![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
ABSTRACT
Research on socioeconomic homogamy was developed by stratification researchers who used marriage patterns to describe how open stratification systems are. In cross-national studies primary concern on marriage homogamy lies in examination of commonality and differences in their social structures. Following large-scale international studies we use the European Social Survey data 2004–2005 to examine the association between spouses’ educational levels. Loglinear analysis is applied to assess: (i) degree of association between education of spouses, (ii) patterns of barriers to intermarriage, (iii) variation in homogamy for partners with the same education for primary, uncompleted secondary, secondary, and university levels, and (iv) asymmetry in marriage patterns between women and men. The strongest association between spouses’ education is in Slovakia, followed by Czech Republic, Norway, Germany, Ukraine, Poland, Hungary, Estonia, and Slowenia, whereas the lowest association displays in Luxembourg, France, Sweden, Finland and Belgium. In addition to previous research we found inter-country variation in division into post-communist and Western democracies. In line with all earlier studies we found – upon examination of parameters estimated for educational levels – a uniform tendency according to which the difficulties of intermarriage varies monotonically with differences between educational level of spouses. The tendency toward in-marriage proved to be the strongest in the lowest educational levels – such pattern takes place in the 14 countries. Finally, our analysis substantiated presence of net tendency to ‘marry up’ higher educated husbands by women but we find that it is by no means an universal rule and in seven, out of 22 countries examined, it is men who ‘marry up’ higher educated wives.
Notes
1For illustration we report percentages of, respectively, married and cohabiting respondents in the ESS samples: Austria (51.5, 7.3), Belgium (51.5,11.2), Czech Republic (56.4, 6.3), Denmark (53.7,13.4), Estonia (42.6,10.9), Finland (49.9, 13.9), France (57.3, 12.9), Germany (54.1, 8.5), Greece (65.8, 1.5), Hungary (53.3, 6.2), Iceland (49.9, 17.6), Ireland (57.4, 2.7), Luxembourg (49.4, 10.5), The Netherlands (60.8, 9.1), Norway (52.8, 13.9), Poland (57.5, 2.3), Portugal (60.5, 2.5), Slovakia (55.5, 3.2), Slovenia (52.2, 7.1), Spain (57.3, 4.3), Sweden (42.8, 20.9), Switzerland (58.2, 7.7), Ukraine (61.9, .7), the United Kingdom (51.2,10.4).
2Nevertheless, even this reduced classification does not capture distinctions between educational levels in some countries in valid way. In samples for Czech Republic, Denmark, Germany, Iceland, Norway, Slovakia and Switzerland category with primary education was relatively small that could influence loglinear parameters. To test whether educational homogamy was affected by our classification we repeat analysis for other, modified (four-categorical) education variables. The differences between fits and parameters of loglinear models are small and do not change our conclusions. More detailed information on these results are available on request from authors.
3It is equal to half of the sum of differences between the percentages of men and women in the same educational strata with value range from 0 (perfect match) to 100 (perfect incongruity).
4We adjusted the original data to ensure comparable sample sizes across 22 tables. As noted by Smits et al. (Citation1998), estimation results based on unadjusted data will be disproportionatelly influenced by spouse-pairing patterns in the tables with the largest number of cases. To adjust sample sizes we shrank all entries by arbitrary constant required to generate an overall table size of 1,000. All models are estimated using LEM software package (Vermunt Citation1997).
5BIC = L2−(df)*ln(N), where N is the number of cases. The BIC statistic is an index that adjusts the L2 for sample size. The more negative value of BIC, the better fit of the model to data (Raftery Citation1999).
6The traditional statistical criteria for model selection (the statistical significance indicated by p) cannot be used here because the total number of cases in our table is so large that it is almost impossible to find a model that does not differ significantly from the saturated model.
7Our model seems to be more realistic than that adopted by Smits et al. in that we allow main diagonal to vary across countries. So, we use parameters Q2*C in our model instead only Q2. Note also that we used multiplicative version of the fixed-distance model whereas Smits et al. (Citation1998) applied it in an additive logarithmic form. It means that parameters estimated for our model (Table ) have to be transformed to make them directly comparable with those of Smits et al.
8Sources for the GDP is the United Nations Common Database/World Bank (http://globalis.gvu.unu.edu). We could not use this data for earlier years because they were unavailable for that time for post-communist societies.
9However, it may well be that educational gradient holds also in Czech Republic. The numerical size of the lowest level of education is very small that might inflate value of parameter for this category.
10In Slovakia values of diagonal parameters for both the lowest and highest categories appear extremely high. It seem to result from its relatively small size that can inflate values of these parameters. We tested whether our results were influenced by under-representation of the lowest educational category by estimation of these parameters in samples restricted to 25–55 age range and using modified classification of the four educational levels. These analyses confirmed that in Slovakia homogamy in lowest level is the strongest – and in the highest level is next to it – although parameter values for them were lower that those presented in Table .