grade 9 and NEPS Starting Cohort of First-Year Students to focus on the different educational choices of female and male students entering different institutions of tertiary education such as professional college versus university, on gender differences in fields of study, and on dropout intentions. Finally, using NEPS Starting Cohort Adults, we traced changes in gender-specific educational trajectories along with the job entry and careers of women and men over successive birth cohorts.

2 The Development of Gender-Specific Mathematical Competencies in Preschool Children and the Influences of the Family

Preschool age is a formative phase for both girls and boys that lays the foundations for further development and differentiation between the genders. Some studies indicate that there are already appreciable gender differences in the competencies of preschool children: Boys are better in mathematics than girls (Artelt et al., 2001; Blossfeld et al., 2009; Bonsen et al., 2008; Dornheim, 2008; Frey et al., 2010; Jordan et al., 2006; Lonnemann et al., 2013; Niklas & Schneider, 2012) and girls have higher literacy skills than boys (Phillips et al., 2002; Ready et al., 2005; Stanat & Bergmann, 2009).

However, there has been little research on how these gender differences emerge in preschool age. We expect that early socialization experiences in the family establish the foundation for the emergence of subsequent gender-specific structures in the life course. In the literature, the explanations given for the emergence of gender-specific interests and competencies range from biological differences, across varying psychosocial conditions, to the role of the family of origin and differences in the home learning environment (Stanat & Bergmann, 2009). Empirical results indicate that parents have a kind of gender-specific bias toward gender stereotypes (Blossfeld et al., 2009). Moreover, several sociological articles have also stressed the role of family characteristics (especially social class; see, e.g., Connolly, 2004, 2006).

When analyzing how gender differences in mathematical competencies emerge at preschool age, we took the following three different approaches: (1) we applied the framework of the home learning environment that assigns central importance to the domain-specific stimulation that fosters domain-specific competencies such as mathematics (Kluczniok et al., 2013). We assumed that the emergence of gender differences in mathematical competencies would relate to early gender differences in numeracy and literacy stimulation. Therefore, possible gender differences could be explained partly by differences in domain-specific stimulation during mother–child interactionFootnote 1 (Hypothesis H1), because a higher input in terms of stimulation should lead to a better outcome in terms of competencies.Footnote 2 (2) We focused on the hypothesis that differences in social origin would have different implications for boys’ and girls’ competencies (Connolly, 2004, 2006). In particular, we expected that families from various social origins would reveal large differences in the way they treat daughters versus sons. Therefore, according to the sociological egalitarian values hypothesis (Farkas, 1976; van Berkel & de Graaf, 1999), we assumed that both girls and boys would be treated more equally by parents with an advantaged social background, whereas more traditional gender attitudes should play a stronger role in less advantaged families with girls and boys consequently not being supported in the same way. We expected that the size of the potential gender gap would be associated at least partly with differences in socioeconomic status (SES) (Hypothesis H2). (3) According to the literature on school-age children (e.g., Eccles et al., 1990; Eccles Parsons et al., 1983), a positive relationship between parental gender expectations, higher mathematical self-concept, and mathematical competencies is assumed to favor boys. We wanted to know whether these relations are also true for preschool age. However, there is only little research on such relations at this age, although the literature does show that even younger children have a domain-specific self-concept that relates to achievement (Arens et al., 2016; Marsh et al., 2002). Based on previous research (e.g., Marsh & Craven, 2006), we expected that boys at age 6 would have both a higher mathematical self-concept and higher mathematical competencies (Hypothesis H3), and that there would be a positive relationship between both concepts (Hypothesis H4). In addition, according to the model of parental influences on children (Eccles et al., 1990), the higher mathematical self-concept favoring boys found at school age is associated with parental beliefs about their children’s abilities. Therefore, we also focused on these relationships for children at age 6. We assumed that parental assessments of their children’s abilities would relate positively to their children’s self-concept (Hypothesis H5), and that parents would assess their sons’ mathematical abilities as being higher than those of their daughters (Hypothesis H6).

Following these theoretical expectations, we empirically assessed the development of early gender differences in mathematical competencies at ages 4 and 6. For the earlier preschool age, the focus was on the influence of social origin and domain-specific stimulation; whereas for children who are about to enter school at age 6, we concentrated on the association between mathematical competencies, mathematical self-concept, and parental assessment of their children’s mathematical abilities.

2.1 Results of the Analyses at Age 4: Are There Already Pronounced Gender Differences in Mathematical Competencies at This Age, and, if So, Do These Relate to Early Domain-Specific Stimulation of Boys and Girls?

