Wednesday, January 10, 2007


This is refering to a news item before the new year, but it's an argument people bring up every now and then, so I feel like addressing it. The thing is, if you look at the statistics, relationships with children that are cohabiting are more likely to split up than those in marriage. This is clear, and is very true. This is not, as far as I know, a manipulated fact.

The problem comes from the conclusions drawn from this. The obvious one to draw, and this is the one politicans draw whenever they mention this particular fact is that marriage is a more stable environment than cohabitation is. And they may well be right. Certainly marriage has things going for it, particularly very clear laws on what to do in the case of divorce. But it certainly isn't necessarily nice. See, the problem here is something called confounders.

When statisticians want to discover if there is a link between two things, they will often be unable to conduct a proper experimental test, and have to use observational data- that is look at people in real life situations, for example smokers, and see if they get above average amounts of lung cancer. And, indeed, they do. However, you cannot necessarily claim that smoking is what causes this. In fact it might well be that people who are predisposed to smoke are also predisposed to eat a particular kind of food, for example, and it is THAT that is causing lung cancer. So reducing smoking would have no real effect- it only appears to be causing it. Of course there are ways to deal with this, by doing lots and lots of surveys and trying to link people by their characteristics- make sure that your group of smokers DOESN'T do anything different- other than smoke that is- than your group of non-smokers. As time goes on and more evidence builds up, you become more and more confident about causality, although you can never be sure- statistics really doesn't allow you to be sure.

In this case we don't have any linking between groups here, any particular difference in characteristics between cohabiters and people in marriage is not displayed. We just have one statistic. It may well be that people who are less committed tend not to get married, and because they are less committed, they are more likely to end the relationship. This is a very simple, and quite probable situation off the top of my head, other factors could be far more complicated. Sadly, of course, most people are blinded by these kind of figures- the phrase "lies, damn lies and statistics" only holds true if you allow yourself to be blinded by them. While there are often different ways to interpret statistics, there is usually a way which is more correct than others. If there isn't, then there probably haven't been enough tests done to gather a coherent picture of the situation.


At 3:49 pm, Blogger Jeff said...

I suppose people just enjoy the hype generated by statistics, regardless if those statistics are scientifically correct or not.

Just one thought on the smoking anology, I thought the damage caused by smoking are indeed well proved and documented :)

At 2:48 pm, Blogger Mr K said...

Well yeah it's pretty much certain, and thats aided in the most part by medical fidnings also that add weight to why smoking might cause lung cancer etc. However statistics alone can probably never assign causality to anything with complete certainty- it can just be highly likely.


Post a Comment

<< Home