key: cord-0979226-zaw2s25b
authors: DAMETTE, O.
title: Zorro versus Covid-19: fighting the pandemic with face masks
date: 2021-01-08
journal: nan
DOI: 10.1101/2021.01.04.20237578
sha: 69499d2624bc4b401e4e27db0edd85eca288f3da
doc_id: 979226
cord_uid: zaw2s25b

To confront the global Covid-19 pandemic and reduce the spread of the virus, we need to better understand if face mask use is effective to contain the outbreak and investigate the potential drivers in favor of mask adoption. It is highly questionable since there is no consensus among the general public despite official recommendations. For the first time, we conduct a panel econometric exercise to assess the dynamic impact of face mask use on both infected cases and fatalities on a global scale. We reveal a negative impact of mask wearing on fatality rates and on the Covid-19 number of infected cases. The delay of action varies from around 7 days to 28 days concerning infected cases but is more longer concerning fatalities. We also document the increasing adoption of mask use over time. We find that population density and pollution levels are significant determinants of heterogeneity regarding mask adoption across countries, while altruism, trust in government and demographics are not. Surprisingly, government effectiveness and income level (GDP) have an unexpected influence. However, strict government policies against Covid-19 have the most significant effect on mask use. Therefore, the most effective way of increasing the level of mask wearing is to enforce strict laws on the wearing of masks.

of others, and perceptions of others' judgment -especially for young people -have also been found to be significant drivers. Demographic factors (e.g., age, gender) were not found to be significant drivers. The aforementioned study found that self-reported mask use differed considerably across the eight countries considered. In contrast to previous findings, sex-and age-specific patterns about threat perception, confidence in the healthcare system, and the likelihood of adopting preventive behaviors were also documented (6) . Older individual-level studies on the severe acute respiratory syndrome (SARS) epidemic of 2003 in Hong Kong (23) reported that women, people in the 50-59 age group, and married respondents were more likely to wear face masks. Some other studies of different but related nature analysed the practices about face masks/N95 respirators utilization in Poland (24) whereas (25) investigated the social and behavioral consequences of mask policies face with the COVID-19 pandemic using 7000 German participants from April, 14 to May, 26, 2020 . They reveal that mandatory policy is more effective than voluntary policy.

To the best of our knowledge, we present here the first statistical analysis of the effectiveness of mask wearing to reduce the spread of Covid-19 at a global scale. 1) We compute dynamic panel econometric estimates to assess the impact of mask wearing on both infections and fatality rates per capita. Although some data is available on the number of tests being conducted, no official data on mask wearing exists. Previous studies used a dummy variable by taking into account the date of mandatory mask policy introduction and the duration variable (10) . For this study, we collected individual data from the Covid-19 World Survey Data API jointly conducted by the University of Maryland and Facebook (26) . We obtained estimates of the percentage of people in a given country that used face masks daily from April, 23 to July, 15, 2020. We developed a dynamic model that included lagged effects to control for the dynamics of the epidemic over time, including delays between infection and case confirmation and incubation period. Our dynamic model indirectly accounts for potential simultaneous impacts of other determinants of Covid-19 outbreak. We also included additional controls to directly account for certain factors that may have influenced viral transmission (e.g., mobility, temperatures). Finally, we controlled for the effects of other non-pharmaceutical mitigation measures during the studied period by adding the number of Covid-19 tests (testing policy) and a stringency index (reflecting all mitigation measures) as additional explanatory variables. 2) We then performed a crosssectional statistical analysis to examine the socio-economic determinants of mask wearing across countries.

For this analysis, we examined a different set of determinants to understand the heterogeneities regarding mask adoption across countries and the factors that can increase the number of individuals that wear a face mask.

3 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021.

We first conducted an econometric exercise for 96 countries for the period April, 23 to July, 15, 2020. To account for lags in epidemic dynamics, we collected data on infected cases and fatalities for a longer period from January, 1, 2020. Our global panel data covers all parts of the globe, but we narrow this down to European countries for robustness checks.