Using NEPS Starting Cohort Newborns, we already observed small gender differences in mathematical competencies at age 4. In contrast to studies on elementary school age, our study showed that mathematical competencies favor girls (see Model 1 in Table 3.1). Girls have a 0.11 higher mathematical competence test score than boys. Yet, this difference is only small.Footnote 3 The interaction term of child’s sex with mother’s SES is not statistically significant (Model 3), implying that in all SES milieus, the observed gender differences (small gaps in favor of girls) are about the same. Therefore, we rejected Hypothesis H2. The observed small gender gap in mathematical competencies cannot be attributed to the numeracy stimulation children experienced in interaction with their mothers at age 2, because the gender gap does not decrease, but even slightly increases (Model 4). Thus, it seems that boys receive a slightly higher numeracy stimulation within mother–child interaction. Including literacy stimulation (see Model 6) leads to a small decrease in the gender gap, indicating that it contributes to the observed gender gap. Nonetheless, the part it explains is not large (6%) and a significant gender gap remains. Therefore, we also rejected Hypothesis H1. The nonsignificant interaction effects between the sex of the child and numeracy/literacy stimulation (see Models 5 and 7) means that girls’ mathematical competencies do not benefit more from higher stimulation in each domain than those of boys.

Table 3.1 OLS models predicting mathematical competence (N = 1956 children)
Full size table

Thus, based on NEPS data, we conclude that at the early age of 4, there is already a gender gap in mathematical competencies that is surprisingly in favor of girls. However, this gap is not large. Furthermore, we cannot explain it well through literacy and numeracy stimulation within the home environment, although the former seems to be surprisingly more important here than the latter. The underlying reasons are small gender differences in stimulation (girls are stimulated more in literacy; boys, more in numeracy), and the comparatively small effects of stimulation on mathematical competencies. Because all models controlled for structural characteristics of the family as well as the child’s characteristics, we conclude that at this early age, girls seem to have a small developmental advantage with regard to mathematical competencies. This development seems to hold in each social origin milieu and at this early age. Therefore, (potential) differences between origin milieus in the socialization of girls and boys do not seem to be very important.

2.2 Results of Analyses at Age 6: Do Gender Differences in Mathematical Competencies Change Between Ages 4 and 6? What Is the Role of the Early Mathematical Self-Concept?

Although at an early age, girls may start out with a small advantage in mathematical competencies, previous studies have demonstrated that at a somewhat later age—at the end of preschool—this gender gap reverses in favor of boys. Is this also true for Germany on the basis of NEPS data, and what is the role of the mathematical self-concept in explaining this gap? To answer these questions, we first analyzed whether the gender gap favoring girls at age 4 does indeed reverse; and, second, whether there is a relationship between gender, mathematical self-concept, and parental beliefs relating to the gender of children at the age shortly before school entry.

Using NEPS Starting Cohort Kindergarten, our results (not reported in the tables) indicate that boys at age 6 indeed have both a significantly higher mathematical self-concept and higher mathematical competencies than girls, although the relationship is not very strong statistically, t(1942) = 2.95, p < 0.01, d = 0.13 for mathematical competencies; r = 0.06, Z(1942) = 2.7, p < 0.01 for mathematical self-concept. Therefore, the gender gap favoring girls at age 4 reverses at the end of preschool age and confirms Hypothesis H3. Furthermore, and surprisingly, there is a negative relationship between the two concepts, t(1942) = 3.4, p < 0.001, d = −0.17. Previous studies interpret this relationship in light of the very optimistic self-concept children have at this young age (Helmke, 1999). Therefore, at age 6, a very high self-concept is not yet associated with high competencies, so we rejected Hypothesis H4. Concerning parental assessment of children’s mathematical abilities, we found no significant relationship with regard to the mathematical self-concept, Z(1942) = −0.28, p = 0.8. Therefore, we rejected Hypothesis H5. However, there is a small relationship with regard to the sex of the child, r = 0.06, Z(1942) = 2.6, p < 0.01. Thus, based on NEPS data, parents assess boys’ mathematical abilities slightly higher than those of girls, confirming Hypothesis H6.

Hence, we found an association between gender, mathematical self-assessment, and mathematical competencies at preschool age that is in line with other findings at school age. Therefore, we concluded that the gender differences we found for the mathematical self-concept and mathematical competencies are related. Second, the parental assessment of children’s mathematical abilities is not related to their mathematical self-concept, although parents assess their children’s abilities differently depending on their sex. Therefore, we conclude that at age 6, there is no relationship between a higher mathematical self-concept favoring boys and parental assessment.

3 Gender-Specific Differences in Elementary School Performance and in the Transition to Secondary Education: Primary and Secondary Effects of Gender and Social Background in the German Secondary School System

Germany is still dominated by the hierarchically ordered secondary school types of Hauptschule, Realschule, and Gymnasium, each offering a particular qualification (along with a small proportion of Gesamtschulen, i.e., comprehensive schools). Whereas girls tend to be more present in the academic tracks of secondary school and Realschule, boys tend to opt more for Hauptschule. Thus, in Germany, the later field and the position reached in the labor market seem to be based on early educational choices such as the choice of secondary school type and vocational training. Types of vocational training differ clearly in terms of reputation and labor market outcomes such as income (Ammermüller & Weber, 2005), unemployment, and occupational status (Reimer & Steinmetz, 2009).

