Date of Degree


Document Type


Degree Name





Michael Grossman

Committee Members

Harold Hochman

Elliot Zupnick

Subject Categories



The object of this research is to develop a model of household demand for prenatal care and attempt to measure the productive value of prenatal care per se on infant health as measured by survival. Traditionally, infant mortality rates have been used as indices of a nation's health status. Since the U.S. has lagged significantly behind other developed nations in reducing infant mortality since the mid-1950's, there have been charges of a malfunction in the U.S. health delivery system. Particularly in the area of infant health, critics have charged that more prenatal care Inputs are needed and that they should be directed specifically towards so-called high risk mothers. Others have questioned the value of input intensive prenatal care, claiming its marginal product is low, cost high and efficacy unproven.

An economic model is developed in which the demand for healthy children is viewed as being derived from the demand for children per se. In a world where families cannot substantially effect the outcome of Individual pregnancies by varying inputs, it is demonstrated that measured infant mortality rates will be not only a function of health status but also fertility decisions, in a world where families can vary inputs, it is argued that prenatal care, as the most pregnancy relevant related input, should be a good index of the total demand for pregnancy related inputs. It is argued that the level of inputs will be positively correlated with income, tend to increase in families who have experienced pregnancy losses and decrease as family size increases, particularly if marginal children are less "wanted" as family size increases.

Demand and production relationships are estimated using data from the 1970 New York City birth cohort. The data set consists primarily of birth and linked death certificates for the period January to June, 1970 and contains 54,000 observations after editing.

Several different dependent variables are utilized to estimate the demand for care. They Include a dichotomous care/no care variable, the interval to the first visit and the number of visits. Significant empirical findings include: (1 ) the decision whether or not to seek care Is most strongly influenced by legitimacy status; (2) the demand for care is effected by past experience as predicted by the model in that families with more live children demand less care and those with a history of losses demand more care; (3) substantially less care is demanded by blacks, foreign born and Puerto Rican born mothers even when other variables are accounted for; (4) less care, other things equal, is obtained in specially designated Maternal and Infant Care Project areas, despite the presence of these special projects to encourage the use of care by high risk mothers of low socio-economic status; (5) the amount of care a mother receives is substantially determined by obstetrical protocol and does not seem to reflect her previous pregnancy experience.

Outcome measures Include birth weight, infant death, neonatal and postneonatal death. Regarding birth weight significant findings include; (1 ) birth weight differentials atrributable to race, ethnicity, nativity or legitimacy characteristics are substantially reduced by taking account of differentials in the level of care received; (2 ) the net gain in birth weight attributable to a full complement of prenatal care (303 grams) as compared with no care is substantial when comapred with the birth weights of "high risk" infants; (3) previous experience of pregnancy successes and losses are reflected by increments or decrements in birth weight.

In comparing results of outcome regressions for neonatal and postneonatal mortality, it is found that other things equal, prenatal care has a positive effect on survival during the neonatal period but no effect during the postneonatal period. Hence, it is argued that care per se has real value in improving pregnancy outcomes and is not primarily acting as a proxy variable for "wantedness" or other unmeasured inputs.

The results of using FIML logit estimators on the dichotomous dependent variables biased on a subsample of observations do not agree with the OLS estimates based on the entire sample. It is suggested that econometricians need to more fully explore the relationship between the value of these two techniques, particularly in very large data sets.


Digital reproduction from the UMI microform.

Included in

Economics Commons