Educational heterogamy

Abstract

Socio-demographic heterogamy is commonly believed to lead to lower relationship stability due to the presence of cultural differences between partners. This explanation lacks empirical support, as psychology studies have not confirmed the link between heterogamy and cultural differences, whereas sociological studies have been characterized by problems related to the choice and measurement of cultural differences. Progressing from previous research, this article examines the specific link between educational heterogamy and cultural differences in the domain of child-rearing. Results partly confirm the alternative hypotheses of child-rearing values and behaviors being subject to gender differences and processes of selection between partners. Nonetheless, they also indicate a positive link between educational heterogamy and differences in child-rearing values and behaviors between partners.

Keywords

Child-rearing couple relationship cultural differences diagonal reference models education heterogamy

The importance of socio-demographic characteristics for partner choice has been the subject of a long research tradition. Since the beginning of the 20th century (e.g., Harris, 1912), research has examined the degree to which partners have similar, rather than dissimilar characteristics; in other words, the degree to which they are homogamous, rather than heterogamous.

Homogamous relationships are commonly found to be standard in partner choice. This is explained in terms of three factors. From a cultural point of view, individuals are believed to look for a partner who guarantees a cultural ‘fit’ (Kalmijn, 1998). A cultural fit can express itself in, for example, similarities in values, tastes and specific expectations regarding relationships and the family. People are believed to consciously or unconsciously prefer certain characteristics when selecting a partner, since similarities in values, tastes and expectations are often correlated with similarities in, for example, age, education and ethnicity (Heaton, 2002; Jones, 1996). In addition, ‘third parties’ such as family and friends are believed to stimulate a homogamous partner choice (Kalmijn, 1998; Kalmijn, de Graaf, & Janssen, 2005). Their reluctance towards boundary-crossing heterogamous relationships is related to the possible threat of heterogamy for the homogeneity and the social integration of the group. Finally, structural aspects are considered important because ‘one first has to meet before being able to mate’ (Bozon & Heran, 1989; Lörincz, 2006). As the available pool of possible partners tends to be limited to similar others, homogamy is stimulated at the supply side of mate selection.

The general preference for homogamy is explained by structural aspects and individual and third parties preferences. In addition, the last two factors are believed to lead to lower relationship satisfaction and stability for heterogamous couples (Kalmijn et al., 2005). Because they have crossed certain socio-demographic boundaries, heterogamous partners could face cultural differences, and because they have violated third parties preferences, they could experience a lower level of social support. Thus, the ubiquity of homogamous relationships and vulnerability of heterogamous relationships are considered two sides of the same coin, as both rely on a central dichotomy, cultural similarity and social support in homogamous relationships versus cultural differences and social disapproval in heterogamous relationships.

Although fundamental to heterogamy thinking, this central dichotomy lacks a strong empirical base. Focusing on the distinction in terms of cultural differences, only a handful of studies have empirically tested its validity. An early study by Jorgensen and Klein (1979) examined cultural differences in a sample of 120 U.S. couples. They did not observe an association between heterogamy in social class, religion, adult status and education on the one hand, and value dissensus about marriage or role dissensus about important marital and parental roles on the other hand. In contrast, some later studies did support the hypothesized intermediate role of cultural differences in the relation between heterogamy and marital quality and stability. Janssen (2001) examined several types of heterogamous marriages in a Dutch sample. Mainly for educational heterogamy, he found the link with divorce to be mediated by a scale measuring the frequency of ‘differences of opinions between the spouses about religion, politics, lifestyle, division of labour, children, sexual behaviour and money’ (2001, p. 98). Hohmann-Marriott and Amato (2008) studied the effect of ethnic heterogamy on marital satisfaction, conflict, and instability in the U.S. They found evidence of a mediating role of cultural differences, which they measured as the mean absolute difference between partners’ responses to questions about the family, gender, marriage, social issues and the father role.

Even though these studies show some support for the link between socio-demographic heterogamy and cultural differences, important limitations can be detected in the way cultural differences are operationalized. First, the results are based on a number of wide and seemingly random domains. Some of the studied values and attitudes have immediate reference to the functioning of the couple, for example, those concerning marriage (Hohmann-Marriott & Amato, 2008). Others, however, are more general and do not appear to be linked to the couple or the type of heterogamy under study, for example, those concerning politics (Janssen, 2001). Thus, there appears to be a lack of systematic and foundation in the choice of studied values and attitudes, which is in line with the generally vague description in heterogamy research of cultural similarity as similarity of values, opinions, taste and knowledge (Kalmijn, 1998, p. 399).

Cultural differences have also been quantified in a number of different ways. Whereas most studies have used the mean absolute difference between the answers of the husband and wife on a set of value or attitude questions (Hohmann-Marriott & Amato, 2008; Jorgensen & Klein, 1979), others have focused on the dimension of conflict using a measure based on questions concerning the frequency of conflict/disagreement (Janssen, 2001). Both methods are open to criticism. The use of difference measures has been contested on methodological grounds for its problems such as reduced reliability (Edwards, 2002). The frequency of conflict/disagreement can be informative, yet it does not necessarily provide a good measure for the degree of cultural differences. Thus, the measurement of cultural differences appears to have its limitations as well.

In addition to the limitations of the heterogamy literature, the link between socio-demographic heterogamy and cultural differences is challenged by the findings of the psychology literature. This literature focuses on explaining cultural similarity and has failed to support the role of socio-demographic homogamy in the U.S. (Luo & Klohnen, 2005), Canada (Booth & Welch, 1978), China (Chen et al., 2009), and Brazil (Van Es & Shingi, 1972). Thus, the psychology literature does not link socio-demographic homogamy to cultural similarity, despite the wide variety of attitudes and values that has been studied.

Overall, the link between socio-demographic heterogamy and cultural differences lacks empirical support and requires re-examination. Progressing from the measurement problems of previous heterogamy research, we will examine the link with cultural differences by means of a technique, diagonal reference models (DRMs), that has been shown to be particularly useful for studying heterogamy effects (de Graaf, 1991; Eeckhaut, Van de Putte, Gerris, & Vermulst, 2013). In contrast to the lack of systematic and foundation in the choice of studied values and behaviors, we will consider a set of specific values and behaviors that can be linked to the type of heterogamy under study. The focus will be on educational heterogamy, as achieved characteristics in partner choice – especially education – have gained importance during the past decennia compared to ascribed characteristics (Hou & Myles, 2008). In addition, research has indicated that cultural differences, as compared to social disapproval, provide a more important explanation for the risks associated with heterogamy in achieved than heterogamy in ascribed characteristics (Janssen, 2001). The role of educational heterogamy will be examined for cultural differences between partners in the domain of child-rearing. The choice for the domain of child-rearing will be elucidated in the next section, after which we will discuss the heterogamy hypothesis in more detail.

Which cultural differences?

The central question of this study is whether or not educational heterogamy leads to differences in child-rearing values and behaviors between partners. The choice for the domain of child-rearing is based on two criteria.