Our project addressed the interplay between social and gender inequalities in shaping families’ educational decisions on the transitions from elementary to secondary education. Specifically, we examined (1) whether social background moderates the gender gap in academic performance in German elementary schools, and (2) whether it further moderates the gender gap in the allocation of students to different secondary school types after fourth grade (net of differences in academic performance). In other words, using a well-established distinction in the educational transition literature (Boudon, 1974), we were interested in whether primary and secondary effects of social background play out differently for children of different genders.

To date, available evidence suggests almost unequivocally that girls generally outperform boys in school, at least when educational achievement is measured in grades (Buchmann et al., 2008). For instance, DiPrete and Buchmann (2013) argue that the major part of girls’ advantage in schools comes from their being more predisposed to the types of behavior that are rewarded in school contexts, such as demonstrating diligence, good conduct, and obedience. Furthermore, similar types of behavior are more likely to be discouraged among boys due to their partial incompatibility with particular concepts of masculinity dominant among youth. In sum, we generally expected that girls would outperform boys in terms of their school grades (Hypothesis H7).

Furthermore, and according to the gender ideology perspective, girls and boys would choose secondary school tracks that enable stereotypically appropriate gender roles (Ridgeway, 2011; Vleuten et al., 2016). Thus, in line with previous findings (e.g., Jürges & Schneider, 2011) we expected that after accounting for their disadvantage in academic performance, boys would be more likely to opt for higher grade educational tracks (Hypothesis H8). Furthermore, and regarding the impacts of social origin, we expected that boys and girls from advantaged backgrounds would benefit equally from their parents, thanks to the more gender-egalitarian norms and values typically held among higher educated parents (Alwin et al., 1992; Dryler, 1998; Guiso et al., 2008). Similarly, boys from more advantaged backgrounds might be more immune to the concepts of masculinity/femininity that stigmatize academic excellence in school, because they have internalized the more successful role models embodied in the success of their own parents (DiPrete & Buchmann, 2013; Francis, 1999). Finally, a complementary argument to predict lower contingency of gender differences on the socioeconomic standing of families can be derived from the theory of compensatory advantage of social background (Bernardi, 2014). This perspective postulates merely that advantaged families are generally better equipped to confront unfavorable events or circumstances. Accordingly, given that boys are generally more sensitive to negative influences from surrounding environments (Buchmann et al., 2008; DiPrete & Buchmann, 2013), more advantaged families should mitigate (or compensate for) the impact of these influences more effectively. In line with these arguments, we therefore expect that gender differences in educational achievement should be less pronounced among children from more advantaged backgrounds than among children from less advantaged backgrounds (Hypothesis H9).

We tested our hypotheses with NEPS Starting Cohort Grade 5.Footnote 4 First of all, our analyses corroborate the widely held belief that girls tend to excel in literacy and boys tend to excel in math (Buchmann et al., 2008; OECD, 2015). Table 3.2 shows that, as of Grade 4, boys’ grades are, on average, 0.24 points higher in math (Model 1), but 0.17 points lower in German (Model 3). In turn, however, the two relative advantages for boys or for girls in these two different domains seem to compensate for each other, and this results in no statistically significant gender difference in terms of GPA (i.e., the average of the two grades, Model 5). Taken as a whole, this does not corroborate our Hypothesis H7 predicting that girls would exhibit higher academic performance than boys. Instead, it reveals that gender differences manifest themselves rather differently in two different domains.

Table 3.2 OLS models predicting elementary school grades in math, German, and their average
Full size table

Do these differences manifest themselves differently when we look at children’s social backgrounds? As far as performance in math is concerned (Model 2), we do find that girls with higher educated parents enjoy a substantially smaller (albeit still statistically significant) difference in grades compared to boys, with the gap favoring boys constituting only 0.19 points (instead of 0.32 for children of lower educated parents). However, we do not find any statistically significant moderating effect of social background with regard to the gender gap in grades in German (Model 4)—if anything, the model suggests that the gap in favor of girls tends to widen slightly among the children of higher educated parents. In terms of GPA (Model 6), the overall pattern replicates the pattern of declining disadvantage for girls in families with higher educated parents, although this is somewhat less pronounced due to the offsetting effects of girls’ persistent advantage in German. In sum, the evidence presented in Table 3.2 provides modest support for H9, according to which we expected the gender gaps in academic performance to narrow for children coming from more advantaged backgrounds.

We now turn to the study of secondary effects of gender and social background—that is, the effects that concern the secondary school track choice over and above elementary school academic performance. Table 3.3 contains the predicted probabilities (and the average marginal difference of these probabilities for boys and girls) of being in each of the respective four states (i.e., Gymnasium, Realschule, and Hauptschule tracks or Gesamtschule) as of Grade 5 (the underlying multinomial regression model is described in the table footer). Statistically significant but small differences emerge only for the choice of the Hauptschule track: Boys are slightly more likely to be drawn into this track in the case of both higher and lower educated parents. Largely, however, the pattern of educational outcomes between boys and girls is quite similar, and it does not appear to be modified whatsoever by the children’s social background. We therefore evaluate this evidence as not supporting our Hypotheses H8 and H9. It also seems to contradict previous findings for Germany, according to which girls are found to be more likely to end up in the academic track even after controlling for academic performance (Jürges & Schneider, 2011).