Our dynamic panel data estimates (Mean Group here ) reveal that masks wearing has a clear (P ¡ 0.01) negative impact on infected cases with a 7-to 28-day lag but that this negative effect disappears after 42 days ( Table 1 ). The most significant negative effect is obtained after a lag of around 27 days (Fig. 1) . After 32 days (the gray area in Fig. 1) , the coefficient associated with the mask use variable becomes insignificant.

We also found a negative impact of mask wearing on fatality rates with a longer lag. Here, we consider 14-and 28-day lagged variables for illustration ( Table 2) . As a consequence, the mask wearing effect is more important on infections in the short term (7 to 28 days), whereas the effect is greater on fatality rates in the medium term (14 to 35 days). This is consistent with the epidemiological rationale that mask protection today can reduce the probability of fatalities after approximately one month, which is based on the dynamics of the epidemic, the incubation period, and the length of time before infection, symptoms, and complications are declared. The lag is thus greater for fatalities than for confirmation of infections.

We conducted a variety of robustness checks (SI Appendix). Considering the fact that the Covid-19 epidemic did not begin at the same time on all continents, we considered European countries for which data was available as a homogeneous and robust sample (SI Appendix, Tables S5-S11). The robustness of the sample is supported by the fact that European countries report a particularly high number of Facebook panel responses for the mask wearing variable.

We also accounted for potential omitted bias and collinearity between mask wearing and other mitigation measures, such as testing policies, school closures, and travel bans. We disentangled the effects of mask wearing from those of other mitigation measures and found a robust negative effect of mask use on the Covid-19 outbreak (Table 3) , taking into account the daily number of new tests per thousand of population and the policy stringency index. The mask wearing effect appears to be robust to the introduction of other controls, such as individual mobility measures -see (27) and (28) for mobility studies -from mobile device location pings (here, we use walking and driving indexes from Apple Trend Reports), temperatures, and other mitigation measures, including Covid-19 testing policy and the Covid-19 policy stringency index.

What are the drivers of mask wearing on a global scale?

As demonstrated in the previous section, face mask use is negatively correlated with infections and fatality rates. The next question is why individuals in some countries are more likely to wear masks wearing than 4 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint (Tables 2 and 3) , we can report some interesting insights.

First, we found that individuals appear to have implicitly understood the crucial role of mask wearing by themselves since we note a convergence to higher levels of mask wearing across countries over time. This is indicated in Figures 2 and 3 by the fact that the distribution of the mask wearing variable moves to the right as we move from the initial period (April, 23 or 24) to the final period (July, 15). In other words, more and more individuals declared on Facebook that they had adopted mask wearing to combat the Covid-19

virus. The increasing dynamics of the pandemics , the communication (both public and private) on the role of masks in reducing transmission, and the changing of habits are potential explanations for this result.

We then estimated the determinants of mask adoption on a global scale by screening a set of potential predictors (see Methods). The cross-sectional results are presented in Tables 4 and 5. We found that population density was positively associated (Fig. 4 , Tables 4-5) with the percentage of individuals wearing a face mask daily. This effect is generally robust (P ¡ 0.01 to P ¡ 0.1). Mask adoption was highest in countries with a high population density and lowest in countries with a low population density. The literature shows 5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 8, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint -6.05e-08*** -5.73e-08*** -1.18e-09** -7.59e-10* (2.26e-08)

(1.73e-08) (5.34e-10) (4.05e-10) new tests (t-14)

-1.58e-07 1.26e-07*** (6.55e-07) (4.23e-08) stringent (t- 14) 9.93e-11 -5.99e-10 (2.21e-08) ( CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; Figure 3 : Distribution of masks wearing across countries on 15th July, 2020 that mask wearing is usually viewed as a complement to social distancing measures; therefore, using masks as a protective measure is more important in densely populated environments (e.g., urban areas, public transport, supermarkets, town centers) than in less densely populated regions (e.g., forests, rural areas). In state capitals and large cities, mask wearing is generally mandatory, especially on public transport. However, when we tested the urban population percentage variable directly, we found the results were less clear cut and that the associated coefficient was never significant. Population density in general seems to be more important than the urban population percentage.

