Assessing the impact of antiretroviral-drug access on future HIV infections in Sub-Saharan Africa requires identification of the behavioral response. This paper combines geocoded information about the timing of introduction of ARVs in Kenyan health facilities with population surveys to estimate the impact of proximity to ARV providers on adolescent risky sexual behavior. A variety of difference-in-difference strategies yield a range of estimates of behavioral effects on pregnancy rates and self-reported sexual activity among 15-18 year-olds in areas where ARVs were introduced, from small to quite large. A simulation combining estimated behavioral responses with medical evidence regarding HIV transmission suggests increasing ARV access will reduce new HIV infections even with a very large increases in risk-taking.
Supplementary notes can be added here, including code and math.