You are on page 1of 18

1210451

research-article2023
PSPXXX10.1177/01461672231210451Personality and Social Psychology BulletinWei et al.

Empirical Research Paper

Personality and Social

People in Tight Cultures and Tight


Psychology Bulletin
1­–18
© 2023 by the Society for Personality
Situations Wear Masks More: Evidence and Social Psychology, Inc
Article reuse guidelines:

From Three Large-Scale Studies in China sagepub.com/journals-permissions


DOI: 10.1177/01461672231210451
https://doi.org/10.1177/01461672231210451
journals.sagepub.com/home/pspb

Liuqing Wei1 , Alexander Scott English2 , Thomas Talhelm3 ,


Xiaoyuan Li4, Xuemin Zhang5, and Shuang Wang6

Abstract
Studies have found large differences in masks use during the pandemic. We found evidence that cultural tightness explains
mask use differences and this association was more robust in tight situations like subways. In Study 1, we observed 23,551
people’s actual mask use in public places around China. People wore masks more in tight situations; however, differences
did not extend to outdoor streets and public parks, where norms are looser. We replicated this finding using a data
from 15,985 people across China. Finally, in a preregistered study we observed mask use with the removal of COVID-19
restrictions, people still wore masks more in tight situations like subways than in loose situations of parks. These findings
suggest that norm tightness has a lasting association with people’s health-protective behaviors, especially in tight situations.
It provides insight into how different cultures might respond with future pandemics and in what situations people adopt
health-protective behaviors.

Keywords
COVID-19, tightness–looseness, social norms, mask wearing

Received October 6, 2022; revision accepted October 11, 2023

In June 2020, the World Health Organization recommended and 73 hospitalized patients (National Health Commission of
wearing masks in public to limit the spread of COVID-19. the People’s Republic of China, 2020a). During the whole
Yet, mask use has varied widely. For example, from April to period of our observation in Study 1 (from May 25, 2020 to
October 2020, mask use was higher in Italy and Japan than in June 28, 2020), there were only 345 new local cases in China.
Denmark and Sweden (Badillo-Goicoechea et al., 2021). Thus, China has entered the “drag on” phase of the virus.
Mask use also varies within countries. An observational In March 2023, we conducted follow-up observations
study in the United States found that people in rural areas (Study 3), at that time China had lifted all restrictions on
(20%) wore masks less than people in cities (47%) and sub- COVID-19. Millions of Chinese people were infected
urbs (49%; Haischer et al., 2020). This study explores why. COVID-19 and recovered from it (Chinese Government,

Vigilance During the “Drag On” Phase 1


Hubei University, China
of the Pandemic
2
Wenzhou-Kean University, China
3
The University of Chicago, IL, USA
4
Two obvious factors that could explain differences in mask City University of Macau, China
5
use are mask policies and objective risk. There is some evi- Beijing Normal University, China
6
The Education University of Hong Kong, China
dence that more people wear masks when COVID-19 cases
are high (English et al., 2022). There is evidence that at least Corresponding Authors:
some people still respond to high COVID-19 risk by chang- Alexander Scott English, School of Psychology, Wenzhou-Kean University,
Daxue Road No. 88, Wenzhou 325060, China.
ing their behavior, even when there is no legal requirement Email: AEnglish@kean.edu
(Lopes, 2022).
Yet, the COVID-19 pandemic has dragged on, through Xuemin Zhang, Beijing Key Lab of Applied Experimental Psychology,
Faculty of Psychology, and State Key Laboratory of Cognitive
periods of lower risk. After the initial outbreak, China intro- Neuroscience and Learning, Beijing Normal University, Xinjiekouwai
duced strict lockdowns and mask mandates. On May 27, Street No. 19, Beijing 100875, China.
2020, Mainland China had only two new confirmed cases Email: xmzhang@bnu.edu.cn
2 Personality and Social Psychology Bulletin 00(0)

