Gans and Leigh on the "Baby Bump" 

 In  Born on the First of July: An (Un)natural Experiment in Birth Timing , forthcoming at the Journal of Public Economics, Joshua Gans and Andrew Leigh examine "introduction effects" (the extent to which people change their behavior to respond to new policies) in the context of a baby bonus that was initiated in Australia in 2004. In May of that year, the government announced that families of babies born on or after July 1 would receive a $3000 cash bonus. Mothers with due dates around that time made special arrangements (mostly delaying Caesarean and other planned deliveries) to get the prize. The authors estimate that over 1000 births were moved; July 1, 2004, witnessed more births than any other day in the period since 1975 for which the authors have data.  


 The authors note two implications of the study. First, policies can provoke not only long-run distortions (e.g. increases in babies born) but short-run distortions from gaming of the system. Second, the "baby bump" constituted a large disruption in regular procedures in maternity hospitals and staff; they don't find effects on infant mortality, but they suggest that the event could be useful for studying the effects of under-staffing in hospitals. 

 My first thought in reading the paper was that it was a cautionary tale for regression discontinuity design. The setup of the study has the flavor of David Card et al's paper " Does Medicare Save Lives? ,"  discussed on this blog  by the intrepid John Graves, in which the authors examine the outcomes of patients who need medical procedures right around the time when they become eligible to receive Medicare benefits; they find that patients who were barely old enough to receive the benefits do considerably better than the ones who were too young. I figured this study was probably a failed attempt to do something similar, ie to study the effect of extra income on child mortality or other outcomes by comparing kids born just before and after the benefit was handed out. This sort of thing doesn't work when the subjects are able to sort around the threshold. In this case, the parents who gave birth just after the cutoff may have been more desperate for money, or had more power with the doctors, or perhaps were generally more in tune with political events, such that differences in outcomes between recipients and non-recipients of the bonus could be due to these factors and not the bonus itself. In Card et al's case, they focused on emergency procedures that could not have been delayed; this study shows that for many people childbirth is quite postponable. So in addition to the implications Gans and Leigh draw in their conclusion, I would add that this is another case where an RDD-style approach is complicated because subjects can effectively sort.  

 I do think you could examine the effect of the bonus on child outcomes if you looked at kids born at least 3 weeks before and after the cutoff date, a point at which the sorting is probably not that big of a deal. And date of birth is probably not itself a strong confounder for whatever you want to study, so there are limited advantages to focusing in on the threshold anyway. 

 At any rate, it appears my initial impression -- that the paper is the artifact of a failed RDD project -- was wrong: the authors have done other examinations of how events affect birth patterns ( the effect of the millennium on conceptions, births and deaths , and the  ability of parents to move births from inauspicious days  like Feb 29 and April 1).