Tuesday, September 27, 2011

Birth Weight and Later Outcomes In Ireland

A report published in 2006 by the Institute for Public Health in Ireland established that there were substantial inequalities in birth weight. Finding that there is a social gradient in this outcome is hardly surprising, it is the norm in other countries, and indeed the norm for most if not all health related outcomes, those with higher incomes tend to fare better. The recent Growing Up in Ireland study provides the opportunity to examine this issue in further detail. The following graph reiterates the differences in birth weight by family background using both income and mother’s education as indicators of SES. The mean birth weight is substantially higher in the later income quintiles, while a year of extra education for mothers is associated, on average with an extra 13 grams. Of course this is a simple linear fit but it turns out that controlling for other factors reduces this slightly, but not by a whole lot. Of course it’s hard to be sure that this is a causal relationship and that maternal education really affects birth weight (and is not just correlated with some other factor which does), but this result is consistent with studies which adopt more experimental approaches such as “Mother's Education and the Intergenerational Transmission of Human Capital: Evidence from College Openings”, Currie and Moretti, Quarterly Journal of Economics (2003) 118 (4): 1495-1532, and the Geary Working Paper “Mother's education and birth weight”, Chevalier, A. and V. O'Sullivan (2007).

A question that arises is whether these disparities in birth weight should be a cause for concern, or rather how much concern. Apart from the higher costs and increased risk of mortality which are associated with infant health, a growing literature has linked birth weight to a number of different outcomes in later life. In other words these inequalities in infant health are likely to persist. I also pursue this with the GUI data, and find that birth weight is predictive of various indicators at age 9, and not just health. The following graph depicts the relationship between Drumcondra maths/reading scores and birth weight. Again, controlling for the rich set of information available in the data does not alter this relationship substantially. Omitted variable bias is still a concern of course, however these results are entirely consistent with the literature, and in this context there is good reason to believe that there is at least some causal component to these effects. For a good review on the health side see the recent paper Almond, D. and J. Currie (2011). "Killing Me Softly: The Fetal Origins Hypothesis." Forthcoming in the The Journal of Economic Perspectives. Numerous twin studies also support this view, for example Black, S. E., P. J. Devereux, et al. (2007) "From the cradle to the labor market? the effect of birth weight on adult outcomes" The Quarterly Journal of Economics 122(1): 409-439.

I will be releasing a working paper on this shortly.


Liam Delaney said...

Nice use of GUI Mark. I have the early draft you have sent also. The standard questions apply but looking at GUI it's really hard to think of convincing identification strategies for assigning birthweights. And it really is time to think more flexibly about what we can get from data like this where it's unlikely that there will be hard data available that will permit the type of exclusion restrictions current in the literature. Having read your early draft I will provide more detailed comments offline but it's clearly an important national contribution.

Kevin Denny said...

I don't share the angst about "identification". But it depends on what question you are going to ask. I am not sure the usual endogenous-treatment framework is the best way of thinking about this.