Estimating the effect of smoking on birth weight in a dynamic model when fertility is a choice
The negative effect of smoking during pregnancy on birth weight outcomes has been a consistent finding in the economics literature on estimating birth weight production functions. An important result in the literature is that the negative effect of smoking on birth weight is generally robust to the introduction of unobserved heterogeneity in family-specific health endowments. In the past two decades, studies have been increasingly interested within the economic and medical literature upon the estimation of birth weight production functions, however, fertility itself is unrelated to either anticipated or realized birth weight outcomes that depend on such endowments.
There are several reasons for this economic-medical crossover of interest. First, infant mortality has been shown to steeply decline with birth weight. Secondly, other literature reports that birth weight is related to outcomes of interest such are morbidity, later health and intellectual achievements. Also health problems such as deafness, blindness, epilepsy, chronic lung disease, learning disabilities and attention deficit disorders commonly in low birth weight infants. In a later study conducted in 1995 by Lewit et al found that health care, education and child care costs in 1988 for the 3.5 to 4 million children between the ages of 0 to 15 who were low birth weight ranged from 5.5 to 6 billion dollars more than had these been of normal birth weight. This lead to the finding of several problems in estimating birth weight production functions such as failing to account for the possibility of different behavioural inputs depending on unobserved endowments which would generally lead to a biased estimate of the birth weight production function through the direction of the bias being a priori unclear.
One defining factor is the women who come from families with “poor” health endownments might quit smoking during pregnancies, or avoid smoking completely to somewhat compensate for their “poor” health endowments. However due to their poor health endowments they might still realise poor birth weight outcomes.
One purpose of the full distritation was to explore the impact of relaxing that assumption on the estimates of the smoking effect on birth weight. To that end, a dynamic model of fertility choice that explicitly incorporates the smoking decision, allowing for its addictive nature, and the birth weight technology, is constructed and empirically implemented using longitudinal data from the National Longitudinal Surveys 1979 youth cohort.
The estimates of the model imply that avoiding heavy smoking during pregnancy will increase the birth weight of the born child by almost 5oz., or approximately 140g. This estimate, however, is only slightly lower than the one obtained from a sibling Fixed-Effects estimation procedure which is based on the simulated data, and is equal to the sibling Fixed-Effects estimate obtained from the actual data.
In addition to obtaining estimates of the birth weight production function that account for fertility choice, the estimates of the model are used to perform counterfactual policy experiments. In particular, the model predicts that preventing women from smoking during pregnancies will increase average birth weight outcomes by 0.7oz. In addition, this policy will reduce the incidence of low birth weight by about 10%. Furthermore, increasing the cigarette taxes will lead to a significant decline in the percentage of women who smoke, both while they are pregnant and while they are not pregnant. Consequently, in the event of a 50% increase in the price of cigarettes about 6% of the pregnant women will realize a significant increase of 4.7oz., on average, in their children’s birth weight.
Shnaps, R. (2001). Estimating the effect of smoking on birth weight in a dynamic model when fertility is a choice. (Order No. 3015372, University of Pennsylvania). ProQuest Dissertations and Theses, 95-95 p. Retrieved from http://search.proquest.com/docview/251759058?accountid=13380. (251759058).