The Long-Run and Intergenerational Impact of Poor Infant Health : Evidence from Cohorts Born During the Civil Rights Era *

The 1960s witnessed the greatest reduction in the mortality rates of black infants relative to white infants of the last 50 years. We use these dramatic relative changes in the health of black birth cohorts to evaluate the long-run effects of early life health conditions. The microdata contained in the annual Natality Detail files provide information on the characteristics and health risk factors of the universe of mothers giving birth in the U.S., as well as the birth outcomes of their infants. We link these data to the infant health conditions that prevailed in the state and year in which the mother was born. The analysis compares differences in the health and birth outcomes of black and white mothers born in the late 1960s to those of mothers born in the early 1960s, while contro lling for age of the mother and the year in which she gave birth. We find that black women born in 1967-1969 have substantially lower risk factor rates as adults and are much less likely to give birth to an infant with low birth weight and APGAR scores than black women born in 1961-1963. The between-c ohort gains for white women are small to non-existent, consistent with the smaller health improvements for white infants born during the 1960s. The timing of the black-white relative birth cohort improvements corresponds with the timing of the 1960s infant health gains and is robust to several tests of internal validity. For example, the between-cohort relative gains are significantly larger for black women born in Mississippi than for black women born in Alabama, exhibiting strikingly similar patterns to the patterns in relative infant mortality rates in the two states during the 1960s. We conclude that the social policies that led to the infant health improvements (e.g., hospital desegregation, Medicaid, Food Stamps) may have had long-run and intergenerational health benefits. Also, studies of changes in racial health disparities over the life-cycle and over time could be severely biased by not accounting for the significant changes in the health of black birth cohorts. Douglas Almond Kenneth Y. Chay NBER Department of Economics 1050 Massachusetts Avenue University of California, Berkeley Cambridge, MA 02138 549 Evans Hall email: almond@nber.org Berkeley, CA 94720 and NBER email: kenchay@econ.berkeley.edu Introduction The 1960s witnessed the greatest reduction in the mortality rates of black infants relative to white infants of the last 50 years. Appendix Figure 1 shows trends in nonwhite and white infant mortality rates within a year of birth (per 1,000 live births) for the United States from 1950-1990 as well as the nonwhitewhite infant mortality rate (IMR) ratio. 1 Since 1950 there has been a secular increase in the black-white IMR ratio with one notable exception. From 1964 to 1971 the black infant mortality rate and the blackwhite ratio declined sharply relative to pre-existing trends. The black infant mortality rate fell 30 percent from 41 per 1,000 live births in 1964 to 28 in 1971, and the black-white ratio fell from 1.9 to 1.65, the only prolonged convergence in the post-World War II era. Almond, Chay and Greenstone ( 2001) find that while the black-white IMR gap narrowed in all regions of the United States, the convergence rates were greatest in the rural South and varied substantially within the South. For example, Appendix Figure 2 shows trends in black and white post-neonatal mortality rates from 1941-1971 for Mississippi, Alabama, I llinois, and New York; states with large AfricanAmerican populations. 2 In 1965 the black post-neonatal death rate in Mississippi was 26 per 1,000 live births -30 percent greater than the black rate in Alabama, 2-2.5 times greater than the black rates in Illinois and New York, and five times the rates for white infants in all four states. However, Mississippi also experienced the sharpest decline in black post-neonatal mortality after 1965, with the black infant death rate falling over 50 percent from 1965 to 1971. Alm ond, Chay and Greenstone ( 2001) argue that these dramatic changes in Mississippi were largely the result of the federal antidiscrimination effort of the mid-1960s. These figures suggest that there were large improvements in the early health of black cohorts born during the 1960s relative to their white counterparts, and that the relative gains varied significantly across the U.S. This study uses these dramatic relative changes to evaluate the long-run and intergenerational effects of early life health conditions. In particular, we examine whether African-Americans born in the 1 The data are from the Vital Statistics of the United States annual publications. In 1965, black births accounted for 92 percent of all nonwhite births in the U.S. 2 Conventionally, post-neontal mortality is defined as the death rate in the period from 28 days to 1 year after birth. 2 late 1960s have better health as adults and healthier infants in the 1980s and 1990s than blacks born in the early 1960s. A finding of a l ong-run and intergenerational link has striking implications. First, it suggests that the causes of the 1960s infant health improvements (e.g., social policy) had multiplier effects that have been unaccounted for in cost-benefit calculations. Second, it would provide an alternative explanation to racial differences in behavior, medical care access, and treatment for the existence of racial health disparities. Observed black-white differences in adult and infant health could partially be the result of racial inequality in the health the adults had as infants. The data requirements for a precise analysis of adult and infant health outcomes across narrowlydefined race, state and year of birth cells are enormous. Thus, we use the microdata contained in the 19851997 Natality Detail files, which are based on a census of all birth certificates in the U.S. These files provide information on the characteristics and health risk factors of the 23 million mothers born in the U.S. between 1959 and 1970 who gave birth during the 1980s and 1990s, as well as information on the birth outcomes of their infants. We link these data to the infant health conditions that prevailed in the state and year in which the mother was born. The analysis compares differences in the health and birth outcomes of black and white mothers born in the late 1960s to those of mothers born in the early 1960s. A wellknown identification issue is that age, birth cohort, and survey year are collinear. To address this, we use flexible controls for mother’s age and year in which she gave birth -e.g., interactions of a cubic polynomial in age with unrestricted survey year effects -and examine whether the estimated birth cohort effects exhibit trend breaks that correspond with the breaks in infant mortality during the 1960s. We find mother’s birth cohort effects that are very large in magnitude, particularly among AfricanAmerican women. Table 1, for example, presents the estimated effects for the incidences of birth weight less than 2500 grams (first four columns) and less than 1000 grams (last four columns) am ong the infants of women who were born in the United States during the 1960s. 3 The regressions include birth cohort dummies, dummies for age of the mother, marital status, educational categories, and survey year and allow each of their effects to vary by race. The first two columns show that when race-specific birth cohort 3 Low birth weight is conventionally defined as a weight less than 2500 grams. Birth weight less than 1000 grams (2.2 pounds) is strongly associated with eventual infant death (Almond, Chay and Lee 2002).