Replication data for: A Reassessment of `Do Teenagers Respond to HIV Risk Information?'

This paper uses simple difference model to exploit the original study's randomization at the school level, comparing only those classrooms that were treated to the controlled. This procedure allows us to reassess the `spillover' problem and to eliminate the bias of the difference-in-difference model in the original paper, because information on the relative risk of HIV infection (RR) affects only 8th grade classes. We then provide a falsification test for the 'placebo effect' theory of treatment spillover by comparing only 7th grade classrooms (the control cohort). Finally, we use subgroup analysis in all of our models in order to disaggregate the true effect of RR from that of RR and TT (HIV prevention teacher training program) together and RR conditional on TT. We find that the RR information program is not as robust as Dupas' original findings suggest, and the treatment is actually harmful because it increases the overall incidence of sex, and it may be especially harmful for boys given their relative risk ratios: The RR program increases the number of both boys and girls who report that they have ever had sex; and, we find no evidence for Dupas' key finding that it decreases the age gap between the girl and her partner. The RR program also increases the number of boys who report not only having sex, but having sex with multiple partners. And there is no corresponding increase in the number who report using a condom.