The income channel operates through the well-documented effect of education on labor earnings. In a comprehensive survey, Card concludes that by the 1990s, the causal effect of an additional year of education on annual earnings for both women and men in the United States was around 15 percent.6 The notion that an exogenous increase in a woman’s income may lead to reduced fertility is present in the earliest treatments of the neoclassical model of fertility (Mincer 1963, Willis 1973). In these models, households do not value children per se, but what Willis terms “child services”—the product of the number of children and the average quality of those children. A key idea is that production of child services is time-intensive relative to other activities for the woman. As the value of a woman’s time rises, she generally substitutes away from consumption that is highly timeintensive (Becker 1965) and hence desires fewer children. These predicted effects of education on fertility map naturally into predicted effects on child quality. Assuming child services are a normal good, falling fertility in response to rising income requires that child quality be an increasing function of income. Cross-price effects such as these were first emphasized by Becker and Lewis and Willis.
Predictions based on the income channel are further sharpened by positive assortative mating, or the tendency for men and women of similar education to pair (Behrman and Rosenzweig 2002). Under this type of stratification, an exogenous increase in a woman’s education leads to a mate of higher education, further increasing household permanent income through a multiplier effect.
In addition to the income channel, the literature has stressed the role of education in augmenting an individual’s stock of health knowledge.8 With respect to fertility, Rosenzweig and Schultz provide evidence that a woman’s education explains ability to effectively use contraception.
With respect to infant health, Thomas, Strauss and Henriques show that education predicts a woman’s ability in regards to, or perhaps interest in, information acquisition and processing.9 One of the most frequently-cited examples is smoking (Currie and Moretti 2003). Through anti-smoking campaigns in schools or health class, children could learn about the dangers of smoking and be discouraged from adopting the habit. Glewwe argues that the most important mechanism for knowledge gain is not directly via curricula; rather the skills obtained in school facilitate the acquisition of health knowledge. Grossman formalizes these ideas by viewing education as a productivity shifter in the household production function for health.
Since education can affect infant health through several different channels and the intensity of these channels may not be the same for all levels of education nor for all subpopulations, the effect of education on infant health may differ across studies. For example, Currie and Moretti use college openings to study the effect of maternal education on infant health. The women whose schooling attainment at motherhood is affected by college openings are those women with a high level of education generally. As we show below, our study focuses on the causal role of education for women with a low level of education generally. Educational levels that appear to be affected in our study are in the range of eighth to twelfth grade, with a muted effect on the first two years of college. This subpopulation is of interest for several reasons. First, the observational infant health return to education is declining in the level of education. Second, the labor market return to education is declining in education (Card 1999). Third, young women at risk of dropping out of school are frequently the target of specific policies aimed at reducing fertility and improving infant health.