A first criterion is that cultural differences are relevant only in so far as they are important for the functioning of the couple or family. This guarantees that they will be ‘discovered’ by the partners and have consequences. In line with this first criterion, child-rearing differences are likely to surface in couples with children. In addition, it is widely supported that differences in such a relevant domain have consequences. Empirical research, for example, confirms that differences in child-rearing can lead to a heightened divorce risk (Block, Block, & Morrison, 1981), a lowered marital quality (O’Brien & Peyton, 2002) and parental stress and negativity (Feinberg, Kan, & Hetherington, 2007; Harvey, 2000), and for the children, emotional and behavioral problems (Baril, Crouter, & McHale, 2007) and problems in terms of social competence (Feinberg et al., 2007). Thus, child-rearing differences fit both demands of the first criterion.

A second criterion is that the type of heterogamy under study needs to be taken into account. This pertains to the fact that it is logical to first study those values and behaviors that are associated with the studied characteristic, as this is where one can expect a link between heterogamy and cultural differences. In line with this second criterion, education and child-rearing are linked by a long tradition of large-scale research. Starting in 1960, an exploratory cross-national study (Inkeles, 1960) showed an association between the social class position (measured by means of education, among other factors) and the importance that is assigned to child-rearing values such as ambition, obedience, decency and honesty. This line of research was continued in the well-known work of Kohn, who tried to explain why social class position matters for the value that parents assign to obedience, self-control, autonomy and so on (Kohn, 1959a; Kohn & Schooler, 1969; Slomczynski, Miller, & Kohn, 1981). Kohn mainly explained this link by means of individual occupational conditions, yet he repeatedly stressed the importance of education as well: ‘Emphasizing the importance of occupational conditions does not minimize the importance of education, which both the original study and the 1973 NORC data show to be powerfully related to parental values’ (Kohn, 1976, p. 544). Other studies confirmed the importance of education (Spade, 1991) and even indicated it as the most influential component of social class (Alwin, 1984). In addition to the child-rearing values, evidence has been found of a link with the child-rearing behaviors. Both Kohn (1959b) and others (Stright & Bales, 2003) found education to be associated with the two basic child-rearing behaviors, control and support.

The focus will be on cultural differences in the domain of child-rearing. Both child-rearing values and behaviors will be considered because the importance of heterogamy could differ for values and behaviors and the specific way in which cultural differences manifest could affect the type of consequences. More specifically, we will examine an important child-rearing value from the research tradition of Kohn, that is, adaptation, in addition to the two basic child-rearing behaviors, negative control and support. Adaptation refers to the degree to which parents value academic and societal achievement, adjustment to the external social environment and adaptability. Negative control indicates the degree to which parents use punishment in their child-rearing and get angry or ignore the child when the child is transgressing. Support refers to the degree to which parents encourage the child to be independent and self-responsible. Whereas support has generally shown a positive link with education, negative control and adaptation have generally shown a negative link (Kohn, 1959a; Stright & Bales, 2003).

The importance of heterogamy

The idea of cultural differences in heterogamous couples is central to heterogamy thinking and upheld by a wide theoretical base compassing both structuralist and interactionist theories. Symbolic interactionism, for example, ascribes the essential advantage of homogamous couples to the fact that they ‘will have organized their stock of experience in similar fashion’ (Berger & Kellner, 1970, p. 58). This similarity is considered crucial, as it allows partners to successfully redefine their subjective realities at marriage. Heterogamous couples, on the other hand, could encounter problems, as their lack of a similarly organized stock of experience, that is, their lack of cultural similarity, will hinder them to successfully develop a new, shared subjective reality. Structuralist theories similarly link social position to values and behaviors, thus raising the issue of how heterogamous partners’ values and behaviors relate.

In line with theory, we expect heterogamous couples, in contrast to homogamous couples, to display differences in child-rearing values and behaviors, with the higher educated partner scoring higher on support and lower on negative control and adaptation (heterogamy hypothesis). However, we also consider several critiques that challenge this heterogamy hypothesis. These critiques weaken the idea of cultural similarity in homogamous couples and question the expectation of cultural differences in heterogamous couples.

The first critique is the finding of gender differences in values and behaviors, independent of the influence of factors such as age and education (Beutel & Marini, 1995). A cross-cultural analysis in 70 countries (Schwartz & Rubel, 2005) showed that there are consistent (though small) differences in the relative importance that men and women assign to most basic values. Even though gender differences are not unanimously confirmed (e.g., Prince-Gibson & Schwartz, 1998), this suggests that certain cultural differences could be present in homogamous couples, due to differences in gender or associated factors. This leads to a first additional hypothesis, the gender differences hypothesis, which states that homogamous partners on average display differences in support, negative control and adaptation due to gender differences in values and behaviors.

A second critique concerns the influence of characteristics such as age and education. Implicit in the homogamy–heterogamy dichotomy is the assumption that the studied characteristic will influence men and women analogously, as this is necessary for assuming an association between the degree of heterogamy and cultural differences. However, the question is whether it is plausible for all characteristics to show a similar association with the values and behaviors of men and women. In the case of education, for example, it is possible that a high education will be associated with different values and behaviors for women than for men. Especially for older couples, such differences could be substantial, as suggested by the proportion in 1975–1976 of 65 women per 100 men in higher secondary education in the Netherlands (Eurostat, 2003). Even though studies have not supported such an interaction effect (Prince-Gibson & Schwartz, 1998; Verdonk, Harting, & Lagro-Janssen, 2007), it appears advisable to consider the gender interaction hypothesis, which states that homogamous partners on average display differences in support, negative control and adaptation due to gender differences in the link between education and values and behaviors.

Not only the cultural similarity of homogamous couples but also the expectation of cultural differences in heterogamous couples can be questioned. A third critique is that multiple selection effects can be effective in partner choice. For instance, it is possible, and even likely, that individuals not only prefer a partner with a similar social position but also, and maybe primarily, a partner with similar values and behaviors. In a situation where 90% of highly educated men score high and 10% score low on a value, while 90% of lowly educated women score low and 10% score high on the same value, it could therefore be that when a highly educated man chooses a lowly educated woman, it will be a woman belonging to the 10% high scoring on that value. Research found evidence of such a selection on values and behaviors (Houts, Robins, & Huston, 1996; Luo & Klohnen, 2005), suggesting that substantial cultural differences are ‘bypassed’ in heterogamous couples. This leads to a third additional hypothesis, the selection hypothesis, which states that heterogamous partners on average display differences in support, negative control and adaptation that are smaller than expected based on educational differences, due to the processes of selection.

A final critique is that individuals can influence each other’s values and behaviors. Research labels such processes as attitude alignment (Kalmijn, 2005) or value transmission (Roest, Dubas, Gerris, & Engels, 2006) and shows that these processes have a positive effect on the amount of accordance, mainly in close relationships such as the partner relationship. In addition to the direct effect of the values and behaviors, the partner can have an indirect effect. Van der Slik, de Graaf, and Gerris (2002), for example, found that a woman’s child-rearing values are affected by her own and her husband’s level of education. Overall, this suggests that differences between heterogamous partners could be smaller than expected due to processes of convergence over relationship time. Thus, the convergence hypothesis states that heterogamous partners on average display differences in support, negative control and adaptation that are smaller than expected based on educational differences due to processes of convergence.