Table 3.3 Predicted probabilities of track choice by education of parents and gender (N = 3801)
Full size table

4 Gender-Specific Choices in Transitions to and Within Tertiary Education: Do Gender Differences Persist?

Do the gender differences in educational outcomes reported above persist in higher, tertiary education? Do women enter higher education institutions as much as or less than men after Gymnasium graduation, and how (and, if so, why) do study field choices differ between the two genders? Regarding the entry into higher education, one might first think that gender no longer plays any role. The proportions of German women and men who obtain higher tertiary education degrees have become more equal in recent decades (Mayer, 2008), and for several years now, more women than men are entering higher education institutions (Statistisches Bundesamt, 2019). Nevertheless, this does not mean that women aim as high or even higher than men in their educational choices. The growing share of women in higher education may simply be due to the (growing) overrepresentation of women at Gymnasium and their higher exam grades compared to men. Net of these factors (e.g., with similar exam grades), women may still opt more often than men for the lower tertiary education levels. Women may, for example, “more often shy away from university education than men, because the vocational training system provides many attractive opportunities for women to enter qualified service occupations such as nursery or kindergarten teacher” (Blossfeld, 2017, p. 143). Another form of gender segregation in tertiary education is the choice of field of study. If girls opt for the highest tertiary educational level, they less often choose STEM majors (science, technology, engineering, and mathematics) than boys (Charles & Bradley, 2009; for Germany, see Lörz et al., 2011; Uunk et al., 2019)—majors that are assumed to deliver better career and earning prospects than non-STEM majors (Blossfeld et al., 2015).

In our project, we tested two competing theoretical perspectives on (tertiary) education choices of young men and women: the culturalist perspective and the rational choice perspective. The culturalist perspective suggests that female educational choices should be analyzed in light of a preference for “fields characterized by functional or symbolic proximity to the traditional female domestic role” (Charles & Bradley, 2002, p. 581). Males are assumed to be more attracted to fields delivering higher status occupations, and their ambitions will be directed toward majors that lead into higher paid “masculine” and “breadwinner” jobs. The rational choice model dedicates less attention to the role of gender norms and sex stereotypes and explains gender differences in educational choices as being a consequence of gender-specific achievement in different subjects that are indicative of the study success in a particular subject (Jonsson, 1999; van de Werfhorst et al., 2003). On the basis of these arguments and additional assumptions on gender differences in relevant determinants, we hypothesized that we would find only a weak gender difference in the probability of entering general university among Gymnasium graduates (Hypothesis H10): Girls’ relatively higher achievement in German over math (in contrast to boys) can make them opt more often for socially and culturally oriented majors at a lower tertiary level (vocational training and applied university). Yet, the, on average, higher exam grades of girls than those of boys may offset this and lead them more often to general university. We did not assume that boys’ and girls’ orientations toward a job career would affect this decision much, because this life goal appears not to differ strongly by gender (see in Table 3.5 the item “to earn a lot”; cf. Mann & DiPrete, 2013). Moreover, the value of university education can be assumed to be broader. However, we hypothesized that we would find a large gender difference in the choice of the field of study, with women being underrepresented in STEM fields of study and overrepresented in non-STEM fields (Hypothesis H11). This could be expected on the basis of the relative strengths of boys and girls at Gymnasium, with boys showing better math than German performance; and girls, vice versa (see Sect. 3).Footnote 5 Again, we expected that the alternative culturalist model via life goals would be a less powerful explanation for the gender gap in the field of study, because gender differences in these life goals are at most modest (see Table 3.5).

Our analyses on the longitudinally followed NEPS Starting Cohort Grade 9 improve on prior studies on Germany (as well as on other countries) in several ways: First, we improved on the comprehensive work of Lörz et al. (2011) on gender and entry into higher education by investigating realized study choices instead of study intentions. Study intentions may give a distorted picture of gender differences in educational choices if one of the genders abandons its intentions more often (and that appears to be the case in women, as Lörz et al., 2011, themselves suggest). Second, we studied whether relative strength in math compared to German at secondary school can explain the gender gap in the field of study choice (cf. Jonsson, 1999).Footnote 6 The study by Lörz et al. (2011) underestimated the role of relative strength in math (and other explanatory factors) when accounting for the gender gap in study field choice, because it controlled for the choice of taking an advanced math course at secondary school (choosing an advanced math course might actually be another expression of a STEM choice).

