Use of Modern Medical Care for Pregnancy and Childbirth: Does Schooling Matter?

Objectives and method Controversy exists over whether, and to what extent, the estimated effects of schooling on health care use reflect the influence of unobserved preferences, knowledge, and motivation of individuals. To the extent that schooling outcomes reflect such attributes, existing estimates may overstate the schooling effect because of the failure to control for background variables. Alternatively, estimates of the schooling effect may be downwardly biased due to measurement error in the schooling variable. This paper contributes to the resolution of this debate by adopting an instrumental variable (IV) approach to estimate the impact of female schooling on maternal health care use in Indonesia during the 1990s. A large-scale school construction programme that took place in Indonesia in the 1970s, the Sekolah Dasar INPRES programme (SDIP) is used to construct an instrumental variable for education, following Duflo (2000). The interaction between an individual's exposure to the school building programme and programme intensity in her region of birth is used to instrument years of schooling. The choice between use and non-use of maternal health services is estimated as a function of schooling and other individual and household level attributes. Data from the Indonesia Family Life Survey are used for this paper, along with administrative data on SDIP. Results Standard regression models estimated in the paper indicate that each additional year of schooling does indeed have a significant, positive effect on the timing of ante-natal care visits, use of hospital care and skilled assistance at delivery. IV estimates of the schooling effect are larger. The results suggest that schooling has a positive impact on maternal health care use even after eliminating the effect of unobserved variables and measurement error. The analysis also highlights several other correlates of poor maternal health use such as low levels of access to services and household socioeconomic status, which could potentially offset the impact of schooling. Contributions to research and policy This paper moves beyond previous work on the impact of education on health care use by adopting an IV approach to address the problem of endogeneity and measurement error. IV methods have been used widely in the labour economics literature to examine the impact of schooling on wages and other labour market outcomes but have rarely been used to estimate the effect of schooling on health outcomes. A second contribution of this paper is to provide a more accurate assessment of the impact of education on health care use relative to other determinants. Public policy has generally emphasised the role of female schooling as a primary driver of preference change. This paper does not seek to identify the effects of other factors driving changes in women's preferences for health care use. It does, however, provide an assessment of the size of the schooling effect relative to other key determinants of health care use such as household socioeconomic status and access to health services.