But neither wedding cake nor the marriages they celebrate cause observed “marriage premia” any more than dances on tarmacs caused airplanes landing on Melanesian islands. In fact, for the most part, the evidence we have suggests that marriage is an effect of other things that facilitate good social outcomes rather than a cause on its own. In particular, for poor women, the availability of suitable mates is a binding constraint on marriage behavior. People in actually observed marriages do well because they are the lucky ones to find scarce good mates, not because marriage would be a good thing for everyone else too. Marrying badly, that is marriage followed by subsequent divorce, increases the poverty rate among poor women compared to never marrying at all. Married biological parents who stay together may be good for child rearing, but kids of mothers who marry anyone other than their biological father do no better than children of mothers who never marry at all.
Read the whole thing. The author has a much better handle on the research than most of what I've read in blogs on the subject and makes a convincing case.