Research

Abstract: Information campaigns are widely used to promote preventive health behaviors. Yet the beliefs they change may not persist, and it is unclear whether these changes drive behavior. I study this question in the context of condom demand in Ghana, where condoms are widely available but underused. In a randomized experiment with 2,151 unmarried male students in Ghana, I estimate the effects of health risk information on beliefs and revealed demand through voucher redemption and examine whether adding information about peers’ views changes these effects. Health risk information raises condom redemption by 4 percentage points immediately and 7 percentage points eight weeks later relative to the control mean of 13 percent, while its effect on beliefs about condom safety is no longer statistically detectable after eight weeks. Correcting peer misperceptions moves beliefs toward the truth but does not increase demand. The persistence of the health information effect is consistent with experiential learning, whereby an initial information shock induces demand and subsequent experience may sustain it even as the initial belief change is no longer statistically detectable.

Abstract: Understanding what drives people to get tested for HIV is essential for designing effective communication strategies that promote test uptake. In this study, I use a randomized experiment to examine whether and how the format of information affects HIV testing behavior among university students in Ghana. Providing factual information on HIV incidence and the availability of nearby testing services increased actual testing rates by about 1 percentage point from a near-zero baseline. In contrast, adding a story about the testing experience to this statistical information did not generate any additional effect. Financial incentives, introduced non-randomly, raised testing rates to 11 percent. Interestingly, the impact of the original information treatments diminished when a financial incentive became available. Analysis of belief outcomes indicates that the information treatment primarily worked by increasing awareness of local testing services and correcting misperceptions about peer testing behavior, rather than by heightening perceived risk. However, stories did not enhance the treatment effect on beliefs or information recall beyond the impact of simple statistical facts. These results suggest that factual information can effectively address informational barriers to HIV testing in this context, while narrative elements offer no measurable added benefit for influencing this high stakes health behavior.

Abstract: While some studies find that health risk information changes sexual behavior, others find no effect. I synthesize 2,027 treatment effect estimates from 107 randomized controlled trials conducted across low- and middle-income countries to systematically assess the effects of these interventions on sexual health outcomes. A random-effects meta-regression yields an average effect of 0.12 standard deviations (95 percent CI[0.11, 0.13]), which remains statistically significant after correcting for publication bias. This average effect masks a clear gradient along the knowledge-attitude-behavior pathway. Effects are largest for knowledge (0.21 SD), smaller for attitudes (0.12 SD), and smallest for behavior (0.08 SD). The weaker behavioral effects reflect, in part, differences in how much individuals can act on information. Information has larger effects on behaviors individuals can undertake unilaterally, but no detectable effects on behaviors requiring partner cooperation. Effects also attenuate over time, with most of the decline occurring within six months before stabilizing at a smaller but persistent effect. Finally, the magnitude of these effects varies systematically across populations and intervention settings. Overall, the results show that health risk information reliably changes what people know and believe, but its effects on behavior depend critically on whether individuals can act on that information themselves.