« Tips for panhandlers, from panhandlers | Main | Does this count as an event study? »
The economic consequences of unwed motherhood
This was published in the American Economic Review in 1994:
We estimate the long-run and life-cycle effects of unplanned children on unwed mothers by comparing unmarried women who first gave birth to twins and unmarried mothers who bore singletons. We find large short-term effects of unplanned births on labor-force participation, poverty, and welfare recipiency among unwed mothers, but not among married mothers. Although most of the adverse economic effects of unplanned motherhood dissipate over time for whites, there are larger and more persistent negative effects on black unwed mothers.
Notice that comparing one birth to two, rather than zero to one, tries to address the identification problem, namely that early pregnancy may be correlated with other unfavorable conditions. For the curious, here are many related articles. And here is a very useful literature review, which suggests inconclusive results.
Posted by Tyler Cowen on September 3, 2008 at 10:58 AM in Economics | Permalink
Comments
ARE YOU SERIOUS??
The difference between twins and singleton birth is a measure of the impact of unwed mothers having children????
Shouldn't it be a comparisons between having no children and one child ???
Posted by: spencer at Sep 3, 2008 11:29:36 AM
Is the above comment sarcasm? The authors' method was exceedingly clever imo.
Posted by: a person at Sep 3, 2008 11:58:25 AM
Prof. Cowen,
This is a great pedagogic approach to the headlines.
Posted by: Hopefully Anonymous at Sep 3, 2008 12:10:40 PM
The above comment was serious.
It is a perfect example of the economist looking under the street light.
What does it say that you though it was great because of the methodology rather than it having any bearing on reality?
Posted by: spencer at Sep 3, 2008 12:25:18 PM
agreed with person. you don't think it has any bearing on reality, spencer?
Posted by: josh at Sep 3, 2008 12:47:54 PM
I'm with Spencer here. Twins may be a greater financial burden than a single child, but it seems to me the opportunity costs of unwed motherhood are going to be roughly the same in both cases, and drastically different than having no children at all.
Posted by: Sol at Sep 3, 2008 1:02:28 PM
Spencer - The problem is one of identification. Comparing mothers with one child to the same person who has no child is going to be problematic since motherhood is not randomly assigned to young people. There's many unobserved differences between young mothers and a young woman, many of which may be correlated with having a child. We can try to control for them, but the unobserved heterogeneity problems will still persist. Twins are often used in studies like these because having a twin is basically a random event, and thus conditional on women who have children, we can observe the effect of an additional child on some maternal outcome, as well as on the child's outcome. It's of course only valid at that local margin, so it'll depend on whether you think we can learn something about going from 0 to 1 that is similar to what we learned by going from 1 to 2.
Posted by: jason voorhees at Sep 3, 2008 1:15:12 PM
In an ideal situation, we could collect a sample of people with exactly similar characteristics and, in this case, "treat" half of them with a child. How do you find a representative sample for such a population? How do you find the people whose characteristics point to a liability for wedlock children, and simply have not had that child? Such an exercise could prove exceedingly difficult if impossible when one considers the sheer number of variables that would need to be selected in the control group.
With this (clever) method the problem of finding this obscure sample of, in a sense, unwed mothers without kids, is removed. Of course the opportunity costs of the second child may be different than the opportunity costs of the first child - I'm sure the authors are aware of this possibility. That does not make this paper ridiculous. A paper like this can tell us quite a bit about the costs of unwed pregnancy, and likely more than the comparable paper (filled with a strong dose of sampling error) that examines the cost of the first child.
Posted by: a person at Sep 3, 2008 1:24:58 PM
It might not make the paper ridiculous, exactly, but if the opportunity cost difference between 0 and 1 is much higher than the difference between 1 and 2 (as I think it is), then it makes it not terribly accurate.
Though I guess it probably does set a good solid lower bound on the cost of having a child out of wedlock, which is a worthwhile piece of information.
Posted by: Sol at Sep 3, 2008 1:53:07 PM
voorhes -- the problem of identification is exactly why it is a great example of
the economist looking under a street light.
This was presented as an estimate of the cost of unwed motherhood and I find it unbelievable that the difference between having one and two children in any way remotely resembles the opportunity cost of the first child.
Give me one reason to think that.
Posted by: spencer at Sep 3, 2008 2:38:30 PM
To your point. Going from 0 to 1 does seem to me, too, to be different than going from 1 to 2. I don't know if it's a difference in degree or kind, though. The problem is that, again, we can't randomly induce a child to mothers and do the comparisons you're wanting. We can, though, do the latter. The twin experiment at least allows us a chance to understand at the margin what an additional child does, conditional on the characteristics of women who already choose to have adolescent births. You're saying you can't learn anything about this to generalize elsewhere. I'm not so sure. Seems like what we can learn here is that probably going from 0 to 1 will be at least as bad as what we found going from 1 to 2 - a lower bound, like Soi said. That may be the best we can actually learn, unless you have a better identification solution. Sometimes, you have to settle for the second best when it comes to social science.
Posted by: jason voorhees at Sep 3, 2008 2:58:37 PM
I have no problem with it being an estimate of the lower bound and on that basis saying it is good work.
But that is not at all what Tyler presented it as.
Posted by: spencer at Sep 3, 2008 3:23:58 PM
Another important issue is that in single parent households there is one potential income, but in two parent households there are two.
So, even if becoming a single parent does not impact your labor supply or educational attainment, it still makes you more likely to be in poverty and using government transfer programs--because there is only one potential earner. i think a lot of this literature misses this point.
Posted by: Ben at Sep 3, 2008 3:34:07 PM
My Daddy (Mr. Beefy IV) always said "Early to ripe, early to rot!" :).
Posted by: Mr. Beefy at Sep 3, 2008 4:09:45 PM
Without reading the paper: Are women more or less likely to abort/put up for adoption if they know they are having twins (seems like this would be easy data to get from clinics, agencies)? To what extent is having the twins an indicator of family/local support or optimism? But maybe I'm missing the point.
Posted by: NE1 at Sep 3, 2008 4:32:37 PM
the summary said:We estimate the long-run and life-cycle effects of unplanned children on unwed mothers.
Both me and those who took acception to my comments now agree that is not at all what this paper does.
Posted by: spencer at Sep 3, 2008 7:40:05 PM
@A person:
filled with a strong dose of sampling error
A bit trivial perhaps but don't you mean selection bias?
Posted by: JSK at Sep 4, 2008 9:11:52 AM
Yes! sorry about that one...
Posted by: a person at Sep 4, 2008 1:05:27 PM
