Bryan Caplan says the marriage premium (pdf) - the well-established fact that married men tend to earn much more than single ones - is due in large part to causality. Being married, he says, causes men to earn more, rather than there being selection effects, whereby men who are attractive marriage prospects are also attractive to employers. Several things make me agree with him.

- Introspection. I strongly suspect that if I'd been married, I would have earned more. This is partly because a wife might have inspired me to overcome my distaste for job search, and partly because having someone to provide for would have caused me to move along the wage/job dissatisfaction indifference curve. It's also plausible that a wife can nag a man to get a job.

- The marriage premium is bigger for straight men than gays (pdf). This is not obviously consistent with selection effects; if it were the case that the same things that make a man attractive to an employer also make him marriageable, shouldn't this effect be as powerful for gays as straights? It is, however, consistent with a causal effect. Having a wife frees a man from housework, thus allowing him to focus on his job, and gives him a chance of having children for whom he must provide.

- Top baseball players earn more if they are married. Again, this is inconsistent with selection effects; most such men are surely highly marriageable. But it is consistent with a causal role. Maybe employers discriminate in favour of married men, believing them to be more reliable. Or maybe a wife emboldens a man to take a more aggressive stance in wage bargaining.

- Even a study which favours the selection hypothesis finds that there's a causal effect at the lower part of the wage distribution.

However, on the other hand, something else leads me to favour the selection effect. It comes from happiness research. There's good evidence that married people, on average, are happier than singletons. However, Andrew Clark has shown that the effect on happiness of marriage fades away quite quickly. How can we reconcile these facts?

Simple. Happier people are more likely to be married; who wants to marry a miseryguts? Marriage doesn't cause happiness, but it selects for it.

And here's the thing. There's evidence (pdf) that happier people earn more (pdf) than less happy ones.

Perhaps, therefore, the positive link between marriage and wages (for men) isn't wholly causal, but reflects the fact that happiness increases both wages and marriage chances.

If all this sounds like an abstruse issue of labour economics, it shouldn't. There's an ideological undertow here. Bryan thinks that poverty (in the US) is due to individual bad behaviour, and so changing behaviour "get married!" is a route out of poverty. His interlocutors, however, disagree.

However, we must avoid the trap of motivated reasoning. The marriage premium is causal or not, whether you want it to be or not. I suspect that some of it is causal. But this doesn't mean I have to accept that poverty is simply due to individuals' failings.