In this thesis, I investigate policies and programs to improve reproductive health. My thesis
makes a substantive contribution to reproductive health policy and a methodological contribution to
quasi-experimental research.
In chapter 1, I evaluate the impact of a mobile phone intervention for adolescent girls. I design
and implement a randomized controlled trial in Ghana to test whether sending information via mobile
phones is an effective way to improve girls’ knowledge of sexual health and to ultimately reduce teenage
pregnancy. I find that mobile phone programs are effective not only in increasing knowledge, but also in
decreasing risk of pregnancy among sexually active adolescents. I discuss the results in the context of
sexual education policy in Ghana.
In chapter 2, I explore the complex interactions between migration and reproductive health. I
reconstruct the complete migration and reproductive health histories of women residing in the urban
slums of Accra, Ghana. Using individual fixed effects to reduce selection bias, I find an increased risk of
pregnancy, miscarriage, and abortion in the 48 months after migration, with no significant increase in the
chance of live birth during this time period. With half of abortions in Ghana classified as unsafe, these
results suggest that policies which target the rapidly growing number of urban migrants by providing
access to contraception and public hospital services may reduce unsafe abortion and improve maternal
health outcomes. In chapter 3, I investigate the bias of standard errors in difference-in-differences estimation,
which typically evaluates the effect of a group-level intervention on individual data. Common modeling
adjustments for grouped data, such as cluster-robust standard errors, are biased when the number of
clusters is small. I run Monte Carlo simulations to investigate both the coverage and power of a wide
variety of modeling solutions from the econometric and biostatistics fields, while varying the balance of
cluster sizes, the degree of error correlation, and the proportion of treated clusters. I then apply my results
to re-evaluate a recently published study on the effect of emergency contraception on adolescent sexual
behavior. I find that the study’s results claiming that emergency contraception increases risky sexual
behavior may be spurious once proper adjustments for grouped data are applied.