Do Teenagers Respond to HIV Risk Information? Evidence from a Field Experiment in Kenya

Publication Details

American Economic Journal: Applied Economics, January 2011, v. 3, iss. 1, pp. 1-34. Available From:

Link to Source
Pascaline Dupas
Sub-Saharan Africa (includes East and West Africa)
Health Nutrition and Population
HIV and AIDS, Primary Health- including reproductive health, Sexual Behavior
Equity Focus
Evaluation design
Difference-in Difference (DID), Randomised Control Trials (RCT)
Journal Article


This study examines whether, and in what way, receiving specific human immunodeficiency virus (HIV) risk information will influence teenagers’ sexual behaviour. The author postulates that teenagers who receive only ‘risk avoidance’ (abstinence) education will make riskier sexual behaviour decisions than teens who receive ‘risk reduction’ (condom use) education. The author also discusses sociological (largely income based) drivers: cross-generational relationships between teenage girls and older sexual partners, who typically have a higher prevalence of HIV infection than teenage boys.
To test the impact of receiving risk reduction information on teenagers’ sexual activity choices, a randomised field experiment (RCT) was conducted in two rural districts of western Kenya, surveying students in 328 primary schools. After stratifying the schools by location, test scores and student sex ratio, schools were randomly selected to (1) teach the official national HIV/acquired immune deficiency syndrome curriculum, which consists of risk avoidance messaging, after additional teacher training (TT) in the curriculum; (2) provide the Relative Risk Information Campaign (RR), including information on HIV prevalence rates by age and gender, which was conducted by a trained project officer from a nongovernmental organisation; or (3) continue to provide the national curriculum, without any additional training. RR and TT treatment and control groups were randomly assigned.
To measure the incidence of risky sexual behaviour, follow-up surveys were conducted to collect self-reported sexual behaviour information, as well as data pertaining to marital status, sexual partner age differentials and incidence of childbearing, which was used as a proxy for unprotected, and therefore risky, sex.
The author estimates the effect of each programme using four different regression specifications: (1) an ordinary least squares (OLS) estimation of the simple difference, (2) a probit estimation of the simple difference, (3) an OLS estimation of difference-in-differences (DID), and (4) an OLS estimate of difference-in-differences (DID) with school fixed effects.

Main findings

The results indicate that receipt of the risk reduction information reduced the incidence of childbearing by 1.5 percentage points, which amounts to a 28 per cent decrease in the childbearing rate for treated girls relative to girls in the RR comparison group, the bulk of which corresponds to a decrease in childbearing outside of marriage. Despite increasing the amount of HIV education actually delivered, the TT programme was shown to have no impact on the incidence of childbearing, which the author posits may be due to the ‘abstinence until marriage’ curriculum.
The results also reveal that the impact of the risk reduction information on the age differential between childbearing partners is both negative and significant at the 1 per cent level. On average, the age differential between RR treated girls and their partners is 1.7 years smaller than for RR control girls and their partners. Coefficients for the TT programme in this area were not statistically significant.
Overall, the results suggest that providing teenagers with relative risk information led to a large decrease in incidences of cross-generational unprotected sex and did not increase incidences of intra-generational unprotected sex. Therefore, the study indicates a positive impact of the RR programme on reducing teenagers’ risk for HIV infection.
The author notes that although the students who completed the follow-up survey were balanced across RR treatment and control groups, a higher fraction of students in the TT control group were represented in the secondary school survey than those from the TT treatment group. Additional limitations to the study include using childbearing as a proxy for risky sex, as it would fail to reflect other types of risky sex, terminated pregnancies or those that are the result of long-term monogamous relationships, and the fact that self-reported data on sexual behaviour is subject to reporting biases.

Scroll to Top