Abstract

Introduction

During its first six months, the global COVID-19 pandemic killed more than a million people. Reliable data about the contagiousness of the virus and exactly how it was transmitted was initially scarce and conflicting, but by June 2020 both the Centers for Disease Control (CDC) and the World Health Organization (WHO) were sending clear and consistent messages that behavioral measures such as staying at least 6-feet (~2 meters) from others who don’t live with you, regularly washing your hands, and wearing a facial mask covering your nose and mouth were effective ways to curb the transmission of the virus. By September 2020 more than eight-in-ten U.S. adults (85%) reported wearing a mask all or most of the time when they were inside stores or businesses, and 74% of U.S. adults reported always wearing a face mask when they were outside of their homes [3,4]. Despite an initial slowing during widespread lockdown and stay-at-home orders, COVID-19 infections continued to grow, and by October 2020 had been transmitted to more than 7.4 million Americans, killing 210,000. As the world awaited safe and effective vaccines, individual behaviors were the frontline defense against the spread of COVID-19.

Despite evidence from scientific and global public health data about its effectiveness, wearing a mask or cloth facial covering became a source of contention in the U.S. and elsewhere. Statistical models of per capita COVID-19 mortality rates during the onset of the pandemic showed that deaths were highest in regions that were slow to adopt mask-wearing mandates. This was further illustrated by data regarding the low levels of coronavirus mortality seen in Asian countries that adopted widespread public mask usage early in the outbreak, either via mandate or social norms [5].

In the United States and abroad, legal mandates were the strongest predictor of mask-wearing, but there was also systematic variation by level of education, gender, age group, race/ethnicity, and political affiliation. Survey data showed that:

• In the U.S., college graduates were more likely to report having worn a mask all or most of the time (76%), compared with those without a college degree (60%).

• Women (87%) were more likely than men (83%) to report wearing a mask.

• Adults over 65 reported wearing masks at higher rates than younger age-groups.

• Whites were less likely to report wearing masks than other racial and ethnic groups.

• Democrats were more likely than Republicans to report wearing masks (76% vs. 53%), and conservative Republicans were among the least likely to say they have worn a mask all or most of the time (49%) compared with 60% of moderate Republicans [3,6].

Psychologists have stressed that psychological factors are essential to understanding and responding to COVID-19. European data reveal that psychological factors such as perceptions of risk, the observed behavior of others, concern for self-protection and protecting others, and perceptions of the judgment of others when wearing a mask were also found to be significant predictors of mask wearing [7,8]. Self-reported concerns about the behavior and judgement of others suggests that social norms may influence decisions about when, where and with whom to wear a mask in public. A large body of evidence from social psychological research attests to the power of normative information about close others and individuals in the immediate vicinity to shape individual attitudes and behavior [9–12]. While the majority of Americans claimed to regularly wear a mask in public spaces, they report substantially lower rates of mask-wearing by others. For example, only 44% of Americans said that all or most of the people in their area have been wearing masks in stores or other businesses [6].

Prior research has established the causal influence of normative beliefs on behavior [13]. From classic research on conformity, to more contemporary research on normative beliefs and social marketing, behavioral research has shown that perceptions about the common and accepted behaviors of others can produce conformity and pressure to align one’s behavior with that of a group [10]. Research on normative social influence has been conducted across a range of behavioral domains, including environmental protection, gambling, paying taxes, and health behaviors such as alcohol and tobacco use. While the influence of social norms has been well-established, the current COVID-19 situation provided an opportunity to study a novel behavior (mask-wearing) in a context where the behavior had both a direct personal benefit (self-protection) and social benefit (slowing the spread of the virus).

Observation of public mask-wearing and the rate of mask-wearing in the immediate social context can advance our understanding of normative social influence, and directly inform efforts to promote wide-spread behavioral adoption. Two overarching questions guided this research: how often do adults correctly wear a mask in public spaces, and to what extent are people influenced by the mask-wearing behavior of proximal others? Specifically, we predicted that observed mask-wearing would be significantly lower than self-reported rates, and positively associated with the behavior of both proximal others and distant others in the immediate vicinity.

Methods

Restricted guidelines covering human subjects research during COVID-19 require novel approaches to collecting behavioral data. This study used structured field observations of mask-wearing behavior via livestream public webcams, with IRB approval from California State University San Marcos (CSUSM). Given the nature of collecting observational data, informed consent was waived.

Location

Nine livestream public webcams across the U.S., the U.K., and Ireland were selected as locations for data collection. These locations represented all available livestream public webcams positioned such that the mask-wearing behavior of participants entering the frame was clearly visible (www.earthcam.com). Data were successfully collected from seven locations: 1) Bourbon Street, New Orleans; 2) Abbey Road, London; 3) Times Square, 4) and Little Italy, New York City; 5) Duval Street, Key West; 6) Essex Gate, Dublin; and 7) Freemont Street, Las Vegas (see Fig 1). Webcams in Miami and New Jersey were also selected but the cameras were not operational during the data collection period. All locations were urban streetscapes in high pedestrian traffic areas; at the time of observation, the U.S. locations had city-wide mandates to wear a mask indoors and outdoors in public spaces [14–18] while wearing masks was only mandated indoors for London and Dublin [19,20].

Open in a separate window

Fig 1

Livestream webcam locations.