Table 3.4 Study choice of male and female Gymnasium students after graduation: Level of (first) further/tertiary education and (first) field of study at general university
Full size table

Our analyses do not display a gender difference in the probability of entering general university after Gymnasium graduation (as in Hypothesis H10).Footnote 7 An equal share of both genders (55% men, 56% women) enters general university as a first further education (see Table 3.4, Panel A). Female Gymnasium graduates opt somewhat less often for the applied university level (Fachhochschule) than male Gymnasium graduates and more often commence vocational training (vocational training, or Lehre, is an increasingly significant pathway for Gymnasium students in Germany; see also Blossfeld, 2017). Yet, these gender differences are not statistically significant.Footnote 8 Furthermore, the gender difference in the probability of entering general university is not larger for children from lower educated parents than for children from higher educated parents, although this might have been expected on the basis of distinct gender socialization patterns within these origin milieus (see Sect. 3 of this chapter). Explanatory analyses revealed that the exam grade received at Gymnasium is the most powerful of the explanatory factors studied for the choice of a general university education and it has a positive effect (better grades, higher probability), whereas the relative math grade does not have a significant effect (findings not reported in the table, yet using the same variables as in Table 3.5). Given equal exam grades, girls appear to have a somewhat lower probability of entering general university than boys, but the difference is not statistically significant. Thus, we concluded that in Germany, girls from Gymnasium “aim as high” as boys in their choice of level of tertiary education.

Table 3.5 Factors explaining the gender gap in STEM fields of study at general university (N = 1088 students)
Full size table

In line with other studies for Germany (Lörz et al., 2011; Uunk et al., 2019) and with our own expectations (Hypothesis H11), our analyses do reveal a large gender gap in the field of study (see Table 3.4, Panel B). Girls much less often choose a STEM major at general university than boys: nearly one-quarter of girls (23%) compared to more than one-half of boys (51%)—a difference of 28 percentage points.Footnote 9 A similarly sized gender difference in the field of study has been noted for other industrial countries (see Charles & Bradley, 2009), though the general share of students choosing STEM is larger in Germany (OECD, 2017). The explanatory analyses in Table 3.5 report that the largest part of this gender gap in STEM study choice can be attributed to relative math performance, in particular with regard to grades (it also has the strongest relative effect on the choice of a STEM major). Almost one-quarter of the gender gap (24%) can be attributed to this factor, which outperforms the contribution of other potential explanations of the gender gap, including academic self-assessment, students’ life goals, and gender-role attitudes (contribution of each of these factors is at most 8%). This implies that the rational choice explanation of the gender gap in STEM fields of study prevails over the culturalist explanation, and that girls partly choose STEM majors at general university less often than boys due to their relative strength in German compared to math (cf. Jonsson, 1999). Notwithstanding, this “relative strength” explanation does not suffice, as is shown by the substantial unexplained gender gap (43%): In other words, even if girls show the same relative strength in math as boys (or vice versa: boys in German), girls less often opt for a STEM major. Therefore, future research needs to further disentangle this important form of gender inequality within tertiary education.

Because about 20% of students leave the higher educational system without a certificate (Heublein & Schmelzer, 2018), we also tested how far the intention to drop out from the chosen field of study differs between women and men in terms of the opinion of significant others, especially for those enrolled in a gender-atypical major. We know that young women face a lower dropout rate than men in general and with regard to all subjects (ibid.). However, research on gender-specific differences in dropout within gender-atypical fields of study is scant and focuses mostly on women in STEM subjects. A common explanation of the larger dropout in STEM than non-STEM is that women show a lower identification with that field of study (e.g., Wolffram et al., 2009). Referring to Tinto’s (1987, p. 93) model and the work of Nora (2001), we know that for students, significant others play a crucial role not only in the phase of integration in the academic system but also in their separation from the well-known home environment. Therefore, the support of parents and peers is important when students think about staying or leaving the system or field of study. We thus expected that the more that significant others support the chosen field of study, the more likely it would be for students to stay in the system and the more the probability of dropping out would decrease (Hypothesis H12). Based on former results (Choy et al., 2000; Wells et al., 2013), we expected the following differences between female and male students: Positive parental and peer influences should show a stronger effect on women than men (Hypothesis H13). Moreover, in gender-atypical fields of study, the opinion of significant others should have a stronger influence on decision-making processes with regard to “doing gender” (West & Zimmerman, 1987) than in gender-typical fields (Hypothesis H14).

Using NEPS Starting Cohort First-Year Students, we first looked at the influence of significant others (parents and peers) on women’s and men’s intention to drop out; and, second, we estimated the influence of significant others by field of study (FEM vs. STEMFootnote 10). The structure of our analyses was based on previous findings (results are not reported here in detail) and indicate that male and female students face different dropout intentions. Therefore, the probability is lower for women in general as well as within FEM and STEM fields.

Table 3.6 shows the influence of significant others on the dropout intention of women and men. Here we find influences of both the parents’ as well as the peers’ attitudes toward the students’ fields of study for both sexes. The more that significant others support the choice concerning the field of study, the lower the probability of dropout intention. This is in line with Hypothesis H12. We further see in Table 3.6 that parental influences are slightly higher than those of peers. Furthermore, the influence of parents’ and peers’ opinion is stronger for men than for women. Contrary to our expectations, the probability of dropout decreases by about 9 percentage points for male students but by only 5 percentage points for female students when the choice of the field of study is supported by parents, compared to students whose parents do not support their field of study at all. We could identify a similar effect for peers: Here the dropout intentions decrease by about 5 percentage points for men and women. Therefore, we could reject Hypothesis H13.

