Divorce and Education in Croatia: Evidence of a Changing Relationship from a Couples’ Perspective

Petra Međimurec, University of Zagreb, Faculty of Economics and Business, Department of Demography
Ivan Čipin, University of Zagreb, Faculty of Economics and Business

This paper adds to a broad literature on educational differentials in divorce by examining data from Croatia. Our contribution to previous research is manifold. First, we investigate divorce from a couples’ perspective and include both husbands’ and wives’ characteristics into our analyses. Existing studies are often limited to female education only, and hence fail to account for marriage partners’ joint status. Second, we test whether the relationship between education and divorce varies across marriage cohorts. In many countries, marriages are becoming increasingly stable for highly educated spouses. However, studies that empirically evaluate temporal patterns in divorce by spouses’ relative education are scarce and mostly refer to the US. Third, our data link individual vital statistics records on marriage and divorce from 1989 to 2014. Similar data have rarely been used to study divorce in European countries. These data conveniently allow us to estimate the effects of education on divorce during a period of marked improvements in female education, which are reflected also in a growing share of hypogamous marriages and a diminishing share of hypergamous marriages. We restrict our analyses to first marriages (n > 500,000) and control for a limited set of additional variables. These include calendar time, as Croatia underwent profound social and economic change over the years considered. Using event history methods, we find a large shift in the influence of women’s education on divorce – the once positive gradient has eventually turned negative. For men, the negative influence of education on divorce has widened over time. Couples in which husbands hold less education than their wives appear to be more likely to divorce relative to other couples, but our results suggest that this association has weakened or even reversed among newer marriage cohorts.

Inmany countries, divorce has become more common among the lower-educated (e.g.Härkönen and Dronkers, 2006; Matysiak et al., 2014); this can strengthen its –on average – negative outcomes (McLanahan, 2004; McLanahan and Percheski,2008). Education is a widely studied predictor of divorce. Researchersrecognize its manifold meaning: education is an economic indicator because itcorrelates to earnings potential and labour market activity, but education is alsoan attitudinal indicator because it reflects values, beliefs, life-styles, andcultural backgrounds. Theory provides opposing arguments on how educationrelates to marital (in)stability. Empirical studies report mixed evidence aswell – unlike most predictors of divorce, education shows effects that varybetween countries and over time. Nonetheless, a growing body of literatureidentifies similar trends in the relationship between education and divorce (Matysiaket al., 2014; Härkönen and Boertien, 2014). This study explores whether Croatiafollows the same pattern of change: are the lower-educated exhibiting thefastest rise in divorce?

Divorce is a two-personevent, and the ability to observe the marital couple as a unit is a keyadvantage of this study. If – as theory predicts – the husband’s and the wife’s(socio-economic) attributes jointly produce the marital couple’s risk ofdivorce, then it is preferable for empirical models to take in each partner’seducational resources (Jalovaara, 2003; Lyngstad, 2004). The lack of detailedenough data unfortunately prevents many researchers from doing so (e.g.Härkönen and Dronkers, 2006; Salvini and Vignoli, 2009). Empirical models inthis study offer valuable insights on the relationship between education and divorcefor at least two reasons:

      i.         theyestimate how both spouses’ educational resources – in absoluteand relative terms – modify the risk of divorce in Croatia, and

     ii.         theytest for change in the effects of both spouses’ educational resources –yet again, in absolute and relative terms – on the risk of divorce acrossmarriage cohorts.

Despite theoreticalefforts that have linked the weakening (or, in most instances, the reversal) ofthe gender gap in education to family outcomes (e.g. Van Bavel, 2012; Esteve etal., 2016), empirical work on temporal change in the effects of spouses’ relativeeducation on the risk of divorce remains scarce (and this is distinctly so forEuropean countries; see, however, an American study by Schwartz and Han, 2014).

Divorce in Croatia has notbeen widely discussed. A large majority of existing knowledge aboutsocio-economic determinants of divorce refers to northern-European andwestern-European countries, and to the United States (Lyngstad and Jalovaara,2010; Härkönen, 2014). This study considers a different setting. Croatia is anex-Yugoslav country with a relatively unfavourable public attitude to divorce,getting married and having children still come high on the list of lifepriorities, and the overall prevalence of less traditional family practices israther low (EVS, 2016). As a further matter, Croatia underwent major economic,political, and social turmoil during the 1990s, when there began a rapidtransition to a market-oriented system. The far-reaching implications of suchprofound change extend to family behaviour (Thornton and Philipov, 2009), andthis makes temporal dimensions of divorce in Croatia very interesting from ademographic point of view. Applying a life course approach, this studyappropriately accounts for the role of period and marriage-cohort effects.

In order to conduct theanalysis, this study has merged couple-level vital statistics records onmarriage and divorce into a unique database. The database includes informationon all couples who got married between 1989 and 2014: it picks up couples onthe date of their marriage and follows them until their marriage is dissolved(or until the observation is censored). In other words, the database joinsdivorce records to a full set of marriage records spanning more than two and ahalf decades. These are seldom available data, and only a handful of studieshave used a similar input to explore educational gradients in divorce (see Parkand Raymo, 2013).

Our focus is on firstmarriages, and we estimate event history models to assess the education-divorcenexus in Croatia. The results point to several conclusions; selected findingsare briefly presented below.

Among earlier marriagecohorts, higher-educated women were more likely to divorce, but the positiveinfluence of women’s education declined, and eventually turned negative. Formen, the influence of education on divorce grew stronger – more negative withtime.


Theheightened risk of divorce for hypogamous marriages diminished; there areindications of a cross-over pattern in the influence of relative education ondivorce.

Ourpaper discusses viable explanations. A feasible extension is an analysis ofeducational gradients in divorce (and of possible change therein) separatelyfor homogamous, hypergamous, and hypogamous marriages.

* Thefinal paper will include a full list of references (the extended abstract omitsit due to a word-count limit).

Presented in Session 1110: Families and Households