Adapted from “Can Facebook Ads Prevent Malaria? Two Field Experiments in India,” by Dante Donati of Columbia Business School, Ana Maria Muñoz-Boudet of the World Bank, Victor Orozco-Olvera of the World Bank, and Nandan Rao of Virtual Lab and UAB.
Key Takeaways:
- Social media campaigns had a small but significant impact on the use of bed nets to avoid malaria transmission in India.
- That impact was largely on urban areas and households that already had a low malaria risk.
- The researchers recommend that governments and NGOs assess their target demographics before launching public health campaigns and adjust their targeting and engagement goals with those results in mind.
In 2020, at the height of the pandemic, governments and nongovernmental organizations (NGOs) were particularly interested in determining the effectiveness of public health campaigns. At the time, Dante Donati, assistant professor of business in the Marketing Division at Columbia Business School, had been working with the World Bank for six years to evaluate the impact of health campaigns in different media — including an extensive study in Nigeria analyzing the efficacy of sexual and reproductive health videos in efforts to change teenagers’ behavior.
“We observe that a lot of governments and organizations around the world are leveraging social media for their public health objectives,” says Donati. “But these types of ads are very poorly understood because there's a lack of data. This is very different in commercial marketing. When a company tries to sell a product online, there are standard techniques to link that sale to ad exposure from social media marketing. For public health, it’s much more complicated because the outcomes are usually off-line and longer term.”
During the COVID lockdowns, Facebook approached Donati’s research team: “They were launching a massive campaign in India to encourage people to take preventive and response behaviors to avoid malaria, and they wanted to evaluate the impact of that campaign,” Donati says.
How the researchers did it: Because mosquitoes are the primary vectors for malaria, one of the most effective preventive measures is to sleep under a bed net — mosquito netting that drapes over a bed. “We can use ads to remind or inform people to sleep under a bed net, but then how do you know whether people are actually taking that action?” says Donati. “That is the main challenge we face in this research.”
To overcome that hurdle, Donati’s team recruited 8,000 Facebook and Instagram users to answer surveys. The researchers segregated respondents according to their disease risk, as defined by their type of dwelling. Half of the respondents lived in low-risk “solid dwellings,” typically made of brick or cement, while the other half lived in higher risk “non-solid dwellings,” made of straw, mud, or wood.
Donati says, “We started asking questions: Did you sleep under a bed net last night? How many people in your household used the bed net last night? Did you have a high fever, and did you seek medical help within 24 hours?” By collecting these answers systematically, the team was able to measure the outcomes over time.
The researchers then needed to link those results to ad exposure. To do that, they considered 80 districts in India. A randomly selected group of 40 of them received an ad campaign from the Facebook page of the local implementing partner, Malaria No More, and 40 of them did not. “In that way, we were able to attribute actions more directly to the campaign,” Donati explains.
What the researchers found: The social media ad campaigns did have a positive yet moderate impact on the use of bed nets, which increased by 4.5 percent among bed net owners. However, the researchers drew the most interesting conclusions when they compared results across dwelling types, Donati says: “The campaign only increased the likelihood of sleeping with a bed net for households living in solid dwellings — already at lower risk for malaria.”
They also found that members of those households were much more likely to seek medical treatment promptly if they experienced a high fever. “Moreover, after the campaign, we found that self-reported malaria incidents in households at lower risk went down by about 43 percent, but we didn't find any effect among households living in these non-solid dwellings that are more at risk of malaria,” says Donati. “So we face very strong and significant heterogeneous effects across dwelling types.”
The researchers then compared those results to publicly available health data in India. They found that there was a reduction of malaria incidents only in urban areas, but in those areas, there was a 30 percent reduction in India’s incidence rate. “So overall, we find that the campaign works,” says Donati, “but only among a specific subgroup, which is less at risk of the disease.” The results were not driven by the ad content being ineffective for rural populations but rather by the insufficient reach of ads in those areas.
Why it matters: “Our results help illuminate how to target those at risk — how to make sure that the intended audience for the campaign receives the ads,” says Donati. For public health campaigns to be successful, it’s important to understand how the target audience engages with social media and who is more likely to respond to these ads. “Fifty percent of the world uses social media,” Donati says, “but what about the rest?” He recommends using survey research to determine who you’re likely to reach before launching a health campaign.
The research also highlights the importance of privacy policies in public health campaigns. “By restricting access to data that allow better targeting, recent privacy policies might hinder the success and assessment of these campaigns,” he notes. “Our methodology integrates these privacy considerations and shows how information directly collected from users (first-party data) can be used to inform targeting and help reach the most valuable segments of the population — in our context, those most at risk of malaria.”
Donati also recommends avoiding typical engagement objectives for these types of campaigns. “Facebook optimizes for interactions such as clicks, likes, and comments,” he says. “But you might miss your target audience if its engagement is low. When you can, opt for reach, not engagement.”
