The Negative Female Educational Gradient of Divorce: Towards an Explanation in Five European Countries

Pearl Dykstra, Erasmus University Rotterdam
Maike van Damme, University of Cologne

How can we explain a negative female educational gradient of divorce? In this paper, we continue research from Boertien and Härkönen (2014, 2018) for the UK and Raymo, et al. (2013) for Japan on the mechanisms explaining such a gradient. We formulate hypotheses based on Levinger’s (1965, 1976) social exchange theory on ‘attractions’ and ‘barriers’ and assess whether there are mediating effects of affectional rewards (relationship satisfaction), economic rewards (family income, durables possession, his unemployment), symbolic rewards (age and educational similarity between the spouses, his education, his social status, and communication or conflict resolution skills), affectional barriers (having young children), material barriers (her social status, her unemployment, and being a home owner), and symbolic costs (being committed to marriage, having broken up before or having parents that broke up, being religious, going to church frequently, and living in a rural area).

We analyse the Generations and Gender Survey (GGS) [2004-2013] for two waves for Bulgaria, Russia, France, Austria, and Czech Republic. Using the khb-approach with lagged independent variables (from wave 1), we examine the probability that women broke up between the two consecutive waves and perform a mediation analyses to explain the female educational gradient of union dissolution.

Indeed, we find a negative female educational gradient for the pooled country fixed effects model. Instead of.being explained by ‘attractions’, ‘barriers’ were explaining the negative educational gradient of union dissolution in the five countries we studied. We found suppressor effects of ‘attractions’.

A research gap still exists in how we can explainthe negative female educational gradient of divorce in European countries.Micro-level explanations of educational differences have rarely been studied(but see Boertien and Härkönen (2014) for the UK andRaymo, et al. (2013) for Japan). Weinvestigate educational differences more detailed and not only consider women’seducation (actor effects), but also her husband’s education (partner effects)and dissimilarity effects. We formulate hypotheses based on Levinger’s (1965,1976) social exchange theory on ‘attractions’ and ‘barriers’. Weassess effects of relationship satisfaction as affectional rewards, familyincome and durables possession as economic rewards, aggregated gender roledifferences between men and women, age and educational similarity between thespouses, his education, and his social status as symbolic rewards,communication or conflict resolution skills, having young children asaffectional barrier, her social status and being a home owner as materialbarriers, and being committed to marriage, having broken up before or havingparents that broke up, being religious, going to church frequently, and livingin a rural area as indicators of symbolic costs.

We use the Generations and Gender Survey(GGS) [2004-2013] for two waves for Bulgaria, France, Austria, and Czech Republic.Using the khb-approach with lagged independent variables (from wave 1), weanalyse the probability that women broke up between the two consecutive waves andperform a mediation analyses to explain the female educational gradient ofunion dissolution.



Thehigher her education is, the higher the symbolic rewards (his education andeducational similarity) and the less likely she will break-up (H1).Her higher education is expected to lead to a lower break-up rate dueto the higher (symbolic, material, and affectional) attractions to staytogether (H2). We expect that her higher education is related to higherbreak-up rates because she has lower barriers (less material and symbolic costs)to disrupt the relationship (H3).

Weindeed find a negative educational gradient. Both the difference between hermid education and low education and the difference between her high educationand low education are significantly negative. His education slightly explainsthe relationship between her education and break up. Also, her education seemsto be a proxy for educational dissimilarity in the relationship. When she ishaving a high education, she is much more likely to be in a female hypergamousrelationship, and such female hypergamous couples in turn apparently have a lowerbreak up risk than homogamous couples (indirect effect is negative). Malehypergamy, on the contrary, suppresses the negative female educational gradientby about half to 60%. This means that if we could take into account that herhigher education often does not coincide with traditionally specializedcouples, the stability of higher educated couples would have been even greater.Unfortunately, we cannot take simultaneously his education and educationaldissimilarity into account. We therefore continue with his education only.

Fromall the mediating and confounding variables included in the model, only fewexplain the negative educational gradient. Her lower occupational statusexplains the largest part (16% of mid-low difference and 37% of high-low difference).If she is higher educated, she is less likely to have a lower status and havinga lower status is positively related to the separation risk. Thus, not economicindependence matters, but rather the fact that women belonging to the lowereconomic strata have less to lose. His social status suppresses therelationship between female education and union dissolution. His educationallevel hardly explains anything anymore. Note that family income also suppressesthe negative educational gradient (those with a higher family income turn outto be more likely to break up, rather than less). Note also that the onlyindirect effects that are significant are that of family income and commitment.

Overall,‘attractions’ suppress -27% of the total educational difference between highand low education in break up rate, while ‘barriers’ explain 54% of the totaleducational effect. This is not in line with our hypotheses where we expectedthat attractions would explain, but not suppress the negative educationalgradient and barrier variables would be suppressors. It turns out to be theother way around. A similar pattern is observed for the union dissolutiondifference between her middle education and low education.

Notethat we also ran models only including variables that contributed at least |3%|to the total effect of education on relationship break up. The observed patternremains the same. The explained percentages change somewhat, but her socialstatus remains the most important mediator.

[notdiscussed: country-specific effects]


Tables.Khb-mediation analyses. Country fixed effects weighted pooled model four countries

Explainingeducational difference between middle and low education in separation risk.


Explaining theeducational difference between high and low education in separation risk.

Presented in Session 1110: Families and Households