Method

Data

The hypotheses are examined using cross-sectional data from the survey Child-Rearing and Family in the Netherlands. The survey data, collected in 1990, were aimed primarily at studying the relation between environmental factors and the internal functioning of the family in the Netherlands (Gerris et al., 1992). Accordingly, they provide information concerning a range of child-rearing values and behaviors useful for this study.

In the first stage, a selection of Dutch municipalities, distinguished by regional zone and degree of urbanization, was made. Within these selected municipalities, a random sample of 9- to 16-year-old children who live with their parents was taken. In most cases, these parents were married (90.3%) and had one to three children (M = 2.5, SD = 1.0, max = 12). The average marriage duration was high (17.2 years, SD = 3.7), whereas parents’ age ranged from 27 to 65 years for men (M = 42.4, SD = 5.0) and 26 to 54 years for women (M = 40.0, SD = 4.3). Most of the spouses were born in the Netherlands (95.1%) and specified their religious group membership as Catholic (41.8%) or Protestant (27.3%) (approximately 1% unknown and remainder did not belong to a religious group). The socio-economic characteristics of the sample were heterogeneous with 9.5% of individuals (♀/♂: 4.1/14.7%) performing unskilled jobs, 13.9% (17.9/10.1%) skilled jobs, 37.0% (23.6/49.9%) lower technical/managerial jobs, 9.2% (12.9/5.7%) small trade jobs, 15.7% (22.5/9.2%) midlevel technical/managerial jobs and 14.7% (19.0/10.5%) higher professions. The final sample of 788 families was questioned using both structured interviews and questionnaires. Both parents were questioned, resulting in a multi-actor perspective. The response rate of 43% is comparable to that of other Dutch surveys, such as the Netherlands Kinship Panel Study (Dykstra et al., 2005).

In the analysis, we examine a subset of the families included in the survey. We exclude non-married couples (89 families) due to their small number in the sample. We also exclude families in which one or both partners were born outside the Netherlands (55 families), to reduce heterogeneity on non-studied factors. Because the studied values and behaviors partly concern the rearing of children in the family, we exclude families in which the target child does not live at home (2 families) and families in which the parents indicated that they are not the genetic parents of the child (10 families). Finally, one family is not included in the analysis due to incomplete information on education. Thus, our final sample consists of 631 families, 80.1% of the original sample.

Variables

The focus is on cultural differences in the domain of child-rearing (Table 1). The main advantage of the survey Child-Rearing and Family in the Netherlands is that it uses scales to measure parents’ child-rearing values and behaviors. The child-rearing behaviors are based on the average scores of the individual on three scales: ‘punishment’ and ‘ignoring’ for negative control (α_♀/♂ =.69/.68, N _♀/♂ = 548/553) and ‘autonomy’ for support (N _♀/♂ = 543/553). The child-rearing value of adaptation (α_♀/♂ =.73/.73, N _♀/♂ = 618/629) is based on the average scores of the individual on three important values from the research tradition of Kohn: ‘achievement’, ‘conformity’ and ‘adjustment’. The original six items were questioned using seven-point Likert scales (see Appendix 1). The internal consistency and the construct validity of the items have already been validated by previous research (Gerris et al., 1993). The correlations between the partners’ scores range from .26 (p <. 01) for autonomy to .49 (p <. 01) for punishment. The correlations between the various items and variables are presented in Appendix 2.

Table 1.

Descriptive statistics.

	Percentage/M (SD)
	Men	Women
Child-rearing behaviors
Negative control	1.7 (1.0)	1.6 (1.0)
Support	3.7 (0.8)	3.6 (0.8)
Child-rearing values
Adaptation	3.8 (0.8)	3.8 (0.8)
Education of the (wo)man
Less than elementary school	2.5	1.6
Elementary school	6.0	12.8
Lower technical and vocational training	31.9	29.5
Lower general secondary education and the first three classes of the higher general secondary and pre-university education	13.2	21.4
Intermediate vocational education	17.0	18.1
Upper school of the higher general secondary and the pre-university education	7.1	6.0
Higher vocational education	12.4	8.1
University education	10.0	2.5
Age of the (wo)man	42.5 (4.8)	40.1 (4.2)
Number of children	2.5 (1.0)
Age of the target child	12.8 (2.2)
Gender of the target child
Boy	49.1
Girl	50.9
Degree of urbanization of the place where the (wo)man grew up
Big city	29.6	29.8
Small city	29.0	29.5
Urbanized rural	28.8	28.7
Rural	10.5	10.9
Level of education of the father of the (wo)man
Low	41.7	37.9
Middle	31.9	35.5
High	12.5	10.8

The main independent variables in the analysis are the education of the man and the education of the woman. These are based on information concerning the highest level of education attained: (1) less than elementary school, (2) elementary school, (3) lower technical and vocational training, (4) lower general secondary education and the first three classes of the higher general secondary and pre-university education, (5) intermediate vocational education, (6) upper school of the higher general secondary and the pre-university education, (7) higher vocational education and (8) university education. The educational categories are also used to test the effect of educational heterogamy in the analysis (see section on Analysis Technique).

Finally, we control for a number of variables that could affect the link between educational heterogamy and child-rearing values and behaviors: age of the man and age of the woman; number of children; age of the target child; gender of the target child; the degree of urbanization of the place where the man grew up and the degree of urbanization of the place where the woman grew up and the level of education of the father of the man and the level of education of the father of the woman. All these variables are centred or effect coded in order to estimate the effect of educational heterogamy for the ‘average couple’. Missing values are imputed (multiple imputation, five imputations (fully conditional specification)) to retain the maximum number of cases in the analysis.

Analysis technique

Starting with the heterogamy hypothesis, we focus on the individual values and behaviors of the men and women. The effect of educational heterogamy on these individual values and behaviors is examined by means of DRMs. DRMs, originally named diagonal mobility models, were suggested by Sobel (1981) as a way of studying the effects of social mobility. In addition, this technique has been favourably judged in relation to other models for examining status inconsistency, mobility and heterogamy effects on a number of class-related behaviors and outcomes (de Graaf, 1991; Eeckhaut et al., 2013; Tolsma, de Graaf, & Quillian, 2009). Its main advantage is that it offers a parsimonious and interpretable approach to analyzing the effect of educational heterogamy, over and above the effects of education, while controlling for other covariates.

DRMs represent the theoretical idea that homogamous couples represent the core of their group. When cross-tabulating the education of the man by the education of the woman, homogamous couples can be considered diagonal referents for couples who are located off the diagonal, heterogamous couples. Each heterogamous couple with characteristics i and j is determined by the values as estimated for the homogamous couples, the diagonal values μ _ii and μ_jj . The relative impact of these diagonal referents, that is, of the characteristic of the individual and partner, is estimated by means of a salience parameter – p – with a value between 0 and 1.

The baseline model describes the value of the dependent variable Y _ijk as a function of both spouses’ educational levels (i and j), complemented by a stochastic error term with expected value 0. To this model, covariates can be added, which means that, for example, control variables can be included in the models (in equation (1): l different variables X).