2023), but mask use was still a common practice. In this April to September 2020). In Study 3, we observed people’s
study, the “drag on” phase refers to a long period during mask use in March 2023, a time when all mask restrictions for
which the risk of getting infected by COVID-19 is much COVID-19 were lifted in China. We observed 1,287 people in
lower, but the risk is not completely relieved. five Chinese cities and one Portuguese city. This study can
We posit that mask use is a valuable feature of the “drag eliminate the confounds of mask mandates and enforcement
on” phase even after vaccinations, because it is (a) an observ- of mask policies, reflecting people’s true free choices of mask
able behavior representing COVID-19 norm, and (b) in tight use under different situations.
cultures, individuals must follow the recommendations of
the public health officials like wearing masks on public
transportation or indoor shopping malls. On the opposite
Culture Influences Mask Use
end, in a loose culture like the United States, United Studies have linked many factors to people’s preventive
Kingdom, or Denmark, masks might fade out of public behaviors against COVID-19. Perceptions of COVID-19
quicker even on public transportation or indoors, hence it is risk is often a strong predictor (Bruine de Bruin & Bennett,
up to the individuals’ discretion to wear a mask or not. 2020; Mahdavian et al., 2022; Rubaltelli et al., 2020;
However, in tight norm societies, masks still symbolize vary- Savadori & Lauriola, 2022; Schneider et al., 2021). However,
ing degrees of the “drag on” phase and tolerance of residents the factor that comes out as the most important varies across
in these cultures (Lopes, 2022). countries. For example, a study found different results in
Morocco and India (Jadil & Ouzir, 2021). In Morocco, peo-
ple’s perceptions of COVID-19 severity were the best pre-
Mask Policies dictor of protective behaviors. In India, perceptions of the
Mask policies are another obvious cause of differences. For probability of getting COVID-19 were the most important
example, in the United States, many cities have had stricter predictor of protective behaviors (Jadil & Ouzir, 2021).
mask policies than rural areas. In this study, we find evidence However, perceived risk is not the only meaningful pre-
that culture matters beyond mask policies. We do this in two dictor. A study in Germany and United Kingdom found that
ways. First, we measure differences within a single country, personal health risk perceptions, official message quality, use
which holds constant national mask policies. Second, to of official news sources, political orientation, and age all pre-
account for regional differences in mask policies, we gath- dicted people’s protective behaviors (Mahdavian et al.,
ered data on mask policies across the seven cities where we 2022). However, these were all factors at the individual level.
observed people’s mask use. The comparison shows that the Rarely has research looked at culture-level predictors.
policies had few variations between cities (Table S1). Furthermore, most studies used self-report. In this study, we
use objective observations to improve on the limitations of
self-report. Furthermore, we test culture-level factors, not
Study Overview just individual factors.
In three studies, we investigated what economic and cultural Studies have found evidence that mask use differs along
factors were associated with people’s mask use in the pandemic cultural and demographic factors. For example, around the
“drag on” phase in China, and whether the associations were world, people in collectivistic cultures were more likely to
more robust in tight situations like subways than in loose situa- report wearing masks (Lu et al., 2021). Similarly, people in
tions like public parks. We started by observing people’s actual China’s more collectivistic rice-farming areas were more
mask use. Many prior studies have relied solely on self-report likely to wear masks, search the internet for masks, or report
surveys of mask use (such as Lu et al., 2021; Rieger, 2020). wearing masks than people in less collectivistic wheat-farm-
Self-report can be problematic because people may exaggerate ing areas (English et al., 2022). People in more densely pop-
their mask use, especially when there are social norms or laws ulated areas of China were also more likely to wear masks
mandating masks. A study in Kenya found that 90% of people (English & Li, 2021).
in phone interviews said they wore masks in public, while in In this study, we test whether tight social norms explain
direct observations, only 5% of people wore masks correctly differences in mask use. “Norm tightness” refers to the
and another 5% of people had masks with them but did not strength of social norms across cultures (Gelfand et al.,
wear them correctly (Jakubowski et al., 2021). 2011). Although all cultures have social norms, research has
In Study 1, we measured people’s actual mask use in pub- found that norm tightness varies across nations (Gelfand
lic across China. We observed 23,551 people in 15 districts et al., 2011). For example, Pakistan scored as the tightest cul-
across seven cities. We measured mask use for 5 weeks, start- ture in a study of 33 nations (Gelfand et al., 2011). Another
ing May 25, 2020. Then we tested whether this pattern would tight culture is Singapore, which is famous for caning thieves
hold in a large-scale survey (Study 2). The survey asked and killing people for carrying marijuana. In contrast, Brazil
15,985 people across China to report their weekly mask use and Australia have looser norms.
when going outside. Like our observational study, this survey Tight cultures have more rules about what behaviors are
covered the “drag on” phase of the pandemic in China (from acceptable, and they have stricter punishments for deviant
Wei et al. 3

