The Behavioral Effect of Facial Masks During the COVID-19 Pandemic

We test whether a proximate person (subject) keeps a different distance to a proximate person (experimenter) depending on wearing a facial mask.

2020-04-18

2020-04-24

The primary outcomes of the analysis will be the measured distance between the experimenter and the subject standing behind in the line and an indicator for whether or not the subject complies with the distance of 150cm.

The experimenters position themselves at the end of a line to enter a business. As soon as an individual joins the line behind the experimenter, the experimenter records the distance to the next individual in line behind them. Shortly after the measurement, the experimenter leaves to rejoin the end of the line. The experimenter varies whether or not to wear an FFP2 mask according to a predefined protocol.

The experimenters position themselves at the end of a line (e.g. in front of the post office, see experimental protocol for details). The experimenters are adjusting themselves facing away from the line such that the person in front of them stands to their one side and the arriving subject on their opposing side; this allows the experimenters to better register incoming subjects and turn toward them to conduct the subsequent distance measurement. Then, the experimenters calibrate a measurement app on their mobile device and await the subject’s arrival to start the measurement. After the measurement is taken, the experimenters immediately record the observed variables as described below on their mobile device. Experimenters vary whether or not they wear a facial mask according to the experimental protocol.

As described in the experimental protocol, experimenters measure the key dependent variable of this study using an augmented-reality measuring app. Specifically, the distance in centimeters between the experimenter and the subject next in line at the moment of joining the line and standing still will be used to assess to what extent the subject adheres to the public safety recommendation of keeping a distance of at least 150 cm.

Furthermore, visibly accessible data about the subjects is recorded to be used in the analysis. Specifically, experimenters record the subjects’ gender, age, whether the subject is wearing any kind of mask themselves, and whether they are accompanied. If subjects are accompanied, experimenters record how many individuals are joining the subject and whether this group includes children or only adults. Furthermore, the time and date, as well as the address and store of observation, are recorded by each experimenter.

Before each session, each of the experimenters set an even target of observations to be recorded. Half of the observations should be executed with, and the other half without a mask on. The order is decided randomly using a fair coin or any random number generator by the experimenter before starting the respective session. Example: The experimenter sets the number to 20. After tossing the coin, the sessions start with 10 observations with the mask on. Then the experimenter removes the mask and performs another 10 without.

The treatment is randomized within experimental sessions as described in the randomization method.

Yes

The data will be collected by 5 experimenters, each collecting 60 observations. The number of clusters depends on the number of sessions onto which the experimenters will split their data collection. Details are specified in the experimental protocol.

The planned number of total observations is 300.

150 with mask, 150 without mask

Overall 22 observations were collected in a pilot study, thereof 12 without and 10 with the face mask. The initial data showed that subjects kept a greater distance to the experimenter in the mask condition (M=152.5 cm, SD=52.35) compared to no mask condition (M=139.8 cm, SD=34.88), the difference, however, was not significant. Using the pilot data and the power command in Stata, we compute that in a sample with 300 observations, a two means t-test with an alpha of 0.5 and an effect similar in size to the one in the pilot data (raw difference of 12.7cm) and comparable standard deviations would have a power of 0.7. This corresponds to a minimum detectable effect size of about 14.4cm difference in the means for a two-sample t-Test with unequal variances at a power of at least 0.8.