Results

Overall mask-wearing

Despite contemporaneous national survey data showing that 74% of American adults claimed to wear a mask in public, we observed that only 51% of focal participants across the seven locations were correctly wearing a mask covering their nose and mouth. Forty-six percent were not wearing a mask, and 3% had a mask around their neck or in their hand or were engaged in an activity that temporarily precluded them from wearing a mask (e.g., eating or drinking). We noted that none of the webcams captured visible signage encouraging mask-wearing or social distancing. A chi-square goodness-of-fit test revealed that the observed incidents of mask-wearing was statistically significantly lower than the self-reported rates of 74%, χ²(1) = 315.63, p < .001. Our analyses of mask-wearing are restricted to focal participants in the no mask or mask worn incorrectly (coded 0), and mask worn correctly (coded 1) categories.

Mask-wearing differed significantly across location, ranging from 90% of focal participants wearing a mask correctly in New York’s Little Italy, to fewer than 20% in Dublin (See Fig 2). The intraclass correlation coefficient (ICC) across the seven sites was .49 indicating variability both within and between locations. Mask-wearing was more common in the U.S. locations (59%) where mask-wearing was mandated for outdoor public spaces compared with the European locations where mask mandates applied only to indoor public spaces (26%), χ²(12) = 29.95, p < .001. However, the observed rate of 59% of the U.S. participants still fell short of the expectation of 74% obtained from survey data, χ²(1) = 30.70, p < .001. In line with national survey data, mask-wearing differed by estimated sex, with 58% of females compared with 48% of males correctly wearing a mask, χ²(1) = 12.38, p < .001. Mask-wearing also differed as a function of the density of the crowd (and ability to social distance), with 70.3% of participants wearing a mask in a crowded situation, compared to moderately or sparsely crowded situations (45% and 53% respectively), χ²(2) = 18.92, p < .001. Mask-wearing did not vary by age group in our sample, χ²(2) = 3.97, p = .137.

Open in a separate window

Fig 2

Mask-wearing by location.

Discussion

In the months following the start of the worldwide COVID-19 pandemic, the primary tool for slowing the viral spread and protecting public health involved changes in individual behavior. Among the behaviors recommended by health professionals, one of the most important was wearing a mask in public. Our analyses of the mask-wearing behavior of individuals across seven U.S. and European cities showed that significantly fewer individuals correctly wore a mask covering their mouth and nose while in public spaces than was reported by national surveys.

Our reported results examined the role of descriptive norms on mask-wearing behavior. Across the seven public sites included in our study, the intraclass correlation coefficient (ICC) was .49. The ICC is calculated as the ratio of variability between groups, divided by the sum of the variability both within and between groups. An ICC of zero would indicate no variability between groups, and a high degree of variability within groups. In contrast, an ICC of 1.0 would indicate variability between groups, but no variability within groups. In our current situation, the ICC of .49 indicates that there was more variability between groups than would have occurred at random.

We interpret this as evidence for conformity. However, an alternative explanation is that individuals within a group (in our case, cityscape) are affected by the physical and social aspects of their surroundings, including political inclinations of the region, local restrictions, and national mandates. Thus, the observed ICC across the selected sites may be due to shared features of the site, rather than to social norms. However, importantly, our reported results also showed clustering of behavior within each site. Focal individuals in groups tended to match the mask-wearing tendencies of the rest of the group. Mask wearers were in groups with an average mask-wearing rate of 84%, while those without a mask were in groups with an average mask-wearing rate of 34%. This second finding lends more support for our conformity interpretation.

This second finding showing a greater tendency for the mask-wearing behavior of focal individuals to match that of others in close proximity can also be explained as homophily—that is, the tendency for individuals to self-select into groups based on similar preferences. However, our third finding provides some additional support for our conformity interpretation in that distant others in the same setting showed behavioral similarity. We found that focal participants who were wearing a mask correctly were in social contexts that averaged 33% mask-wearers; focal participants who were not wearing a mask were in social contexts that averaged just 19% mask rates.

Our results suggest that an individual’s decision to engage in behaviors that can help slow the spread of a virus—such as wearing a face mask in public—has an important social component. Despite the clear personal and social benefits associated with wearing a face mask in public, a substantial number of individuals in our observations of public spaces chose not to wear a mask. These findings have important implications for efforts to promote health protective behaviors. The primary tool used by health official during the pandemic focused on protecting oneself and others. While such messages were influential—as indicated by the 51% of observed individuals wearing masks in our observation sample—these health-based messages ignore an important social consideration.

Based on our findings, public health communications that highlight the large number of individuals who are wearing masks can go a long way toward creating a support context for the target behaviors. Indeed, messages that highlight the large number of individuals who deviate from the norm have been shown in other studies to reduce conformity [12]. Instead, images and messaging highlighting widespread support for targeted public health measures can serve to amplify the protective personal and social health benefits. Similarly, messages that highlight the trending norm of increased behavioral adoption have also been found to promote higher rates of conformity [23,24].

In conclusion, behavioral data collected from public livestream webcams demonstrates significantly lower rates of mask-wearing than indicated by national survey data. This underscores the need to promote mask-wearing, and strongly indicates that the behavior of close others and strangers in the social context may have a potent influence on the decision to wear a mask.

Acknowledgments

Funding Statement

The author(s) received no specific funding for this work.

Data Availability

All data used in this study are available from the OPENICPSR database at http://doi.org/10.3886/E144983V1.

References