A divorce paper

On the Variation of Divorce Risks in Europe: Findings from a Meta-Analysis of European Longitudinal Studies:

“The aim of this article is to integrate empirical research on divorce risks in Europe and to explain the variation of empirical findings between European countries by the different levels of modernization and differences in the strength of marriage norms. We focus on the effects of premarital cohabitation, the presence of children, and the experience with parental divorce on marital stability. More than 260 studies on divorce risks could be identified, and 120 were used for further meta-analytical examinations. We show that there is considerable heterogeneity of divorce risks within as well as between countries. Explaining the variation of effect sizes between European countries, it could be shown that in countries where more rigid marriage norms prevail cohabitation has a stronger effect on marital stability than in countries where marriage norms are weaker. Furthermore, the lower the divorce barriers are, the weaker is the association between the parental divorce and the divorce risk of the offspring.”

Some data and results from the paper (click tables and figures to see them in a higher resolution):

Table 1

The table shows the estimated effect sizes of premarital cohabitation on the divorce risk in various European countries; a positive effect size indicates a higher likelihood of divorce among couples who lived together before they got married, whereas a negative effect size indicates a smaller divorce risk for couples who did not cohabitate before they got married. They note in the paper that, “The European overall effect indicates a positive relationship between cohabitation and the risk of divorce, that is, cohabiting couples have a 33 per cent higher risk to divorce than couples who do not share a common household before marriage.” However the effecs are highly heterogenous across countries, and more specifically they find that: “In countries in which traditional marriage norms are strongly institutionalized, cohabitation has a stronger effect than in countries in which marriage norms are weaker.” The institutional framework is important. The Q-statistic is a heterogeneity-measure – read the paper if you want the details..

What about children? Here’s a brief summary:


Effect sizes are almost universally negative (children = smaller risk of divorce) and a lot of them are highly significant (more than half of them are significant at the 1% confidence level). As they note, “The presence of children strongly decreases the risk of divorce”. Note that the effect sizes vary but tend to be large; in the Netherlands, the country with the largest effect size, married couples with children are 70% less likely to divorce than are couples without children. The average estimated effect size is 50% so this is a huge effect. However I would be cautious about making a lot of inferences based on this finding without at the very least having a closer look at the studies on which these results are based; for example it’s unclear if they have taken into account that there may be unobserved heterogeneity problems playing a role when comparing married couples with- and without children here; lots of marriages break up early on (using Danish data I have previously estimated that once the marriage has lasted 9 years, half of the total divorce risk the Danish couple confronted ex ante will basically have been accounted for; i.e. the total risk that you’ll divorce your partner during the first 9 years is as big as is the risk that you’ll do it at any point after the 9th year of marriage – see the last figure in this post), and it does not seem unlikely e.g. that sampled marriages involving children may, ceteris paribus, have lasted a longer time on average than have sampled marriages without children (most European couples get married before they have children so the likelihood that a couple will have children is positively correlated with the marriage duration), meaning that these marriages were less likely to get broken up, regardless of the children. If they conditioned on marriage duration when calculating these effects this particular problem is dealt with, but I don’t know if they did that (and I’m not going to go through all those studies in order to find out..) and there may be a lot of other ways in which marriages with and without children differ; differences that may also relate to divorce probability (education, income, labour market status, …). Note that the fact that the studies included in the meta-study are longitudinal studies does not on its own solve the potential ‘duration problem’ (/selection problem); you can easily follow two couples for the same amount of time and still have radically different (ex ante) divorce likelihoods – and comparing unadjusted (group?) hazard rates and making conclusions based on those seems problematic if you have selection issues like these. Researchers aren’t stupid, so the studies here may all have taken care of this particular potential problem. But I’m sure there are problems they haven’t handled. Caution is warranted – part of the estimated ‘children effect’ is likely not to go through the children at all.

How about the parents? How does the fact that your parents got divorced impact your own likelihood of divorce?

Parents divorce

“Nearly all the reported effect sizes indicate positive associations between the stability of the parental marriage and the stability of children’s marriage”. There are huge cross-country differences – in Italy an individual whose parents got divorced is almost three times as likely to get divorced him/herself as is an individual whose parents did not divorce, whereas the risk increase in Poland amounts to only (a statistically insignificant) 14%.

Lastly, I’ll note that:

“No empirical support was found for any of our hypotheses which link the level of modernization to the risk of divorce. A least with respect to the divorce risk, we considered the level of socioeconomic development not to be an important macro-variable. Also, we could not find any significant relationships between the strength of divorce barriers and the effect of children on marital stability.”

I would not have expected these results if you’d asked me beforehand. Then again e.g. the differences in socioeconomic development among the countries included here are not that big, so it may just be a power issue.


October 25, 2013 - Posted by | Data, Demographics, marriage, Studies


  1. That cohabitation increases divorce chances was mentioned in the Psychology in modern life textbook that lukeprog promoted on LessWrong some years ago. It surprised me as well.

    Comment by Emil | November 5, 2013 | Reply

    • “divorce chances”? Really? That’s a funny way of putting it, I’d say… (

      With nothing but this review to go on (I haven’t read Lukeprog’s stuff), I’m not convinced what to make of that finding. What I’m sure of is that it’s not the whole story. Maybe parental disapproval covary with premarital cohabitation in socieities with ‘traditional marriage norms’, and parental disapproval (of the woman especially – remember that a great majority of divorces are initiated by women) is the trigger/driver because it leads to relationship-related stress. Stuff like that. There’s likely to be a lot of stuff ‘playing a role behind the scenes’. It’s clear from the analysis that the cultural setups matter but note also that although the effect is (insignificantly) negative in Norway it’s highly positive and very significant in Sweden. I don’t have a good explanation for that, and I don’t think the authors do either; it’s obvious that something semi-weird is going on here, those countries are quite similar in terms of norms, institutions and so on, and I don’t think it’s easy to come up with a good reason why premarital cohabitation should have such radically different effects in the two countries. Measurement error? Biased estimates? I don’t know what’s going on, but it’s weird.

      Comment by US | November 6, 2013 | Reply

      • I think Emil used ‘chance’ and ‘probability’ synonymously.

        Comment by Miao | November 7, 2013

      • Presumably. And I doubt he’s the only one doing it, because I’ve seen that kind of usage before. But everytime someone uses these words (risk, chance) that way, it gets harder to use them the way they are meant to be used; to distinguish between probabilities of positive outcomes and probabilities of negative outcomes. We already have multiple terms for ‘neutral probabilities’ (probability, likelihood).

        Comment by US | November 7, 2013

      • ‘Probability’ and ‘likelihood’ mean very different things (though I am sure you already know that), and I have seen people using them interchangeably as well.

        Comment by Miao | November 7, 2013

      • I think using those terms interchangeably is much less problematic than is the risk/chance muddle, as most people probably don’t have any real need to retain the ability to distinguish those concepts; unless you’re a statistician or a mathematician the distinction between the two is probably not important (‘academic’), and the people who are familiar with the distinction and the ‘proper usage’ probably won’t care much about how people who don’t talk about those things. (Indeed they may also not care much).

        Comment by US | November 7, 2013

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: