9 Short Observations about the Marriage Premium
In the past, I’ve faulted economists for ignoring the marriage premium (here, here, and here for starters). Last week, when Pascal-Emmanuel Gobry and Megan McArdle joined my fault-finding expedition, Justin Wolfers pushed back on Twitter:
There’s no credible evidence justifying the claim that the marriage wage premium is causal.
Justin, you don’t sound very Bayesian. [and linked to this post on economists’ selective agnosticism].
The Bayesian reading is to confess to a prior that there’s strong selection into marriage.
After reading this, reasons for Justin to rethink his position kept popping into my head. Here are the top nine:
1. No credible evidence? How about the mere fact that the marriage premium (for men) and marriage penalty (for women) persists after controlling for age, education, race, and a long list of other confounding variables?
2. “Strong selection” hardly implies zero treatment effect – and the observed male marriage premium is so large (often over 40%) that the absolute treatment effect would still be large even if 75% of the difference were due to selection.
3. After controlling for observables, do married and unmarried men really seem radically different? What unmeasured pre-existing traits of married men could conceivably lead them to earn 40% more than unmarried men?
4. Following Heckman’s lead, education economists increasingly emphasize the importance of non-cognitive skills (conscientiousness, ambition, organization, etc.). Isn’t it plausible that marriage would causally raise men’s non-cognitive skill via expectations, praise, nagging, devotion, etc.?
5. It’s easy to see the appeal of the selection story: Married people have many traits in common: willingness to commit, to defer gratification, to conform to social norms. Why then, though, do married men earn a large premium, while married women earn a modest penalty? Shouldn’t favorable selection enhance women’s earnings, too?
6. If you say, “Women de-emphasize their careers after they marry,” you’re appealing to a treatment effect. And if you can believe that women de-emphasize their careers as a result of marriage, why can’t you believe that men emphasize their careers as a result of marriage?
7. Kids substantially reduce female earnings. Few doubt that this effect is causal: Kids don’t just take time; they change priorities. If you can believe that kids change their moms’ labor market behavior, why can’t you believe that spouses change each others’ labor market behavior?
8. If marriage is just a piece of paper, how can it possibly cause higher or lower earnings? Because marriage is also a state of mind. Some people (like Justin Wolfers and Betsey Stevenson) can enter this state of mind by sheer act of will. But most people need formal social recognition to effect this transformation.
9. Social psychologists have produced a lot of experimental evidence that merely identifying as a member of a group can have a large effect on behavior; see e.g. the famous Robbers Cave Experiment. Why is it so hard to believe that identifying as “Husband of X” or “Wife of Y” can have such an effect? Or simply identifying as “a Husband” or “a Wife”?
Are any of these arguments – or all of them together – good enough to convince a skeptic like Justin that a substantial part of the marriage premium is causal? Probably not. But my arguments are good enough to convince reasonable skeptics that this issue deserves far more empirical attention from economists of Justin’s caliber.