So, given their track record, it was no great surprise to read another negative piece on marriage in yesterday’s Guardian by an Australian academic who seemed utterly mystified that marriage could possibly matter. ‘Does marriage matter for children?’ asked the author. No! The biggest Australian study of children agrees that married families appear to do better. But it’s not marriage or having two parents that make the difference (e.g. p37 of the report). Its the higher income and better parenting that you tend to find more often among two parent compared to one parent families, and among married compared to unmarried families.
Now even if I perceive that the Guardian has a blindspot about marriage, I need to take note when professors and government bodies produce credible research. What am I to make of their conclusion and findings? Can we now close down Marriage Foundation because our coffin has been well and truly nailed?
Not so fast.
The usual scientific method for finding out ‘whether‘ something makes a difference is to conduct what’s called a Randomised Controlled Trial or RCT. You find a group of appropriate people, randomly assign half of them to do that something and half not, then see who does best.
With marriage, clearly this is not going to happen! You can’t exactly turn round to people and say ‘right, you lot get married and you lot don’t‘. So the usual way to test ‘whether‘ marriage matters is to look at who does best and try to take into account as many background differences as you can – such as income or parenting style – using clever statistics. On the face of it, married families tend to do better in all sorts of ways. Almost all researchers agree that some of this is down to the kind of people who marry – known as the selection effect. If there’s still a difference left over after removing these selection factors, it’s either down to the act of marriage itself or to some other unobserved variable more commonly found among married couples. There’s a sensible debate to be had about precisely ‘how‘ marriage might make a difference. That’s why I love the work of Professors Scott Stanley and Galena Rhoades in the US on ‘sliding, deciding and inertia‘. Their theory of commitment specifically addresses the ‘how‘ of marriage and cohabitation.
In my view, where the debate goes badly wrong is in focusing exclusively on the ‘whether‘ and ignoring the ‘how‘. There are two main errors.
The first error is to forget that if these different family types do actually change people, they are likely to do it by changing behaviour. So its no good discovering that the behaviours make a unique difference but being married doesn’t. Take away the ingredients and its not a surprise that there’s no cake left over.
- This is the fundamental flaw in the Australian studies quoted in the Guardian. It’s fair enough to think of income as being independent of marriage. It’s not fair enough to think of parenting style in the same way. If two adults commit to a plan for their future together and therefore get married, its at least plausible to expect that commitment to spill over into the way they parent their children. What the Australian studies didn’t mention is the strong correlation between family type and parenting style.
- Last week’s UK Department for Education study used much the same statistical method yet, amazingly, still found a ‘small’ unique effect of marriage. If parenting behaviour is in any way influenced by the decision to marry, then the full effect of marriage would be very much larger.
- The much cited UK study by the Institute for Fiscal Studies on the gap in break-up rates between cohabiting and married parents tries to play the same game. IFS were right to knock out background factors, such as mother’s education, for example. They were wrong to exclude behaviours, such as whether the birth was planned or relationship quality was strong, and then claim that marriage doesn’t matter.
- And finally, another recent study in the US by the Brookings Institute received a lot of media coverage by claiming the children of married parents do better because of the parenting they receive and not that the parents are married. Brad Wilcox at the National Marriage Project wrote a detailed critique of this if you want to know more!
The second error is to forget that ‘correlation’ can mean ’cause’. A common throwaway line is to dismiss the benefits of marriage as down to ‘correlation, not cause’. But that doesn’t mean you should rule out the possibility so long as there’s a plausible mechanism. Commitment theory provides that mechanism.
Here’s a silly example that illustrates both of these errors. Women often change their surname after getting married. There’s clearly a correlation here between marriage and name-changing. But correlation doesn’t imply cause. It could be that the kind of women who change their name also get married. Alternatively, since women also often wear wedding rings after getting married, it could be that the mechanism is not marriage but the behaviour of wearing a wedding ring. Taking wedding rings into account, you then find there’s no link at all between marriage and name-changing.
The absurd conclusion from this way of thinking would be to conclude that women change their name because they wear wedding rings. If we want fewer women to change their name, we should discourage rings. However if we wish to encourage name-changing, we should promote the use of wedding rings as a compelling new social policy …
So the next time you read about a study that says it’s not marriage but X, ask yourself if X is a behaviour or not. If it isn’t, the study is probably a good one. If it is, you need to ask how they can be so sure. After all, correlation doesn’t necessarily imply cause. But they shouldn’t rule it out.