Another main driver of mask wearing on a global scale is the level of pollution. Indeed, pollution levels proxied by CO2 emissions were positively associated (P ¡ 0.01 in most cases) with the proportion of mask users only for high levels of CO2 emissions, but this effect was reversed for low levels of emissions (Fig. 5 ).

It is probable that inhabitants of countries with high levels of pollution are more likely to wear masks to fight Covid-19 because they are in the habit of using them as a protective measure against harmful particles.

Indeed, in some countries, face masks are used to reduce the negative effects of pollution in normal 'non-Covid-19' times. Thus, the marginal cost of adopting mask wearing behavior in those countries is low or null.

No change of habits is required for individuals from these countries as they just continue to wear masks as they had done prior to the pandemic. Recent studies have also shown that mask wearing is especially necessary in highly polluted environments as Covid-19 fatality rates are exacerbated by high levels of pollution (29).

8 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint

The most important predictor of the level of mask wearing in a given population was the Covid-19 stringency index. This index, compiled by (30) , measures the stringency of government responses to the Covid-19 pandemic across the world based on nine indicators: a higher score indicates a stricter government response, here on July, 15, 2020. This index was positively correlated with mask wearing, and this effect was particularly robust (P ¡ 0.001). In other words, when government responses are stricter, mask use is significantly higher (Fig. 6 ). If the results we report in the previous section are correct -that is, if increasing mask use reduces the negative consequences of the Covid-19 virus -the more stringent countries may obtain better results in fighting the pandemic. We also tested the lagged stringency index (one month lagged variable) and the growth rate of the index over the month previous to the date of variables counting (July, 15, 2020) .

In addition, we tested whether the proportion of vulnerable individuals in a population affected the likelihood of individuals to adopt mask wearing to protect themselves and others. Linear partial regressions (SI Appendix Tables S13-S16) show that a high proportion of diabetic individuals in a population is positively associated with mask wearing. However, this effect appears not to be robust in a multivariable framework (Table 5 ). Similar way, results are inconclusive regarding the proportion of overweight individuals variable (SI Appendix Tables S13-S16).

Government effectiveness and GDP appear to be negatively correlated (P ¡ 0.01 in most cases) with the proportion of people wearing masks. This is surprising since we expected that rich countries would be more likely to make masks free of charge or subsidize part of the cost and that individuals in these countries would have higher incomes and, therefore, less economic constraints to prevent them from buying the required quantity of face masks to protect them in public and private areas. Regarding the government effectiveness variable, countries with a high level of government effectiveness are generally characterized by better policy formulation and implementation and higher credibility of government commitment, which is likely to increase the probability of mask wearing and higher compliance with government policy. Indeed, it has been found that strong public guidance is necessary to promote and enforce mask wearing (31 and also 25). It should be noted countries with high a GDP and those with a high government effectiveness are likely to be similar (the correlation is around 0.81 between these two variables), which explains why both coefficients are moving in the same direction. To avoid multicollinearity issues, we alternatively added government effectiveness or GDP in the tested models.

