The least controversial measure of health is mortality. Mortality is affected both by underlying health status and by medical care received, and it is not possible to separate out these factors using mortality data alone. In the case of infants, it is possible to proxy underlying health status using birth weight. Currie and Gruber (1996a) focus on the recent extensions of Medicaid eligibility to pregnant women and infants and ask whether these extensions reduced the incidence of low birth weight in addition to reducing the infant mortality rate. They use state rules to calculate the fraction of 15 to 44 year old women in the March CPS who would be eligible for Medicaid coverage in the event of pregnancy in each state and year from 1979 to 1990. They then estimate models in which the fraction of low birth weight infants in the state, and the state infant mortality rates, are functions of the fraction of women who are potentially eligible. State and year dummies are included in the models in order to control for any state or year specific determinants of mortality(4).
They find that the observed 20 percentage point increase in Medicaid eligibility over the 1980s reduced the incidence of low birth weight by two percent and the incidence of infant mortality by 8.5 percent. Cole (1994) reports similar results regarding the incidence of low birthweight using county-level data. This finding supports an earlier study by Hanratty (1992) which showed that the introduction of universal health insurance in Canada was associated with a decrease in the infant mortality rate. As in the U.S., public health insurance was adopted by the Canadian provinces at different rates.
Currie and Gruber (1996b) estimate models in which aggregate state-level child mortality rates depend on the fraction of children eligible in each state, year, and age group. Using this objective measure of child health they find that the 15 percentage point increase in the fraction of children eligible for public insurance between 1984 and 1992 was associated with a .2 percentage point decline in child mortality, which translates into a 5.1 percent decrease in child mortality. Moreover, this difference is statistically significant for deaths from internal causes such as disease (which one might expect to be affected by medical intervention), but not for external causes such as accidents and homicides.
Thus, the evidence suggests that the expansions of public health insurance of the 1980s and 1990s were effective in reducing infant and child mortality. However, they probably did not achieve this goal in the most cost effective way. For example, some of the infants saved by Medicaid coverage of their births, might have been better served by earlier access to prenatal care. Moreover, as discussed above, Medicaid does not necessarily give children access to private physicians, and it is much costlier to treat children in settings such as emergency rooms. Children served in emergency rooms and outpatient clinics are more likely to be hospitalized (Gold and Greenlick, 1981) and audits of hospital records suggest both that 20 to 30 percent of pediatric hospitalizations are medically unnecessary, and that Medicaid coverage increases the probability of unnecessary hospitalization (Kemper, 1988). Medicaid coverage also seems to be associated with an increase in necessary hospitalizations for conditions that could have been prevented with adequate primary care. Both uninsured and Medicaid patients are more likely than privately insured patients to be hospitalized for chronic conditions such as asthma, and communities in which people report barriers to medical care also have higher rates of such hospitalizations (Bindman et al., 1995; Weisman et al., 1992).