October 21, 2025 | News
COVID-19 Pandemic: Facebook Ads for Survey Recruitment
Study investigates whether the content of Facebook survey ads influences responses
Researchers at the Max Planck Institute for Demographic Research (MPIDR) investigated whether the recruitment material used for online surveys during the COVID-19 pandemic influenced respondents' answers. To this end, they analyzed the responses of people recruited via Facebook ads with or without a reference to the virus. The results showed no consistent correlation between advertising images and the wearing of medical masks or handwashing frequency. Instead, the age of the respondents and the month in which the survey was conducted were found to be more closely related to these behaviors.
Images used in the Facebook advertisements to recruit respondents for the “COVID-19 Health Behavior Survey” conducted between March and August 2020 in eight countries in Europe and North America. © Collage created by MPIDR/Photo credits in image
When the COVID-19 pandemic broke out in 2020, many researchers collected data by using online surveys. During the first wave, researchers at the Max Planck Institute for Demographic Research (MPIDR) conducted a large online survey study across seven European countries and the United States, called the COVID-19 Health Behavior Survey (CHBS) (N=120.184). Jessica Donzowa, Daniela Perrotta, and Emilio Zagheni investigated whether the way participants were recruited – through Facebook ads with or without references to the virus in the ad images – influenced how they responded.
“Using regression analyses with separate models for each country, we estimated the probability of wearing a medical mask, increased handwashing, and the perceived threat posed by the virus. To make sure our results were not influenced by other factors, we also took into account differences such as gender, age, educational attainment, and the month of survey participation,” explains lead author Jessica Donzowa, a researcher in the Department of Digital and Computational Demography at the MPIDR.
Predicted probability of wearing a face mask (green) and increased handwashing (orange) according to advertisement image and country.
Adjusted for sex, age, education and month of survey participation. The vertical line corresponds to the lowest confidence interval and serves as a visual indicator of overlapping confidence intervals. © MPIDR
In their analysis, Donzowa and her colleagues found that the advertising images were not consistently related to respondents' answers regarding mask-wearing, hand-washing frequency, or their perception of the threat posed by the virus. "Rather, the month in which respondents took part in the survey and their age were more strongly related to these behaviors," Donzowa says. When ad images appear to influence responses, more explicit content about the virus was not always linked to stronger protective behaviors or higher threat perceptions. The patterns varied depending on the country and the question. For instance, an image of a couple blowing their noses was associated with greater mask-wearing behavior in Germany, but to a lesser extent in the United Kingdom. “This suggests that surveys using Facebook ads during health crises should take country-specific contexts into account and test materials in advance before final data is collected,” Donzowa explains.
However, the researchers do offer one caveat in their study: social media ads are designed for commercial advertising, not for recruiting participants for scientific surveys. The placement of ads is controlled by algorithms that are difficult to monitor, which makes it hard for researchers to control who sees which survey ad. This uncertainty must always be taken into account when interpreting the results. The overall recommendation is that a careful campaign design is needed in order to account for this type of bias.
Social media ads enable rapid recruitment for surveys. However, users' decisions to participate depend on their interest in the topic. This can affect data quality. "We investigated whether the content of survey ads is related to different responses. Our aim is to use our understanding of these effects to improve the design of social media surveys in future public health crises,” Jessica Donzowa says.
Donzowa, J.; Perrotta, D.; Zagheni, E.:
PLOS One 20:7, e0326884–e0326884. (2025)
Keywords
COVID-19, health behavior, social media recruitment, Facebook survey, ad design, self-selection