Table 3.6 Average marginal effects of influence of significant others on dropout intention
Full size table

In a last step, we looked at the influence of parents and peers simultaneously. It appears that it is only the parent’s opinion that has a significant effect for male students, whereas we find no statistically significant differences for female students.

We also wanted to know whether there are differences between fields of study. Tables 3.7 and 3.8 show the influence of significant others on the probabilities of the dropout intention for STEM and FEM fields of study split for women and men. When we look at the influence of parents’ opinion on their children’s study subject, we find significant effects for female and male students that do not differ significantly between STEM and FEM fields of study. For men, the probability of dropping out decreases 9 percentage points when the choice is supported by parents. For women the effect is lower: Here, the probability decreases 5 percentage points compared to students whose parents do not at all support the decision for the chosen field of study.

Considering the effects of peers’ opinions about field of study, we find similar results: Again, the effect is (nearly) identical between STEM and FEM fields, but it is stronger for male students (6 percentage points for men and only 4 percentage points for women).

Table 3.7 Average marginal effects of influence of significant others on dropout intention within STEM fields
Full size table
Table 3.8 Average marginal effects of influence of significant others on dropout intention within FEM fields
Full size table

For FEM fields, we find no significant effects of peers’ or parents’ opinions on dropout intentions. These findings support the effects found in our first analyses (see Table 3.6).

Overall, contrary to our theoretical expectations, effects are stronger for male students when we look at the support of significant others. However, for female and male students, the impact of parental opinions is stronger than the impact of peers’ opinions on the intention to drop out. So far, we could not identify systematic differences between male and female (STEM or FEM) fields of study. Therefore, we evaluate this evidence as not supporting our Hypothesis H14.

5 The Role of Mothers in Their Daughters’ Educational and Occupational Career over Cohorts

The role of the family of origin is crucial on two opposite sides: On the one side, performances, aspirations, and choices depend on social origins (Dustmann, 2004). There is a link between overall educational achievements and parental background. On the other side, parents influence children via stereotypes and the cultural norms surrounding gender and gender-specific parenting (e.g., Eccles, 1987) (Hypothesis H15). The literature demonstrates that whereas sons are more oriented toward taking fathers as examples for themselves, daughters tend to refer to their mothers (Huttunen, 1992; Updegraff et al., 1996). The theoretical explanation for this connection is based on the theory of sex-role socialization. It suggests that parents tend to act as models and provide more valuable information if they have the same sex as their child (Boyd, 1989) because they are perceived as having more relevant information and expertise (Acock & Yang, 1984; Boyd, 1989).

The proportions of German women and men who obtain tertiary degrees have recently become more equal (Mayer, 2008). This change in modern families has led to a modification across cohorts in the relative positions of mothers in terms of education and with respect to fathers (Blossfeld & Drobnič, 2001): Mothers are more and more educated, and the education gap between mothers and fathers has been reduced drastically. The occupational level of mothers is particularly important for the job status that daughters achieve (Hypothesis H16): The background characteristics of mothers prove to be more important for daughters than those of their fathers (Khazzoom, 1997).

Following our life course perspective, we wanted to know which consequences these structural changes on the side of parents are having for the development of gender-specific educational trajectories in general as well as for job entry and careers of men and women in Germany across cohorts. Therefore, this section focuses on the intergenerational transmission of education and occupation with special regard to the mothers’ role. It aims, first, to compare the long-term gender-specific changes in educational opportunities and early job careers across birth cohorts; and, second, to analyze the effects of educational expansion on gender-specific structures of occupational segregation at job entry and early careers.

Starting from the theoretical perspective of sex-role socialization (Acock & Yang, 1984; Boyd, 1989; Khazzoom, 1997), we analyzed the separate role of mothers and fathers in determining the educational achievements of daughters and looked at the maternal-line job mobility of daughters by comparing successive cohorts of German students. First, we calculated the predicted probabilities of reaching low, medium, or high educational attainment on the basis of the level of education of the mothers and the fathers in order to study whether the same-sex parent has a diverse role in influencing the same-sex child’s highest level of education (Minello & Blossfeld, 2016). Second, we focused on the maternal-line relation, demonstrating that the association between mothers’ and daughters’ occupational career has changed over time, and that education plays a fundamental role for daughters’ job mobility (Minello & Blossfeld, 2014).

Our first aim was to understand the separate role of mothers and fathers over time and across different levels of education in influencing the educational attainment of their daughters. Using NEPS Starting Cohort Adults, we first analyzed the educational attainment of daughters on the basis of the level of education reached by their parents. We entered the level of education of the mother, then that of the father into the models. First, we investigated the children of parents with low education (at least a lower secondary education or intermediate level with no vocational training); then, those whose parents have medium education (at least an intermediate level or having completed secondary education); and last, we compared the educational attainment of children with highly educated parents (tertiary degree). We compared children born in 1945–1954, 1955–1964, 1965–1974, and 1975–1980.

