Because differences in realized schooling were very small at age 17, the correlation between schooling and smoking cannot be the result of any causal effect of the former on the latter. Rather, the authors posit that unobserved “third factors” (e.g. heterogeneous discount rates) that influence both smoking and schooling are responsible for the observed correlation between the two variables.
This paper proposes a different explanation for the phenomenon observed by Farrell and Fuchs [1986 [1982]]. I hypothesize that teenagers who eventually go on to realize different schooling levels anticipate (with some degree of accuracy) their future outcomes; and that differences in youths’ prospects for future schooling attainment cause them to choose different levels of engagement in risky health behaviors.
An implication of my hypothesis is that factors that affect youths’ expectations of future schooling events (for reasons that are exogenous to their health decisions) will in turn have an effect on their involvement in risky behaviors as teenagers. This implication is not shared by the“third factors” hypothesis of Farrell and Fuchs [1986 [1982]], which contends that schooling and health are only related spuriously. Furthermore, though similar arguments have been noted in papers such as de Walque [2010], Becker [2007], and Grossman [2005], empirical work examining whether schooling has a causal effect on health has focused almost exclusively on health outcomes that are realized subsequent to schooling decisions.4 In contrast, little is known about how health practices change with respect to anticipated schooling. This is the focus of this paper.