We expected that countries with a high proportion of elderly individuals are expected would have a high level of mask wearing, based on previous findings from (23) . However, the sign of association between masks and the proportion of elderly individuals was negative in partial regressions (SI Appendix Tables S13-S16) and not significantly robust in multivariable regressions (Table 5) . There are several possible explanations 9 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint for this: countries with a high proportion of individuals aged over 65 are generally industrialized countries in Europe where, unlike in Asian countries with younger populations , mask wearing is not habitual. It is possible that older individuals, whose flexibility and capacity for adaptability are generally lower, are less likely to wear masks. Countries with a high proportion of elderly individuals are also likely to be more conservative and, thus, their inhabitants may take longer to subscribe to new habits and rules. In a letter, some authors underlined that socially responsible behaviors may not be intuitive and that behavior change takes time (32) . This is probably truer for countries with older populations . Our multivariate regressions results do not indicate any significant negative relationships, which suggests that demographics structure is probably not a key factor affecting mask wearing disparities across countries.

It is also crucial to consider how the current dynamics of the pandemic might influence the habits of individuals regarding mask wearing and the adoption of other social measures to reduce the transmission rate.

Public information on the dynamics and intensity of the Covid-19 epidemic is probably an important driver of mask wearing, and this variable indirectly captures this effect. When the number of cases is increasing, individuals are more informed and aware of the epidemic's negative consequences and are, therefore, probably more inclined to take control measures such as mask wearing and social distancing more seriously. The higher the rate of infected cases, the higher the probability that individuals will personally know infected individuals (e.g., neighbors, family members, work colleagues). This may make individuals perceive the virus as more frightening and lead to increasing levels of social responsibility, making individuals more likely to wear a mask and obey rules to reduce the transmission rate. We found that the link between the infection count (we used the official count on July, 15, 2020 but also tested a lagged variable) and mask wearing was positive but not always significant at high confidence levels (Table 5) .

Finally, we attempted to evaluate whether individuals in countries with higher levels of altruism, solidarity, and tolerance were more likely to use masks. A scatter plot showed a very modest positive association between mask wearing and the recent (2018) altruism index of Falk and co-authors (33) . However, we found no clearcut correlation, especially in our multivariable analysis (Table 5 ). We also tested the effect of the tolerance index from the World Value Survey database (34) . Tolerance level was significantly associated with mask use for countries with low to intermediate tolerance values, but the association was negative for countries with high tolerance levels (SI Appendix, Table S14 ). This effect may be due to the fact that Scandinavian countries, which are characterized by high levels of tolerance, have implemented liberal policies on mask wearing . We also test if the trust in politicians and government can affect the mask wearing proportion in line with (5) but do not find a clear significant relationship: in other words, we do not find that a high proportion of people that do not trust in government 'at all' is associated to a weak proportion of mask wearing in the population but due to the limited number of observations, the inference work should be interpreted 10 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint Table S20 ). More generally, note that the number of available observations for regressions including altruism and tolerance variables is limited (around 50 observations).

Although we expected that education levels might affect the learning of useful hygiene rules and suitable behaviors to confront contagious diseases, education level (schooling variable from the World Bank or Pisa score) was found to have a non-significant negative effect on mask wearing (see Appendix Table S17 ).

In sum, our results complement those of previous studies, extending them from the level of individual surveys to the country scale. While previous studies investigated factors such as age, gender, or threat perception, our study explores the role of country-level macroeconomic and socio-economic determinants in explaining the substantial heterogeneity across countries regarding the wearing of face masks.

Understanding the links between face mask use and Covid-19 is crucial since the effectiveness of masks has been widely debated and contested among the general public. Our panel econometric exercise demonstrates that the wearing of face masks is negatively associated with infections and fatalities at the country level.

Therefore, mask wearing has an important role to play in controlling the spread of Covid-19. Given the effectiveness of mask wearing in significantly curbing the transmission of the airborne Covid-19 virus and, thus, reducing the number of infections and fatalities, it would be helpful for governments to focus more 11 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint 12 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint on promoting the use of masks than on invoking the precautionary principle. We document a process of convergence over time in favor of mask adoption by showing the increasing percentages of people wearing masks across countries. This is good news as it suggests that individuals have implicitly understood the substantial impact of mask use on fighting the Covid-19 pandemic. However, our data shows that the mean proportion of inhabitants wearing masks was only 56% on July, 15, with a large standard deviation. Even in countries with high levels of mask use, the number of individuals that avoid mask wearing is probably still too large.

