Aggregation aggravation

A 70-year study of 500 juvenile offenders born in the Twenties — the longest-running crime study in the world — has found that those who married were far more likely to go straight later in life than those who remained single.
Melanie Phillips, The Daily Mail, 16 January 2004

Melanie Phillips is one of the best-known socially conservative commentators in Britain. Like many or her ilk, she sees the evidence that “marriage works” as incontrovertible. But although there is no reason to doubt the accuracy of the statistics she cites, her case is flawed because it fails to take into account the effects of what statisticians call disaggregation: the breaking down of statistics into their component parts.

Statistics do indeed regularly indicate a correlations between marriage and various goods, both social and personal. Similarly, divorce and single-parenthood are correlated with various ills. Phillips is quick to highlight these. “Married men were more likely have a full-time job,” she writes about another survey, and “They were less like to use drugs or abuse alcohol, and less likely either to commit crime or become its victims.”

Of course, one point that should always be made about such surveys is that a correlation is not the same thing as a cause. (See previous Bad Move, Correlation/cause confusion) But in this case there is another major problem with interpreting the data. The trouble is that such surveys deal only with very broad categories: the married, cohabiters, divorcees, single parents. But what would happen if we broke down these categories into smaller component parts? What would we see then?

Most obviously, we would notice that cohabitation covers a wide spectrum of arrangements, from people living together with no commitment to the future, right through to those who have spent a lifetime together without exchanging wedding vows. If we looked at the married, we would also find – as well as the strong, stable relationships Phillips values – domestic disasters which are bad for both spouses.

All of this would be consistent with a hypothesis contrary to that espoused by Phillips: that it is not marriage per se which is good for people, but long term committed relationships. We would expect to find more of these kinds of relationships in the married group than the cohabiting group, simply because all marriages are at least in theory long-term commitments, whereas cohabitations need not be. It is therefore more than possible that marriage is not a prerequisite for the social benefits Phillips values after all, but is merely the most common social sign that the commitment required to achieve these values is present in a relationship.

That is just one possible finding breaking down the statistics more carefully might suggest. But there could be others. For instance, take the study on juvenile offenders Phillips cites. Again, we might find that those who stayed straight were those who found long-term relationships, whether or not they got married. Or we might find that that the tendency to get married is a result of some other factor which was more important in preventing re-offending.

Another problem with the statistics is that the divorced were, of course, once married. Those who fall into the married category are by definition those whose relationships are, for the time being at least, holding up. So in determining whether marriage is a good thing or not, you really need to look at how those who are still married and those who have divorced fair as a whole, compared to the never married. And if you want to argue that the divorced would be better off if they had stayed together, the statistics cannot help you make your case, because they cannot tell you what counterfactually would have happened if the separated had remained together.

If you are serious about using statistics as part of your case for the benefits or otherwise of marriage, it is no good simply pointing to correlations between marriage and social or personal benefits. You need to dig deeper and see whether what is truly correlated with happy adults and children isn’t something more fundamental than whether or not people choose to march down the aisle. In and of themselves, the bold statistics showing the married and their children to be better off do not show that marriage works.

Comments are closed.