Please don't trickle down off the fiscal cliff

In June of 1988 in Sweden it was announced that survivorship benefits, a sort of government provided life insurance paid to a wife whose husband dies, would be discontinued. There was one interesting exception:  an unmarried couple with a child together born before the change could take up survivorship insurance if they married before Jan 1 1990.  The spike in new marriages in the graph shows the response to this incentive.

That’s the basis for Petra Persson‘s job market paper. Petra points out that the spike is somewhat mysterious because for all of these couples the promise of survivorship insurance wasn’t enough to induce them to marry previously and only when the option was going to disappear did they exercise it.

Of course some of these new marriages were couples that planned eventually to marry (and take up benefits) and who moved their marriage date earlier. But Petra credibly demonstrates that a large proportion of these marriages were marriages that never would have happened had the reform not been announced. What explains those “extra” marriages?

Petra’s theory is that these couples were still uncertain about whether they were a good match and were planning to live together longer before deciding later whether to marry.  After the reform was announced this option to wait and see was no longer costless and therefore many of these couples rushed into a marriage that, given enough time, they might have eventually decided against.

There’s an alternative story that fits equally well. Consider a couple where  there is no uncertainty at all about whether the match is good:  its a bad match and that’s why they are not married.  (Or it could be that they are perfectly happy together but just see no value in being legally wed.) This couple optimally plans to wait until the husband is close to death and then (if he hasn’t married somebody else) get married in order to take up survivorship insurance.  Now once the reform is announced that option is removed and they re-optimize and marry December 31, 1989.  Many of these are extra marriages because if they waited he might die unexpectedly or marry somebody else.

This theory (like Petra’s) also explains some other facts. For example, conditional on the husband not dying shortly after the reform the divorce rate for these marriages was unusually high. And even after controlling for everything a private insurance company would use to assess risk, takeup of the survivorship insurance via marriage is a good predictor of earlier-than-expected death.

I wonder what we could look for in the data to distinguish the two theories.

It’s a great paper and there’s lots more in there, you should definitely take a look. If I were making a list this year (I am not) Petra would definitely be on it.  (Check out her paper on information overload.)