Y_{i j k} = p \times μ_{i i} + (1 - p) \times μ_{j j} + Σ β_{l} \times X_{i j l} + ε_{i j k}

i = 1, \dots, T; j = 1, \dots, T; k = 1, \dots, n_{i j}

Similarly, heterogamy variables can be added (in equation (2): w different variables H) to test for an effect of educational heterogamy, over and above the educational effects. Accordingly, it can be examined whether an individual’s values and behaviors are linked only to education, indicating that educational and child-rearing differences between partners will be related, or also to educational heterogamy, for example, in the form of a counteracting heterogamy effect that cancels out the differentiating effect of educational differences in heterogamous couples.

Y_{i j k} = p \times μ_{i i} + (1 - p) \times μ_{j j} + Σ β_{l} \times X_{i j l} + Σ β_{w} \times H_{i j w} + ε_{i j k}

i = 1, \dots, T; j = 1, \dots, T; k = 1, \dots, n_{i j}

Five commonly used operationalizations of heterogamy are tested. The first three equations examine the effect of different categorizations of couples: a dichotomous (homogamous/heterogamous), a three-category (homogamous/woman higher educated/man higher educated) and a five-category heterogamy variable (homogamous/woman 1–2 categories higher educated/man 1–2 categories higher educated/woman ≥3 categories higher educated/man ≥3 categories higher educated). The last two equations examine the effect of two non-categorical heterogamy variables: the signed (education man–woman) and the absolute difference in educational levels.

All models are estimated using the non-linear regression command in the Statistical Package for Social Sciences. The best-fitting models are identified by means of a model selection procedure (F test). To insure that the results obtained are not vulnerable to the effect of small representation in some (combinations) of the educational categories, all analyses are re-examined using aggregated educational variables. These supplementary analyses use educational variables with only three or four (aggregating categories 1–2, 3–5, 6–8 and 1–2, 3–4, 5, 6–8, respectively), instead of eight educational categories. Unless they yield divergent results, the results of these supplementary analyses are not discussed.

In addition to the DRMs, different analyses are needed to test each of the additional hypotheses. The gender differences hypothesis is examined by comparing the overall differences in scores between men and women on each of the three dependent variables (independent samples t test). To also take into account (a) the possible role of differences on other factors, such as education or age, and (b) the possible role of educational differences between partners, an additional comparison is made between the means for homogamous men and women based on the results of the DRMs with control variables (μ ₁₁–μ ₈₈ in equation (1)).

The gender interaction hypothesis is tested by means of a gender comparison of the explained variation by education and the link with education. This gender comparison is based on the results of education in the DRMs with control variables (equation (1)).

The selection hypothesis is examined by comparing the zero-order and partial correlations between the scores of partners for each of the studied values and behaviors. The partial correlations control for educational heterogamy (five-category variable) to examine whether educational assortative mating is at the base of the zero-order correlations (partial correlation that is much smaller in value or zero) or whether active selection on values and behaviors is taking place (partial correlation that equals the value of the zero-order correlation). In addition to educational heterogamy, four other forms of heterogamy are controlled for that could be linked to educational and child-rearing differences between partners. Heterogamy in age divides the couples into three categories based on the year of birth of the partners (1 missing): (1) homogamous couples (age difference ≤2 years, N = 350), (2) husband older couples (N = 264) and (3) wife older couples (N = 16). Heterogamy in current occupation is based on a division of current occupations in six categories (see section on Data). Based on these categories, the couples are divided into four groups: (1) homogamous couples (N = 79), (2) couples in which the man holds a ‘higher’ occupation (N = 127), (3) couples in which the woman holds a higher occupation (N = 51) and (4) couples in which (one of) the partner(s) is/are currently unemployed (N = 351). The 23 missing values are added to the analysis in a separate group. Heterogamy in the educational attainment of the fathers of the man and woman is based on the same eight educational categories as educational heterogamy of the partners (see section on Variables), and divides the couples into three categories: (1) homogamous couples (N = 272), (2) couples in which the father of the man is higher educated (N = 105) and (3) couples in which the father of the woman is higher educated (N = 110). The 144 missing values are added to the analysis in a separate category. Finally, heterogamy in the degree of urbanization of the place where the man and woman grew up is based on a categorization of urbanization in four groups (see Table 1) and divides the couples into three groups: (1) homogamous couples (N = 346), (2) couples in which the man grew up in a more urbanized place (N = 135) and (3) couples in which the woman grew up in a more urbanized place (N = 132). A separate group incorporates the 18 missing values. Heterogamy in religion and ethnicity are not considered here, as our selection of couples includes only couples in which both parents are born in the Netherlands, and also in terms of religious affiliation, the sample is by and large homogamous (see section on Data).

The convergence hypothesis, finally, is examined by means of the correlations between marriage duration (in years, N = 631, M = 17.4, SD = 3.3) and the absolute differences in values and behaviors between partners. Given the contested nature of difference scores, an additional test is performed that has been suggested by Watson and colleagues (2004, p. 1054). This test consists of executing a regression analysis for each of the individual values and behaviors. In the first model, the independent variables are the analogous individual score of the partner and the marriage duration. The second model adds the centred interaction between these two main variables. Thus, a total of six regression analyses, each build up out of two models are examined.

Results

Exploratory analysis

The link with educational heterogamy is explored by examining the mean scores on the dependent variables according to education. This is done using the aggregated educational variables with only three categories (see section on Analysis Technique, aggregating categories 1–2, 3–5, 6–8) to get a better idea of the general patterns.

The values and behaviors of both men and women are significantly associated with their own level of education. These associations are strongest for the child-rearing values of adaptation and weakest for the child-rearing behaviors of negative control (Table 2(a) and (b); see below for the results of the F tests). In addition, the values and behaviors of both men and women are associated with the level of education of their partner. This is of most importance for the woman, as two out of three reported values and behaviors appear strongly associated with the education of the male partner. Only for support (♂: F _2,550 = 5.93, p < .01; ♀: F _2,550 = 1.88, p > .10), the woman’s own level of education appears to be the dominant factor, whereas for negative control (♂: F _2,550 = 5.72, p < .01; ♀: F _2,550 = 9.89, p < .01) and adaptation (♂: F _2,626 = 21.20, p < .01; ♀: F _2,626 = 22.57, p < .01), the education of the male partner shows a (somewhat) stronger association. For the man, the education of the female partner appears less relevant than the education of the man himself. This applies to both negative control (♀: F _2,545 = 7.15, p < .01; ♂: F _2,545 = 2.45, p < .10), support (♀: F _2,540 = 8.90, p < .01; ♂: F _2,540 = 5.22, p < .01) and adaptation (♀: F _2,615 = 36.53, p < .01; ♂: F _2,615 = 17.69, p < .01).

Table 2.

