Experimental Design
To study the impact of a series of customized preventive information against COVID-19 on adoption of preventive behavior and relevant knowledge, we conduct two rounds of phone call surveys to 16,058 households from 135 villages/sub-villages in Bangladesh. After the baseline survey, households assigned to the treatment group receive message(s) encouraging them to adopt specific preventive behavior(s) relevant to COVID-19. We combine this experimental variation in information provision with data on households' social network to study the role within-village social networks play in adoption of preventive behavior.
In the first-round survey conducted between May 8th, 2020 and June 7th, 2020, we randomly selected 20% of households (3,216 households out of 16,058 households) to participate in a long version survey, in which we asked the households about their knowledge of COVID-10, their adoption of preventive measures against COVID-19, and the economic impact of COVID-19 on the household. In the preventive measure section, we recorded households' self-reported practice of preventive measures such as whether they wear masks when leaving home, how often they are able to keep social distancing, how often they wash their hands with soap, and etc. When a household was not able to answer or answered incorrectly, the enumerator delivered a customized message to explain why that particular protection will benefit the household. For example, when a household said they are not washing hands with soap, the enumerator delivered a message that highlights the importance of using soap by discussing the possibility of contracting the virus after touching a contaminated surface.
In the ongoing second-round survey starting from October 27th, 2020, five months after the intervention, we re-approached these 20% of households to record their adoptions of preventive measures against COVID-19, COVID-19 knowledges, and the economic impact of COVID-19. We randomly selected and recorded another 10% of the remaining 80% of households (1281 households) and recorded their responses to these questions. The second group of households did not receive customized information in the first round.