For daughters of low educated parents, the investment in secondary education and the reduction in low education numbers are already large in the 1955–1964 cohort. We observe an increase in girls obtaining high education, especially among the younger cohorts. The dissimilarities between considering fathers and mothers are accentuated, but only if we consider the youngest birth cohort. We find a strong reduction in numbers of poorly educated daughters over cohorts, accompanied by a strong rise in daughters with medium education and increasing investment in tertiary education, especially when the youngest cohort of mothers is considered.

When their mother or father has a medium education, daughters have a high and increasing probability of attaining a medium education across cohorts. In the past, fathers rather than mothers with medium education had a higher probability of having daughters with low education. However, this tendency has reversed in the young generations. Nonetheless, the differences are very small, and the pattern does not change much when we compare children with either their mothers or their fathers.

Finally, investment in higher education for daughters is different for mothers and fathers, especially when the younger generations are considered. This speaks in favor of Hypotheses H15 and H16. If the mother has a high education, the recent cohorts of daughters have a higher probability of acquiring a high level of education than the daughters of mothers with only medium education. If the father is considered, the gap between the probabilities of receiving medium or high education is narrower. The probability of being poorly educated is close to zero in both cases.

Our second research aim was to test how far daughters experience upward or downward mobility in relation to their mothers when entering the labor market, thereby testing the role of education in affecting daughters’ maternal-line social mobility. We compared the prevalent ISEI position (International Socio-Economic Index of occupational status by Ganzeboom et al., 1992) of the mother and the ISEI position of the daughter at entry in the labor market with respect to the daughters’ level of education and by cohort. The level of education is defined here as it was done before, and the following cohorts are distinguished: 1944–1956, 1957–1973, and 1974–1984.

Figure 3.1 upper panel shows the predicted probability of experiencing maternal-line upward mobility by educational attainment for different cohorts of daughters. It was much easier for the older birth cohorts to experience upward mobility, despite their lower level of education. Daughters born before 1957, with the exception of those without any educational level, have a higher probability of entering the job market with a higher job position than that of their mothers. Daughters born after 1973 can have a high probability of experiencing an upward move relative to their mothers only if they have a high education, hence a tertiary degree (Minello & Blossfeld, 2016).

Fig. 3.1
Two line graphs. a, the graph presents the probability of experiencing maternal line of daughters born in the years on or before 1956, 1957-1973, and on or after 1974. The lines follow an increase in trend from low to high. b, the graph presents the probability of experiencing maternal line of daughters born on or before 1956, 1957-1973, and on or after 1974. The lines follow a decrease in trend.

Predicted probability of experiencing maternal-line female upward/downward mobility by educational level and birth cohorts (to analyze maternal-line educational and job mobility, we identified upward intergenerational occupational mobility as an increase of at least 5 points on the ISEI scale for the daughters’ first job compared to the ISEI of the mother’s job position. We defined downward mobility as a decrease in the occupational status of the daughter compared with that of the mother. We consider the first job position of daughters and the prevalent job position of the mother (intended as the job that the mother has had for most of her life when the daughter was 15 years old). We defined three levels: at least a lower secondary education or intermediate level with no vocational training (corresponding to the German kein Abschluss, Hauptschulabschluss ohne beruflicher Ausbildung, Hauptschulabschluss mit beruflicher Ausbildung, or Mittlere Reife ohne beruflicher Ausbildung); at least an intermediate level or having completed secondary education (Mittlere Reife mit beruflicher Ausbildung, Hochschulreife ohne beruflicher Ausbildung, or Hochschulreife mit beruflicher Ausbildung); and a tertiary degree (Fachhochschulabschluss or Universitätsabschluss). The three levels of educational attainment are renamed “low,” “medium,” and “high” education). (Source: NEPS, Starting Cohort Adults; own computations; N = 3645)

Full size image

Figure 3.1 lower panel reports the predicted probability of experiencing maternal-line downward mobility by educational attainment for different birth cohorts. For the younger birth cohorts of daughters, it is much easier to experience downward intergenerational mobility relative to their mothers. For the oldest birth cohorts, it is least likely, with the exception of women without any educational degree (within the low educated). In every birth cohort, highly educated daughters have a much lower probability of experiencing downward mobility than lower educated women.

6 Conclusions and Discussion

Our project aimed to trace the emergence and stepwise development of gender-specific differences in competence development as well as decision making at the four most important phases/transitions in educational and occupational careers in Germany: (1) experiences at preschool age, (2) achievement during elementary school and the transition to secondary school, (3) transition from secondary school to vocational training or higher education, and (4) occupational careers. Unfortunately, NEPS data do not allow us to observe the educational careers of men and women as a whole over the life course, but we can follow up the development of changes in transitions across the life course for different cohorts thanks to the NEPS cohort-sequence design. Hence, because the different NEPS Starting Cohorts follow different phases of the educational career, it is possible to string them together.

