HIV/AIDS and sexual behaviour in Botswana

Using three waves of the Botswana AIDS Impact Survey (BAIS) collected by Statistics Botswana, we investigate how sexual behaviour, measured by number of partners and condom use, responds to objective and perceived risk; objective risk is captured by the HIV prevalence rate and perceived risk is represented by concern about the likelihood of being infected by HIV. Indicators of knowledge about HIV, which may affect behaviour and risk perceptions, are included. Endogeneity is addressed through instrumental variables, with HIV prevalence rate instrumented by the distance from respondent`s area of residence to the district with the highest rate, while perceived risk is instrumented by duration (in months) between antiretroviral therapy (ART) rollout and the survey date. The HIV prevalence rate has no significant effect on the number of sexual partners or use of condoms but behaviour responds to changes in perceived risk: those who report becoming more concerned about infection or those who perceive their partners to be unfaithful are more likely to report condom use whereas those who report becoming less concerned are less likely to abstain and use a condom. The effectiveness of advice on health behaviour may be undermined by moral hazard.