Values on the dependent variables according to education.^a

	(a) Mean values of men				(b) Mean values of women				(c) Difference between mean values of men and women (men–women)
Education man	Education woman				Education woman				Education woman
Education man	Low	Middle	High	Mean	Low	Middle	High	Mean	Low	Middle	High	Mean
	Negative control, N = 548				Negative control, N = 553				Negative control
Low	1.64	2.10	1.97	1.99	1.18	2.02	1.97	1.83	0.46	0.08	0.00	0.15
Low	10	32	3	45	10	33	3	46	0.46	0.08	0.00	0.15
Middle	2.00	1.77	1.84	1.81	1.84	1.65	1.78	1.69	0.16	0.12	0.06	0.12
Middle	59	254	23	336	60	256	24	340	0.16	0.12	0.06	0.12
High	1.46	1.53	1.47	1.50	1.72	1.46	1.10	1.32	−0.26	0.07	0.37	0.18
High	5	92	70	167	5	93	69	167	−0.26	0.07	0.37	0.18
Mean	1.92	1.74	1.57	1.73	1.75	1.63	1.29	1.59	0.17	0.11	0.28	0.14
Mean	74	378	96	548	75	382	96	553	0.17	0.11	0.28	0.14
	Support, N = 543				Support, N = 553				Support
Low	3.24	3.37	4.14	3.39	3.33	3.79	3.71	3.68	−0.09	−0.42	0.43	−0.29
Low	10	31	3	44	10	33	3	46	−0.09	−0.42	0.43	−0.29
Middle	3.61	3.56	3.85	3.59	3.65	3.49	3.76	3.54	−0.04	0.07	0.09	0.05
Middle	58	253	23	334	60	256	24	340	−0.04	0.07	0.09	0.05
High	4.26	3.81	3.90	3.86	3.29	3.54	3.89	3.68	0.97	0.27	0.01	0.18
High	5	91	69	165	5	93	69	167	0.97	0.27	0.01	0.18
Mean	3.61	3.61	3.90	3.66	3.59	3.53	3.85	3.59	0.02	0.08	0.05	0.07
Mean	73	375	95	543	75	382	96	553	0.02	0.08	0.05	0.07
	Adaptation, N = 618				Adaptation, N = 629				Adaptation
Low	4.33	4.02	4.29	4.11	4.15	4.06	3.68	4.06	0.18	−0.05	0.61	0.05
Low	13	36	3	52	14	37	3	54	0.18	−0.05	0.61	0.05
Middle	4.05	3.94	3.68	3.94	4.07	3.85	3.50	3.86	−0.01	0.09	0.19	0.08
Middle	70	288	25	383	72	291	27	390	−0.01	0.09	0.19	0.08
High	3.66	3.48	3.32	3.42	4.08	3.52	3.36	3.47	−0.42	−0.04	−0.03	−0.05
High	5	105	73	183	5	105	75	185	−0.42	−0.04	−0.03	−0.05
Mean	4.07	3.83	3.44	3.80	4.08	3.79	3.40	3.76	−0.01	0.05	0.04	0.04
Mean	88	429	101	618	91	433	105	629	−0.01	0.05	0.04	0.04

^aBased on the aggregated educational categories (aggregating categories 1–2, 3–5, 6–8). Note. Values indicated with grey shade denote mean is below the means of both homogamous counterparts, while values in light grey shade denote mean is above the means of both homogamous counterparts. Values in italics denote N. Values in boldface denote mean scores for the homogamous individuals/combinations.

The patterns for education overall confirm the results of previous research, as they show a positive link between education and support and a negative link between education and negative control and adaptation (Table 2(a) and (b)). This confirmation of the link with education leads to the question of the importance of educational heterogamy. The distinction between homogamous and heterogamous couples can be explored by examining the differences in the mean values and behaviors of men and women according to educational heterogamy (Table 2(c)). For example, the differences for the heterogamous combination in which the man and woman are located in the educational categories low and middle, respectively, can be compared to the differences for the homogamous combination in which both partners are in the educational category low and the differences for the homogamous combination in which both partners are in the educational category middle. In only one out of three studied values and behaviors, these differences (largely) follow the pattern as expected from the observed associations with education. This is the case for adaptation, as the heterogamous combinations in which the man (woman) is the higher educated show a larger difference in the negative (positive) direction (i.e., the men (women) assign less value to adaptation – one inconsistent cell) as compared to the homogamous combinations. Different patterns are visible for negative control and support, suggesting that educational heterogamy, or another factor, is interfering with the educational effects.

The heterogamy hypothesis

When now turning to the DRMs, we can start from the baseline model, to which we add the control variables (equation (1)). The results presented are those based on a categorization of education in four categories (see section on Analysis Technique, aggregating categories 1–2, 3–4, 5, 6–8) as the results based on a categorization in eight categories showed somewhat unstable.

Education appears most important for determining the child-rearing values of adaptation and less important for determining both child-rearing behaviors (Table 3). This is in line with the results of the exploratory analysis, as are the differences in values and behaviors according to education (Table 4). Despite some minor inconsistencies, the estimated values for the homogamous couples (μ ₁₁–μ ₄₄) confirm the positive link with support and the negative link with negative control and adaptation.

Table 3.

R ² values for the baseline model and the model with control variables.

	Negative control		Support		Adaptation
	Men	Women	Men	Women	Men	Women
Baseline model	.045***	.040***	.040***	.035**	.137***	.099***
Model with control variables	.134***	.106***	.069	.086**	.174*	.161***
N	548	553	543	553	618	629

Note. $p_{R^{2} c h a n g e}$ : comparison to the previous model.

*p < .05; **p < .01; ***p < .001.

Table 4.

Parameter estimates for the model with control variables (SE is given within parentheses).

	Negative control		Support		Adaptation
	Men	Women	Men	Women	Men	Women
Salience parameter
p	1.00 (.27)	.54 (.14)	.69 (.26)	.00 (.15)	.78 (.10)	.46 (.09)
Means (μ _ii ) for the homogamous with educational level i
μ₁₁	1.87 (.16)	2.01 (.18)	3.48 (.16)	3.57 (.11)	4.17 (.13)	4.35 (.12)
μ₂₂	1.83 (.09)	1.78 (.09)	3.59 (.08)	3.47 (.07)	4.04 (.06)	3.93 (.06)
μ₃₃	1.47 (.12)	1.67 (.13)	3.77 (.11)	3.79 (.09)	3.64 (.09)	3.66 (.09)
μ₄₄	1.48 (.09)	1.14 (.12)	3.88 (.09)	3.80 (.09)	3.32 (.07)	3.28 (.08)
N	548	553	543	553	618	629

Note. The parameter estimates for the control variables are not shown here. The salience parameter was restricted [0–1] in the analysis.

The link with education is a prerequisite for studying the link with educational heterogamy, as educational differences can be expected important only in so far as education itself is relevant. The DRMs are used to analyze whether the effects of education link educational differences to differences in child-rearing or whether these effects are counteracted by an effect of educational heterogamy. In other words, do heterogamous partners conform to the overall pattern in values and behaviors as expected based on education, or do they deviate by showing smaller differences in values and behaviors within the couple?

In Table 5, results of a comparison of different heterogamy models (see section on Analysis Technique) are shown. The improvement in fit by heterogamy is the largest for negative control. For the other values and behaviors, only the results for men show a significant, though less marked improvement in fit by heterogamy.