Starting at the earliest phase in the educational career, our analyses show that there are already gender differences in mathematical competencies at early preschool age, but contrary to the usual expectations, they favor girls. This gender gap cannot be explained by previous maternal, domain-specific stimulation in terms of numeracy and literacy, although there is some evidence that girls gain higher literacy stimulation and boys higher numeracy stimulation. We conclude that at age 4, girls tend to make more rapid advances in development, because the female advantage at this age is persistent even after controlling for theoretically important covariates. Going one step further in life, our analysis of the age of 6, which is shortly before school entry, shows that this gender gap reverses to a traditional gender pattern with mathematical competencies in favor of boys. There is some evidence that compared to girls, this might be due to the higher mathematical self-concepts of boys at this age. Furthermore, it is interesting that this higher mathematical self-concept is not related to the fact that parents make slightly different assessments of their children’s mathematical abilities in favor of boys.

In elementary school, these early gender-specific differences are reinforced: Boys perform better in mathematics and girls perform better in German. Nevertheless, because these relative advantages in each domain compensate for each other, there are no differences in overall performance. This finding emphasizes the salience of educational specialization within the school curriculum, and hence potentially suggests that gender inequalities in educational outcomes in the German educational system should manifest particularly in the horizontal stratification dimension (such as, e.g., STEM vs. FEM orientation). With regard to the transition to secondary school, in contrast to previous findings from research, we do not find that girls are more likely to end up in academic tracks. The only difference we find is that boys have a higher probability of transitioning to Hauptschule. Furthermore, our analysis provides evidence that there is a wider gender gap in educational outcomes among students from less advantaged backgrounds. However, this finding is limited exclusively to students’ performance in math, whereas we do not find any declining effect of gender by social background with regard to students’ performance in German, or any similar compensatory effects with regard to the choice of educational tracks.

Concerning the next phase of the educational career—the transition from secondary school to vocational training or higher education—there is some evidence from the data that among graduates, young women tend to opt more often for vocational training than young men, whereas the latter more often choose to study at universities of applied sciences than women. We do not find gender differences regarding university entry, nor do women aim lower with respect to university entry when they have similar grades to boys. However, a clear gender-specific differentiation appears with regard to the field of study at university: Men more often choose a STEM major than women. Our analyses show that to a large extent, this gender gap can be explained by the girls’ higher relative school achievement in languages than in math. But what about those who enroll in gender-atypical fields of study? Do they have a higher dropout intention than those who enroll in gender-typical fields of study? And what role does the opinion of significant others play? So far, we have not been able to identify a systematic pattern of dropout intentions for students within different fields. This might be due to the fact that students in atypical fields are a highly selective group. Furthermore, our analyses show that significant others are important when looking at dropout intentions. The parents’ opinions seem to be particularly important for students, although it turns out that the peers’ opinions do matter as well for the intentions of students to drop out. The more that significant others support the chosen subject, the more the intentions to drop out decrease. However, the effects we find are not as strong as expected, and the differences between males and females remain on a low level. Nevertheless, we find differences between male and female students: Contrary to theoretical expectations, the influences of significant others are generally stronger for male students than for females. This stronger effect for men shows up in STEM and in those fields related more to women (FEM). When considering the opinion of peers and parents simultaneously, we can identify a significant influence only for parents’ opinion on the dropout intention of male students.

Finally, our results show the important role of mothers in shaping the level of education of their daughters. This plays a fundamental role in guaranteeing the chances of maternal-line mobility of daughters, especially among the young generations. We see an increased investment in middle and secondary education for daughters across cohorts. This change is evident when we regard daughters whose parents have low or medium education. When considering highly educated parents, our results demonstrate that when the mother has tertiary education, the daughter has a very high probability of getting a higher education as well, especially when looking at the younger birth cohort (1975–1980). Because this is not the case when we take only fathers with tertiary education into consideration, we conclude that mothers with tertiary education tend to invest more in their daughters than fathers do. Moreover, younger and more educated daughters have greater chances than older and low educated ones of experiencing upward maternal-line intergenerational mobility and avoiding downward mobility. University degrees are, in fact, more and more important for young women living in Germany, because they are the main way to commence a decent job career in a position higher than that of their mothers.

In summary, based on our analyses, the expected cumulative differences among boys and girls and men and women over the life course emerge in line with the so-called Matthew effect hypothesis: Small gender differences at preschool age get bigger over the school career, not so much with regard to competence trajectories, but with regard to the chosen subjects in schools and fields of study in vocational training and tertiary education. Small gender-specific differences in mathematical competencies at the end of preschool seem to be reinforced in elementary and secondary school and lead to gender-specific choices of fields of study in vocational training and tertiary education as well as to gender-specific differences in job careers. Even though our research is limited and only partially longitudinal, we can shed some light on the gender-specific processes over the life course by comparing short-term processes for different NEPS starting cohorts. It is clear that we are only beginning to understand the gender-specific causal mechanisms, and that we need longer individual histories from the data. However, with each additional panel wave of NEPS, researchers will have longer observation windows and a better opportunity to use real longitudinal data to study the educational and job trajectories of women and men from the various NEPS starting cohorts.