Mass Media during the Covid-19 crisis: experimental evidence from rural Burkina Faso

The objective of this project is to evaluate the impact of the COVID-19 communication response in rural Burkina Faso. We will focus on the information campaign implemented by Development Media International (DMI), an organization specialized in this type of program. DMI is currently broadcasting one-minute radio spots 10 times a day at peak times on 39 national and local radio stations in Burkina Faso. These radio stations reach an estimated population of 17.3 million people or more than 80% of the population of Burkina Faso. In line with WHO guidance, the spots cover a range of topics related to COVID-19. They include both informational spots and story-based dramas. Some radio stations are also broadcasting longer format interactive radio shows. The campaign started in March 2020 and was still ongoing in September 2020..

2020-04-01

2020-12-01

Our primary outcomes are the following:
- Standardized index of COVID-19 knowledge
- Standardized index of misinformation on COVID-19
- Number of times the respondent washed her hands when coming back home out of the last 10 time she went out.
- Number of recommended preventive measures implemented.

- Standardized index of COVID-19 knowledge : defined as: Index = (total number of correct symptoms mentioned without prompting - total number of incorrect symptoms mentioned without prompting ) + (total number of correct precautions mentioned without prompting - total number of incorrect precautions mentioned without prompting) + (total number of correct ways to be infected by the virus mentioned without prompting - total number of incorrect ways to be infected by the virus mentioned without prompting).
This index will then be standardized using control group’s mean and standard deviation.

- Standardized index of misinformation on COVID-19: Index based on the following variables:
- Misinformation on how COVID-19 is transmitted. Respondents will be asked to mention all ways someone can be infected by COVID-19 to the best of their knowledge without any prompting. We will create a dummy variable equal to 1 is the respondent declared that COVID-19 can be transmitted through mosquito bites, by the wind, by goods or commodities from countries with many cases, by sexual intercourse or by drinking or eating infected food.
- Misinformation on precautions to reduce the risk of infection from COVID-19: Respondents will be asked to mention all precautions people can take to reduce the risk of being infected by COVID-19 to the best of their knowledge without any prompting. We will create a dummy variable equal to 1 if the respondent mentioned one of the following answers: take traditional medicine or herb potion, take antibiotics, malaria drugs or other similar drugs, eat healthy or specific food, drink some specific beverage or don’t drink alcohol.
- Think that the risk of infection of people is her household is null of small for an incorrect reason: incorrect reasons includes: the COVID-19 does not exist, because we are healthy or strong, because we have good blood, because we believe in God or we are protected by God, because our race or ethnic group is not affected by the virus, or because there is a treatment or vaccine. Additional incorrect reasons will be added after the survey based on respondents’ other answers.
To construct the summary index, we will first standardize all variables using the control group mean and standard deviation, second, we will take the sum of all standardized variable and third, we will standardize this sum using the control group mean and standard deviation. We will interpret the impact of the treatment in percentage of the control group’s standard deviation.

-Number of times the respondent washed her hands when coming back home out of the last 10 time she went out.

-Number of recommended preventive measures implemented: respondents will be asked to list all the different measures that they or other household members are implementing to reduce the risk of being infected by the COVID-19 without prompting. We will create a score from 0 to 5 equal to the number recommended measures implemented by the respondent and her household. The list of recommended measures is based on the measures recommended by the ministry of health in Burkina Faso and includes: wash hands regularly with soap or hydroalcoholic gel ; social distancing ; stay home and avoid non-essential travels ; cover your mouth and nose with a mask or a scarf ; and cough or sneeze into your elbow.

Our Secondary outcomes are the following:
- Standardized index of knowledge of COVID-19 symptoms.
- Standardized index of knowledge of how to limit risks of COVID-19 infection.
- Standardized index of knowledge of how people can be infected with COVID-19.
- Knows which population is more vulnerable.
- Knows that some people infected by Covid-19 may not present any symptoms.
- Respondent declared that she has sufficient information on Covid-19.
- Misinformation on how COVID-19 is transmitted.
- Misinformation on precautions to reduce the risk of infection from COVID-19.
- Think that the risk of infection of people is her household is null of small for an incorrect reason.
- Misinformation about the type of people who are more vulnerable to the COVID-19.
- Respondent declared that she maintained a social distance with people outside her household always or most of the time in the last 7 days.
- Respondent declared that other people in her village maintain a social distance with people outside their household always or most of the time.
- Respondent is wearing a mask or a scarf to cover her mouth and nose all the time or most of the time.
- Share of people living in the same village who are wearing a mask or a scarf these days.
- Share of respondents who declare that they would be okay to meet with someone who was infected by the COVID-19 but is now cured.

More information are available upon request or in the pre-analysis plan.

In June 2017, we identified 2851 women who had no radio in their household and randomly allocated 50% of them in a treatment group that received a radio. The radios offered use solar energy which make them particularly adapted to this context where most women don't have access to electricity.
We will identify the impact of listening to COVID-19 information on the radio by comparing women who were randomly selected to receive a radio to women assigned to the comparison group.

The radio distribution was conducted by DMI employees who individually tracked the 1,557 women randomly selected to receive a radio. They conducted this operation between March and June 2017 and were able to distribute 1,130 radios (72%). The main reason 28% of women targeted to receive a radio did not receive it, was absence when the distribution took place. Indeed, during this period of the year, many women travel to other villages where they have fields and are growing crops.
To encourage women to keep the radio for themselves, they were told that DMI would come back after some time to their village, and that women who still had their radio would be eligible for a lottery and could win a small cash prize (around US$ 3.5). The lotteries were organized at the end of the follow-up survey conducted in 2018. Ninety five percent of women who received a radio and were successfully interviewed during this survey still had their radio. For 62% of them, the radio was still functioning properly.

The randomization was conducted in the office using STATA and was stratified on the following variables: village, using modern contraception at baseline and ever attended formal education.

Those women who belong to the same household were grouped together for the randomization. Therefore, the randomization was conducted at the household level (even if most women in the sample are the only one from their household).

Yes

2,562 households

2,851 women who belongs to 2,562 separate households

Treatment group (randomly selected to receive a radio) : 1425 women in 1295 households
Control group : 1426 women in 1267 households

Under reasonable assumptions, the study sample size and design provide sufficient statistical power to detect meaningful effect sizes. If we assume a 20% attrition rate in the survey and that the proportion of respondents with a given misperception about COVD-19 in our control will be 25% we will have 80% power to detect a 4.9 percentage points reduction (intention-to-treat) among the treatment group. This assumption is based on a previous survey on this population, in which we found that 25% of the study population think that modern contraception can make a woman sterile, a common misperception related to that topic. This 4.9 percentage points effect size is similar to our estimate of the impact of a family planning campaign implemented by DMI, on modern contraception uptakes in a previous study we conducted with the same population.