Table 5.

R ² values for the model with control variables and the heterogamy models.

	Negative control		Support		Adaptation
Model	Men	Women	Men	Women	Men	Women
Model with control variables	.134	.106	.069	.086	.174	.161
Heterogamy models
Two categories	.140***	.119***	.074***	.087	.175	.161
Three categories	.140**	.119***	.078***	.088	.176	.161
Five categories	.141	.134***	.078	.088	.178	.163
Signed difference	.134	.106	.071	.086	.177**	.162
Absolute difference	.139***	.125***	.074***	.086	.175	.161
N	548	553	543	553	618	629

Note. $p_{R^{2} c h a n g e}$ : comparison to the model with control variables. Values in boldface denote the best-fitting model.

**p < .01; ***p < .001.

For negative control, the best-fitting models indicate a more or less parallel heterogamy effect for men (model with two categories) and women (model with absolute difference). The presence of educational heterogamy appears to influence the reported negative control of both men and women, whereas the size of the educational difference appears to matter only for women. The positive coefficients for heterogamy indicate that men (b _{two categories} =.18, p < .05) and women (b _{absolute difference} =.21, p < .01) report more negative control in heterogamous as compared to homogamous couples. This difference is limited for men (.18) but substantive for women as indicated by the expected score for negative control that is more than half a point higher (3 ×.21 =.63) for women in the most heterogamous couples, as compared to women in homogamous couples. Rather than counteracting the differentiating effect of education, this effect of educational heterogamy points to a situation of elevated negative control for both partners in heterogamous couples.

For men, the best-fitting model for support (model with two categories) parallels the result for negative control. Yet, the heterogamy coefficient for support is not significant in the model (p > .05), meaning that there is no evidence of educational heterogamy counteracting the effect of education on support among either men or women.

Finally, the best-fitting models for adaptation could support the presence of a counteracting heterogamy effect for men, as a positive coefficient for the signed difference would (partly) offset the negative association between education and adaptation. Yet, again the heterogamy coefficient is not significant in the model (p > .10), meaning that the idea of a link between differences in education and adaptation is sustained for both men and women.

Overall, the DRMs confirm the link between educational and child-rearing differences. For support and adaptation, the heterogamy effects did not counteract the differentiating effects of education among heterogamous couples, whereas for negative control, the heterogamy effects indicated that all heterogamous partners report more behaviors of negative control. These results appear to confirm the heterogamy hypothesis, yet they do not take into account the critiques as described in the additional hypotheses. That is, the finding of a link between educational and child-rearing differences does not invalidate the idea of child-rearing differences in homogamous couples nor of interfering mechanisms that diminish the child-rearing differences in heterogamous couples. In other words, the additional hypotheses need to be considered, before formulating definite conclusions regarding the heterogamy hypothesis.

The additional hypotheses

The first two additional hypotheses challenge the heterogamy hypothesis by questioning the similarity of homogamous couples. The gender differences hypothesis is partially confirmed as t tests show significant differences in scores between men and women for negative control (t ₁₀₉₉ = 2.38, p < .05, scores not shown here). In addition, absolute differences between the corrected means for homogamous men and women with educational level i (Table 4; μ ₁₁–μ ₄₄) are on average larger for negative control (.18) than for support (.07) or adaptation (.09). Thus, the hypothesis of child-rearing differences for homogamous partners is partially supported, as the presence of gender differences is confirmed for negative control but less for the other dependent variables.

The gender interaction hypothesis shows less confirmation, as the patterns in values and behaviors by education appear by and large parallel for homogamous men and women (Table 4; μ ₁₁–μ ₄₄). However, small gender differences are present in the explained variation by education (Table 3) with education being more important for men. Thus, gender differences in the link with education are limited to minor differences, primarily in the strength of this link, and probably do not result in major child-rearing differences for homogamous partners.

The last two additional hypotheses challenge the heterogamy hypothesis by questioning the expectation of cultural differences for heterogamous couples. The comparison of zero-order and partial correlations for the selection hypothesis (Table 6) indicates that active selection on values and behaviors is at work. That is, the lack of a clear difference in the correlations after controlling for (educational) heterogamy indicates that the zero-order correlations cannot be ascribed to assortative mating. Rather, selection on values and behaviors is taking place in both homogamous and heterogamous couples. This supports the hypothesis of heterogamous couples showing differences in child-rearing that are smaller than expected based on the link between education and child-rearing values and behaviors, due to processes of selection.

Table 6.

Correlations between the values and behaviors of partners; with partial correlations controlling for (a) educational heterogamy, (b) educational heterogamy and four other forms of heterogamy (for review see section on Analysis Technique).

	Negative control	Support	Adaptation
Zero-order correlation	.42***	.26***	.46***
Partial correlation	(a) .43***	(a) .27***	(a) .46***
	(b) .42***	(b) .26***	(b) .44***

***p < .001.

The results for the convergence hypothesis, finally, fail to show significant correlations between the absolute difference in scores of partners and marriage duration (p > .10). In addition, the regression analyses do not show a significant improvement in fit when including the centred interaction between the analogous individual score of the partner and marriage duration (p $_{R^{2} c h a n g e}$ > .10). Overall, these results do not support the idea of convergence causing cultural differences between partners to be smaller than expected based on the link between education and child-rearing values and behaviors.

The results can now be considered in their totality. The heterogamy hypothesis stated that heterogamous couples, in contrast to homogamous couples, display differences in child-rearing, with the higher educated partner scoring higher on support and lower on negative control and adaptation. The DRMs indicated that educational and child-rearing differences are linked as the models did not show a counteracting effect of educational heterogamy. However, results of the additional hypotheses led to a weakened version of the heterogamy hypothesis as they confirmed that homogamous couples are subject to specific child-rearing differences as well, whereas the degree of child-rearing differences in heterogamous couples is influenced by processes of selection.

Conclusion and discussion

The ubiquity of homogamous relationships and vulnerability of heterogamous relationships are commonly considered two sides of the same coin, as both rely on a central dichotomy: cultural similarity and social support in homogamous relationships and cultural differences and social disapproval in heterogamous relationships. Focusing on the distinction in terms of cultural differences, this dichotomy lacks empirical support. The handful of sociological studies is characterized by important limitations related to the choice and measurement of cultural differences. In addition, the link between heterogamy and cultural differences has been questioned by the findings of the psychology literature, as they have generally not supported the idea of social homogamy being associated with cultural similarity. This study reviewed the distinction in terms of cultural differences, while bearing in mind the problems of previous research. It focused on the specific link between educational heterogamy and cultural differences in the domain of child-rearing and analyzed the individually measured values and behaviors of partners by means of DRMs, a technique that has been shown to be particularly useful for studying heterogamy effects.

