"We propose a novel approach to measuring the causal effect of family structure on a child’s outcomes. In a war, some fathers are killed in action and cannot return to their families. This creates a natural experiment in which the effects of a father’s absence can be tested. Using data from the U.S. Civil War, we find no evidence that a father’s death in the war affected his child’s labor income as a young adult. We also find no effect on labor force participation or the chance of being married in 1880. Daughters of fathers who died were less likely to be students in 1880, although we find no such effect on sons".
It would be great to do similar analysis on the effects of the Civil War in Ireland.
It would be great to do similar analysis on the effects of the Civil War in Ireland.
Although I haven't read the paper,it sounds like an example of the poverty of "natural experiment" thinking. The idea presumably is that the treatment is randomly assigned and has no effect on the outcome except via the family structure (Z on X but not Y directly).
ReplyDeleteIs this really plausible about having one's father killed in a war? Can one think of it as just some people randomly getting fewer parents? Bollox.Ever heard of trauma?
Moreover the probability of "treatment" would be correlated with un-observables which are probably partly heritable. Fathers who are braver/more stupid being more likely to die.
I was saying myself to Liam that there's a lot of potential holes with it, particularly the selection effects that would produce fathers that fight/fight more and who are "braver/more stupid being more likely to die".
ReplyDeleteI just flicked through the paper but he seems to discuss a lot of the issues over its 76 pages. I'm going to read a lot more stuff like this next year.
the Civil War in Ireland was probably too contained to offer much by way of analysis. in general, there are not that many (if any) big events in Ireland in the 20th century that would lend themselves to a regression discontinuity or instrument approach in any field really. Even the ROSLA-type instruments dont work!
ReplyDeleteregarding the US paper, i think its a good paper in terms of estimating health and wage equations in a historical setting. there is no real way of proving one way or the other the randomness of violent civil war deaths as far as i can tell from what he says. leaving that aside, the paper still holds as a good econometric analysis though causal conclusions are mitigated.
Even if the death of one's father was random, is it plausible that this would not effect a health,income or education outcome other than through family structure? It doesn't sound like a good identifying assumption.
ReplyDeleteThe abolition of school fees in the late 1960's works dandy as an instrument by the way.
i didnt know that school fees abolition worked - i was thinking of the Harmon/Callan paper that tried Irish ROSLA as an instrument and concluded that it didnt work.
ReplyDeletehas there been a paper on the school fees abolition?
Education Policy Reform and the Return to Schooling from
ReplyDeleteInstrumental Variables
Version 2.10, April 7th 2000
http://www.ifs.org.uk/wps/wp0007.pdf
Abstract:
This paper exploits an unusual policy reform that had the effect of reducing the
direct cost of schooling in Ireland in the late 1960’s. This gave rise to an increased
level of schooling but with effects that vary substantially across family background.
This interaction of educational reform and family background generates a set of
instrumental variables that are used to estimate the return to schooling allowing for
the endogeneity of schooling. Using a standard Mincer type model we find a large
and well-determined rate of return of around 12% which are substantially higher
than the OLS estimates of around 7%.
Meant to say: the above is Denny and Harmon.
ReplyDeletecheers Martin - has anyone done anything yet with the change in the college fees structure to estimate returns to college education? im thinking of a propensity score match before and after the free fee introduction. given the huge growth happening at the time, it would be a noisy task but interesting anyway.
ReplyDeleteNot that I know of Liam. I've been meaning to set about looking at how the abolition of school fees in the late 1960's would work as an instrument in relation to the effect of parental education on participation in higher education. I'm going to do it this month.
ReplyDelete