Implications for policy Appropriate public hygiene and control policies would consist of mandatory mask wearing. Given the clear effect of mask wearing on infections and fatalities and the fact that a mandatory mask policy involves little economic disruption (19) and is economically attractive, we also recommend extending mandatory face mask use even for children (over 6 years of age). Attitudes regarding altruism, tolerance, and solidarity do not appear to be sufficient to achieve the necessary levels of mask wearing. In contrast, strict government responses significantly increase levels of mask use. In this respect, our results are in line with the findings of (25), which show that a voluntary policy leads to insufficient compliance. In addition, since the absence of economic constraints that might prevent the purchase of masks in high-income countries seems not to be a key driver, the main reasons for not wearing masks appear to be subjective, cultural, or the result of a lack of legal sanctions. Legally requiring people to wear face masks, thus, appears to be an effective instrument (22) . Therefore, some countries could introduce stronger penalties for not wearing masks. Indeed, the only effective way of enhancing mask adoption and saving more lives appears to be to implement more stringent policies in line with some Asian countries, such as Singapore.

We conducted an original empirical work based on a 96 countries dataset between the first of January and the 15th of July 2020. For the first section, we obtain a panel with 96 countries and around 7359 observations. We collected (1) the number of confirmed COVID-19 cases and deaths for the countries in our sample from the European Center for Disease, Control and Prevention between 1st January 2020 and July, 15th 2020, 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. 

To test our theoretical assumptions, we use a dynamic panel econometric model as follows:

Where the subscripts i and t represent country index and periods (days) respectively. The dependent variable, perspective of the dramatic outcomes of the pandemic. Note that when y i,t is the fatality rate, we also add the ratio of infected cases per capita in our benchmark specification in order to account for the fact that the level of the pandemic can impact the fatality rate. The reason behind is to control for a level effect and a 16 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint kind of saturation effect of the health system (too many infected individuals to manage is likely to finally increase the fatality rate).

Equation (1) can be estimated by the mean group (MG) estimator introduced by (37). Both estimators are relevant for macro panels such as the one used in this paper: T is equal to 84 and is thus close to N = 96 (see SI Appendix). We also used the dynamic fixed effects (DFE) estimates and many robustness regressions and tests to validate the consistency of our results (SI Appendix, Tables S10-S11). The Mean Group (MG) estimator consists in estimating each regression separately for each panel member i (country here) with a minimum of restrictions. All estimated coefficients are heterogeneous and are subsequently averaged across countries via a simple unweighted average (37-39). An intercept is included to capture country fixed effects as well as a linear trend. In the dynamic fixed effects (DFE), the slopes are homogeneous but the intercepts are allowed to vary across countries (37-39).

Although we apply appropriate macro-panel estimators to our data, several issues can nonetheless emerge.

First, using dynamic models, we are vulnerable to the so-called Nickel bias. Here, this bias is relatively negligible, notably considering the important time length of our series. Second, panel regressions may be exposed to an omitted variables bias. It would be possible to include control variables such as other control measures (e.g. testing, travel controls) or structural determinants (e.g. population density and demographics such as the population over 65, tourists flows, GDP per capita, and measures of health infrastructures).

Considering the so-called problem of 'bad controls', our set of explanatory variables is assumed to be restricted to the mask variable in order to avoid an over-controlling problem (40). In addition, considering data availability and the fact they are time-invariant variables, we capture these unobservables via the lagged term y i,t−1 and above all with country fixed effects. Another identification issue is related to the potential reverse causality bias related to our Covid-19 variables: news about contemporaneous dynamics of the Covid-19 outbreaks and counts can change the human behavior in real time and the social distancing. This is why lags of dependent variables must be added in our model. Finally, persistence and multicollinearity are other usual issues in panel studies. We have controlled for both by computing autocorrelations LM (Lagrange Multiplier) tests and VIF/Tolerance ratios after each estimated regression. In SI Appendix, we also consider endogeneity issues and other tests related to the specification of our econometric model, the choice of an alternative estimator, and several changes in the sample composition.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint

