At the beginning of the pandemic, public health officials waffled over masks: Do they help? By how much? The World Health Organization declined to recommend mask adoption until June 2020, for instance, partly because of a lack of real-world studies of their efficacy (although laboratory experiments clearly show masks partly block viruses). As recently as July 2021, City Journal published an article arguing that “the best medical” research suggests mask-wearing during the pandemic “has likely provided little to no health benefit.”
Because of the lingering uncertainty, we set out last year to conduct the first large-scale study to answer two questions: Can community-wide masking norms be changed in the real world? And does wearing masks measurably reduce community-level coronavirus infections? We now have the results, and the answer to both questions is an emphatic yes.
To the extent there was still a debate about masking, it should be over. Just as importantly, we have a set of tools that have been proved to increase mask usage substantially, at least in the setting we studied.
The study was a randomized controlled trial that involved 600 villages in Bangladesh and 350,000 people. Bangladesh, with a population of 165 million, already had a mask mandate, but (as in many places around the world) it was widely flouted.
For roughly half the 600 villages, we intervened by promoting mask usage: We gave every family in each village at least three masks — one per adult, cloth or surgical — and distributed masks in public spaces, including markets. We enlisted eminent local figures to explain why wearing masks can help slow viral transmission — providing imams with scripts to be read during religious services, for instance. Local leaders also modeled proper mask-wearing. We then sent observers into public spaces to note any changes in behavior.
Simultaneously, we surveyed people in these villages to see if they or their family members had symptoms of covid-19, and we took blood samples — if people consented — to confirm the presence of coronavirus antibodies.
The results were striking. Our interventions tripled the proportion of people wearing masks: 13 percent of people wore them in the control villages, 42 percent in the “intervention” villages. This in turn led to a 9 percent reduction of community-level symptomatic coronavirus infections. Villages where surgical masks were distributed appeared to be especially protected — an effect concentrated in the most vulnerable populations. We found that surgical masks averted 1 in 3 symptomatic infections among those aged 60 and older.
Some public health officials have worried that mask use leads people to change their behavior in ways that make it riskier — doing less social distancing, for example. Our observers found no evidence of this; in fact, distancing was slightly more common in the intervention villages.
The study took place between November 2020 and April 2021. Interventions in the villages lasted eight weeks, and we continued observations for two weeks after that (and then followed up at roughly five months). Mask use persisted throughout the 10 weeks but had fallen by the five-month mark. Even then, masking remained more common where we had promoted masks: 22.4 percent versus 14.1 percent in the control villages. (In the next phase of the study, we will return to a subset of villages to see if the numbers can be brought back up, at low cost.)
Given how polarized the debate over masking has become in the United States, it is far from clear that a similar implementation strategy would work here. But regardless of the promotion strategy, the results show that masks should remain a central tool for reducing covid-19 everywhere, particularly since the delta variant’s prevalence means that even vaccines do not fully prevent transmission. Masks are especially important in countries such as Bangladesh where vaccine distribution has been slow — and in countries (again, like Bangladesh) where population density makes physical distancing difficult. That describes a large swath of the globe.
The project’s immense scale demanded substantial international collaboration: The nongovernmental organization Innovations for Poverty Action, in Bangladesh, conducted the surveys and did the behavioral observations, while the group Greenvoice distributed masks. (The other lead investigators, besides the authors of this article, were Jason Abaluck of Yale University, Stephen Luby of Stanford University and Laura Kwong of the University of California at Berkeley.)
The randomized controlled trial also helped us learn which approaches didn’t work. Among other things, twice-a-week text reminders, asking people to pledge to use masks and financial or other rewards given to community leaders for meeting targets had no effect.
Masks may have even greater potential than our study was able to demonstrate. First, we persuaded only about a third of the population to change their masking behavior. It is possible that more aggressive efforts could lead to even more change and produce greater health benefits. Second, this research was conducted in rural areas where life is generally lived outdoors. Masks should work even better in places where people congregate inside more frequently.
Moreover, our study was designed only to track symptomatic infections. It seems likely that masks reduced asymptomatic infections as well, further curbing transmission, although we do not yet know for sure. We have designed a new trial to provide a more definitive answer to this question.
The study was also conducted at a time when the alpha variant was the dominant strain in Bangladesh. With the increased transmissibility of the delta variant, we anticipate that masks may be even more effective now in reducing covid-19 — something we will also be assessing in our forthcoming trial.
We took ethical questions seriously in this study. Anyone we asked to go into villages had to be clear of covid-19 symptoms with no recent exposures to people with such symptoms; they also masked up, avoided indoor spaces and practiced physical distancing, where possible. We scrutinized the messages we used to make sure they did not shame people who did not wear masks; we concluded that our messages did not conflict with local religious beliefs, whether Muslim or Hindu.
Most importantly, as soon as the data began to suggest that masking had benefits — months before we drafted and released our study — we began to talk with the World Health Organization, the Bill and Melinda Gates Foundation, the World Bank and dozens of other governmental and nongovernmental groups about scaling up, so that others would benefit.
In Bangladesh, for example, we are working closely with the Bangladesh Rural Advancement Committee, or BRAC, the world’s largest NGO, to implement our model of supplying masks, reinforcing their importance and modeling their use across 35 high-risk districts covering 81 million people in rural areas. Other NGOs and governmental organizations are applying this approach — we call it the NORM model — in Pakistan, India and Nepal. We persuaded philanthropists to donate over 100 million masks to support these efforts.
Our research suggests that if we could get even a quarter more people wearing high-quality masks across Latin America, Africa and Asia, we could change the trajectory of the pandemic. Indeed, the reductions in infections we observed in Bangladesh with surgical masks, if extrapolated to the entire country over the past four months, could have saved nearly 25,000 lives.
Our study showed conclusively that masks are a cost-effective way to reduce infections and demonstrated that a mask-promotion strategy can work. But our broader campaign to spread the word about our approach made clear that building large-scale coalitions across NGOs and governments is essential to converting that information into programs that reach millions.
Until the vaccine supply increases — and possibly afterward — mask-wearing will remain essential worldwide. We all need to think hard about the best ways to marshal resources and apply the findings of this study to specific national and local contexts.