The association with education was supported for all studied variables, but the link with educational heterogamy appeared more complex. The additional hypotheses were partly confirmed, as homogamous couples showed specific differences in child-rearing as well, whereas the degree of differences in heterogamous couples appeared influenced by processes of selection. Nevertheless, the DRMs showed a link between educational and child-rearing differences for all studied values and behaviors. In addition, a heterogamy effect was found for negative control that points in the direction of increased negative control for heterogamous partners. Rather than opposing the heterogamy hypothesis, this effect fits the idea of heterogamous couples being faced with negative consequences (i.e., increased negative control) due to their heterogamous nature.

The analyses partly supported the additional hypotheses, in addition to the heterogamy hypothesis. The result is a more complex version of the heterogamy hypothesis, which can be contrasted to the simple idea of cultural differences for heterogamous couples. This points to the need for heterogamy research to specify the exact role of social boundaries and inequalities for heterogamous relationships. On the one hand, heterogamy research should specify the exact link between heterogamy and cultural differences, while bearing in mind the possible critiques that question this link. On the other hand, it should consider which heterogamy type will lead to which specific differences. After all, not all values and behaviors will follow the same logic, as was illustrated by the child-rearing behaviors of negative control. Some differences might even be adaptive, as is suggested by studies regarding authoritative parenting (Fletcher, Steinberg, & Sellers, 1999).

By specifying the expected link, this study showed an association between educational heterogamy and cultural differences in the domain of child-rearing. This association appeared of most importance for the child-rearing value of adaptation, as the large educational effects for adaptation could lead to substantial differences for heterogamous couples. These results were found, despite the high average marriage duration in the sample. This high average marriage duration could have reduced the association between educational and child-rearing differences, because of a selection in terms of more happy (similar?) couples in the sample. Future research should adopt a longitudinal perspective, as such a perspective would provide a more reliable view on the dynamics in terms of cultural differences at different moments in the relationship (e.g., processes of convergence).

In addition, future efforts should be directed towards collecting more up-to-date couple data on (child-rearing) values and behaviors. Due to the limited availability of data regarding these issues, the results of this study are based on data that were collected in 1990. Even though there is no reason to expect the link between education and child-rearing values and behaviors to have changed substantially, or to have become less relevant, recent data would allow for more up-to-date conclusions.

Despite these limitations, the sample had some clear advantages as well. Most importantly, it provided detailed information regarding educational attainment and scale measures of several important child-rearing values and behaviors of both marital partners. Accordingly, DRMs could be used to examine the effects of educational heterogamy on the individual values and behaviors. In comparison to the use of difference scores and conflict/disagreement variables, this dyadic technique offers specific advantages. Most importantly, it opens the way to examining the effect of educational heterogamy over and above the main educational effects. That is, it takes into account the interrelated nature of effects that is characteristic of this type of dyadic questions.

However, the use of an appropriate analysis technique is only part of the answer. In addition, the complexity of the research question needs to be taken into account. This was carried out by considering additional hypotheses and focusing on the specific link between educational heterogamy and cultural differences in the domain of child-rearing. By doing so, it was shown that specification and elaboration of the basic hypothesis can help in clarifying which type of heterogamy is linked to which specific differences and accordingly, in acquiring insight in the possibly wide range of heterogamy consequences.

Footnotes

Appendix 1

Appendix 2

Pearson correlations for the child-rearing values and behaviors of the men (below the diagonal) and women (above the diagonal): correlations between the original six items and the three final variables (N is given within parentheses).

	Punishment	Ignoring	Autonomy	Achievement	Conformity	Adjustment	Negative control	Support	Adaptation
Punishment	1.000	0.513*** (553)	0.029 (553)	0.101* (550)	0.347*** (552)	0.185*** (552)	0.880*** (553)	0.029 (553)	0.259*** (553)
Ignoring	0.528*** (547)	1.000	0.059 (553)	0.126*** (550)	0.174*** (552)	0.117*** (552)	0.860*** (553)	0.059 (553)	0.169*** (553)
Autonomy	0.024 (543)	0.021 (543)	1.000	0.000 (550)	−0.078⁺ (552)	−0.043 (552)	0.050 (553)	1.000*** (553)	−0.054 (553)
Achievement	0.062 (546)	0.082⁺ (545)	0.076⁺ (541)	1.000	0.346*** (549)	0.573*** (550)	0.130*** (550)	0.000 (550)	0.809*** (550)
Conformity	0.336*** (548)	0.251*** (547)	−0.031 (543)	0.387*** (546)	1.000	0.505*** (551)	0.303*** (552)	−0.078⁺ (552)	0.771*** (628)
Adjustment	0.189*** (546)	0.181*** (545)	−0.093* (541)	0.511*** (546)	0.532*** (546)	1.000	0.175*** (552)	−0.043 (552)	0.849*** (552)
Negative control	0.884*** (548)	0.863*** (547)	0.026 (543)	0.081⁺ (564)	0.337*** (548)	0.212*** (546)	1.000	0.050 (553)	0.248*** (553)
Support	0.024 (543)	0.021 (543)	1.000*** (543)	0.076⁺ (541)	−0.031 (543)	−0.093* (541)	0.026 (543)	1.000	−0.054 (553)
Adaptation	0.247*** (548)	0.208*** (547)	−0.016 (543)	0.783*** (546)	0.800*** (618)	0.844*** (546)	0.261*** (548)	−0.016 (543)	1.000

⁺ p < .10; *p < .05; ***p < .001.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Alwin

D. F.

(1984). Trends in parental socialization values: Detroit, 1958–1983. American Journal of Sociology, 90, 359–382.

Baril

M. E.

Crouter

A. C.

McHale

S. M.

(2007). Processes linking adolescent well-being, marital love, and coparenting. Journal of Family Psychology, 21, 645–654.

Berger

P. L.

Kellner

(1970). Marriage and the construction of reality. In Dreitzel

(Ed.), Recent sociology: Patterns of communicative behavior (no. 2) (pp. 49–73). New York, NY: MacMillan.

Beutel

A. M.

Marini

M. M.

(1995). Gender and values. American Sociological Review, 60, 436–448.

Block

J. H.

Block

Morrison

(1981). Parental agreement-disagreement on child-rearing orientations and gender-related personality correlates in children. Child Development, 52, 965–974.

Booth

Welch

(1978). Spousal consensus and its correlates: A reassessment. Journal of Marriage and the Family, 40, 23–32.

Bozon

Heran

(1989). Finding a spouse: A survey of how french couples meet. Population: An English Selection, 44, 91–121.

Chen

Luo

Yue

Zhaoyang

(2009). Do birds of a feather flock together in China? Personal Relationships, 16, 167–186.

de Graaf

N. D.

(1991). Distinction by consumption in Czechoslovakia, Hungary, and the Netherlands. European Sociological Review, 7, 267–290.

10.

Dykstra

P. A.

Kalmijn

Knijn

T. C. M.

Komter

A. E.

Liefbroer

A. C.

Mulder

C. H.

(2005). Codebook of the Netherlands kinship panel study, wave 1. NKPS Working Paper No. 4. The Hague (The Netherlands): NIDI.

11.

Edwards

J. R.

(2002). Alternatives to difference scores. Polynomial regression analysis and response surface methodology. In Drasgow

Schmitt

N. W.

(Eds.), Advances in measurement and data analysis (pp. 350–400). San Francisco, CA: Jossey-Bass.