We collect data from different sources (SI Appendix) to estimate a very simple cross-section model of the following form at a country level:

where y i captures proportion of individuals declaring wearing a face mask on 15th of July, 2020 and X i is a set of control variables including population density, CO2 emissions (and squared CO2 emissions), GDP level (in logarithms), government effectiveness index, altruism index, tolerance index, Covid-19 policy stringency index, urban population percentage, education level, diabetic population percentage and overweight population percentage have been considered. We also assume that the dynamics of the epidemics proxied by the count of cases is also a potential driver of masks wearing propensity. See SI Appendix for complete information about definitions and sources of the data. We also test as a robustness check the same model with mask wearing data on April, 23, 2020 corresponding to the first available and oldest data from the Maryland survey (SI Appendix). Other variables (overweight part of the population, testing, surveillance, travel restrictions and school closures policies) have been introduced (SI Appendix, Tables S18-S19).

Data Archival SI dataset in .dat format and codes are available upon request.

SI Appendix is available at https://sites.google.com/site/olivierdamette/research . 18 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.04.20237578 doi: medRxiv preprint

Face mask use in the general population and optimal resource allocation during the COVID-19 pandemic

Trust and Compliance to Public Health Policies in Times of COVID-19

Face masks for the public during the covid-19 crisis

Behaviors and attitudes in response to the COVID-19 pandemic: insights from a cross-national Facebook survey

More than Words: Leaders' Speech and Risky Behavior During a Pandemic

Mass masking in the COVID-19 epidemic: people need guidance

Association of country-wide coronavirus mortality with demographics, testing, lockdowns, and public wearing of masks

the-community-during-home-care-and-in-health-care-settings-in-the

Social and behavioral consequences of mask policies during the COVID-19 pandemic

Physical distancing, face masks, and eye protection to prevent person-to-person transmission of SARS-CoV-2 and COVID-19: a systematic review and meta-analysis

Surgical Mask Partition Reduces the Risk of Noncontact Transmission in a Golden Syrian Hamster Model for Coronavirus Disease 2019 (COVID-19)

A modelling framework to assess the likely effectiveness of facemasks in combination with 'lock-down' in managing the COVID-19 pandemicProc

Identifying airborne transmission as the dominant route for the spread of COVID-19

Face masks considerably reduce COVID-19 cases in Germany: A synthetic control method approach

Causal Impact of Masks, Policies, Behavior on Early Covid-19

Respiratory virus shedding in exhaled breath and efficacy of face masks

On coughing and airborne droplet transmission to humans featured

To Wear or Not to Wear? Factors Influencing Wearing Face Masks in Germany during the COVID-19 Pandemic

Factors influencing the wearing of facemasks to prevent the severe acute respiratory syndrome among adult Chinese in Hong Kong

Face Masks Use in the Public Domain and its Determinants During the SARS-Cov-2 Epidemic in Poland; a Non-Participatory Covert Observational Study

Social and behavioral consequences of mask policies during the COVID-19 pandemic

Weights and Methodology Brief for the COVID-19 Symptom

Social distancing responses to COVID-19 emergency declarations strongly differentiated by income

Estimating effects of physical distancing on the COVID-19 pandemic using an urban mobility index, medRxiv

Air Pollution Exposure and COVID-19

Oxford COVID-19 Government Response Tracker

To mask or not to mask: Modeling the potential for face mask use by the general public to curtail the COVID-19 pandemic

Do face masks help? is not the question

Global evidence on economic preferences

World Values Survey: Round Six -Country-Pooled

Estimating the COVID-19 infection rate: Anatomy of an inference problem

The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application