12.

Eeckhaut

M. C. W.

Van de Putte

Gerris

J. R. M.

Vermulst

A. A.

(2013). Analysing the effect of educational differences between partners: A methodological/theoretical comparison. European Sociological Review, 29, 60–73.

13.

Eurostat. (2003). Education across Europe 2003. Luxembourg (Luxembourg): Office for Official Publications of the European Communities.

14.

Feinberg

M. E.

Kan

M. L.

Hetherington

E. M.

(2007). The longitudinal influence of coparenting conflict on parental negativity and adolescent maladjustment. Journal of Marriage and Family, 69, 687–702.

15.

Fletcher

A. C.

Steinberg

Sellers

E. B.

(1999). Adolescents’ well-being as a function of perceived interparental consistency. Journal of Marriage and the Family, 61, 599–610.

16.

Gerris

J. R. M.

van Boxtel

D. A. A. M.

Vermulst

A. A.

Janssens

J. M. A. M.

van Zutphen

R. A. H.

Felling

A. J. A.

(1992). Child-rearing and family in the Netherlands: Documentation of a national representative survey of child-rearing, family relations and family processes in 1990. Nijmegen, The Netherlands: University of Nijmegen, Institute of Family Studies.

17.

Gerris

J. R. M.

Vermulst

A. A.

van Boxtel

D. A. A. M.

Janssens

J. M. A. M.

van Zutphen

R. A. H.

Felling

A. J. A.

(1993). Parenting in Dutch families. Nijmegen, The Netherlands: University of Nijmegen, Institute of Family Studies.

18.

Harris

J. A.

(1912). Assortative mating in man. Popular Science Monthly, 80, 476–492.

19.

Harvey

E. A.

(2000). Parenting similarity and children with attention-deficit/hyperactivity disorder. Child & Family Behavior Therapy, 22, 39–54.

20.

Heaton

T. B.

(2002). Factors contributing to increasing marital stability in the U.S. Journal of Family Issues, 23, 392–409.

21.

Hohmann-Marriott

B. E.

Amato

(2008). Relationship quality in interethnic marriages and cohabitations. Social Forces, 87, 825–855.

22.

Hou

Myles

(2008). The changing role of education in the marriage market: Assortative marriage in Canada and the United States since the 1970s. Canadian Journal of Sociology –Cahiers Canadiens de Sociologie, 33, 337–366.

23.

Houts

R. M.

Robins

Huston

T. L.

(1996). Compatibility and the development of premarital relationships. Journal of Marriage and the Family, 58, 7–20.

24.

Inkeles

(1960). Industrial man: The relation of status to experience, perception, and value. American Journal of Sociology, 66, 1–31.

25.

Janssen

J. P. G.

(2001). Do opposites attract divorce? Dimensions of mixed marriage and the risk of divorce in the Netherlands. Nijmegen, The Netherlands: Katholieke Universiteit Nijmegen.

26.

Jones

F. L.

(1996). Convergence and divergence in ethnic divorce patterns: A research note. Journal of Marriage and the Family, 58, 213–218.

27.

Jorgensen

S. R.

Klein

D. M.

(1979). Sociocultural heterogamy, dissensus, and conflict in marriage. Pacific Sociological Review, 22, 51–75.

28.

Kalmijn

(1998). Intermarriage and homogamy: Causes, patterns, trends. Annual Review of Sociology, 24, 395–421.

29.

Kalmijn

(2005). Attitude alignment in marriage and cohabitation: The case of sex-role attitudes. Personal Relationships, 12, 521–535.

30.

Kalmijn

de Graaf

P. M.

Janssen

J. P. G.

(2005). Intermarriage and the risk of divorce in the Netherlands: The effects of differences in religion and in nationality, 1974–94. Population Studies, 59, 71–85.

31.

Kohn

M. L.

(1959a). Social class and parental values. American Journal of Sociology, 64, 337–351.

32.

Kohn

M. L.

(1959b). Social class and the exercise of parental authority. American Sociological Review, 24, 352–366.

33.

Kohn

M. L.

(1976). Social class and parental values: Another confirmation of the relationship. American Sociological Review, 41, 538–545.

34.

Kohn

M. L.

Schooler

. (1969). Class, occupation, and orientation. American Sociological Review, 34, 659–678.

35.

Lörincz

(2006). Rules of attraction: How do people select partners? Review of Sociology, 12, 71–84.

36.

Luo

S. H.

Klohnen

E. C.

(2005). Assortative mating and marital quality in newlyweds: A couple-centered approach. Journal of Personality and Social Psychology, 88, 304–326.

37.

O’Brien

Peyton

(2002). Parenting attitudes and marital intimacy: A longitudinal analysis. Journal of Family Psychology, 16, 118–127.

38.

Prince-Gibson

Schwartz

S. H.

(1998). Value priorities and gender. Social Psychology Quarterly, 61, 49–67.

39.

Roest

A. M. C.

Dubas

J. S.

Gerris

J. R. M.

Engels

R. C. M. E.

(2006). Disentangling value similarities and transmissions in established marriages: A cross-lagged longitudinal study. Journal of Marriage and the Family, 68, 1132–1146.

40.

Schwartz

S. H.

Rubel

(2005). Sex differences in value priorities: Cross-cultural and multimethod studies. Journal of Personality and Social Psychology, 89, 1010–1028.

41.

Slomczynski

K. M.

Miller

Kohn

M. L.

(1981). Stratification, work, and values: A Polish-United States comparison. American Sociological Review, 46, 720–744.

42.

Sobel

M. E.

(1981). Diagonal mobility models: A substantively motivated class of designs for the analysis of mobility effects. American Sociological Review, 46, 893–906.

43.

Spade

J. Z.

(1991). Occupational structure and men’s and women’s parental values. Journal of Family Issues, 12, 343–360.

44.

Stright

A. D.

Bales

S. S.

(2003). Coparenting quality: Contributions of child and parent characteristics. Family Relations, 52, 232–240.

45.

Tolsma

de Graaf

N. D.

Quillian

(2009). Does intergenerational social mobility affect antagonistic attitudes towards ethnic minorities? The British Journal of Sociology, 60, 257–277.

46.

Van der Slik

F. W. P.

de Graaf

N. D.

Gerris

J. R. M.

(2002). Conformity to parental rules: Asymmetric influences of father’s and mother’s levels of education. European Sociological Review, 18, 489–502.

47.

Van Es

J. C.

Shingi

P. M.

(1972). Response consistency of husband and wife for selected attitudinal items. Journal of Marriage and the Family, 34, 741–749.

48.

Verdonk

Harting

A. J.

Lagro-Janssen

T. L. M.

(2007). Does equal education generate equal attitudes? Gender differences in medical students' attitudes toward the ideal physician. Teaching and Learning in Medicine, 19, 9–13.

49.

Watson

Klohnen

E. C.

Casillas

Simms

E. N.

Haig

Berry

D. S.

(2004). Match makers and deal breakers: Analyses of assortative mating in newlywed couples. Journal of Personality, 72, 1029–1068.