behaviors. In contrast, loose cultures have fewer norms lim- tight norms matter more in some contexts and less in others.
iting people’s behaviors and more tolerance for deviation. Research has found that, even in the same society, there are
Although cultures have norms about specific behaviors, loose situations versus tight situations (Kitayama, 2002;
there is evidence that cultures have tighter or looser norms in Price & Bouffard, 1974). To test this, we measured mask use
general, across domains (Gelfand et al., 2011). In this study, in a range of contexts, from places where norms would pre-
we examine whether people’s mask use differs based on tight sumably be tighter (such as the subway) and places where
norms in general, rather than specific norms about masks. norms would presumably be less tight (such as an outdoor
public park).
Among everyday situations, people have rated job inter-
Norm Tightness Differs Within China
views and funerals as the tightest situations (Gelfand et al.,
Tightness also varies within nations. A study found that 2011; Realo et al., 2015). In contrast, people rated public
norms vary across the 50 U.S. states, with looser norms in parks and their own bedrooms as the loosest contexts, with
big cities (Harrington & Gelfand, 2014). A study of more the least constraint on their behavior. Thus, people’s expecta-
than 11,000 people across China found the opposite pat- tions about masks may also differ across situations.
tern—cities have tighter norms (Chua et al., 2019). For We have good reasons to speculate that people’s mask use
example, large cities like Shanghai have the tightest social in tight situations was shaped more by general norm tight-
norms (Chua et al., 2019; Talhelm & English, 2020). ness than in loose situations. Gelfand proposed the idea of
This provincial variation in norm tightness provides an tight–loose “ambidexterity” to describe people’s ability to
opportunity to test for cultural differences in tightness. A shift norms (Gelfand, 2021b). This ambidexterity means that
prior study on tightness and COVID-19 tested for differences cultures can toggle between tight norms and loose norms
across nations (Gelfand et al., 2021). Although valuable, depending on the situation. For example, tight norms can
comparing different countries introduces potential confounds help cultures defend against threats in times of disease or
between countries, such as language, religion, and national natural disaster. But in times of peace, loose norms can help
government. By comparing regions within the same country, cultures achieve the benefits of flexibility, such as creativity
we can rule out many (but not all) of these important and innovation (Gelfand, 2021b).
confounds. During the pandemic, this ambidexterity makes sense in
different contexts. For example, subways are more threaten-
ing contexts, with many people crowded into a tight space. In
Norm Tightness and Mask Use
contrast, public parks are generally less crowded. Parks are
There is some evidence that social norms are related to mask also open air, which makes it less likely that people will
use. In a survey, researchers found that people who perceived spread the virus. Contact-tracing data have found that trans-
strong social norms about masks tended to report wearing mission is rare outdoors (Qian et al., 2021).
masks more often (Freidin et al., 2021). However, that study Public regulations on masks recognized differences
was limited for two reasons. First, it used only self-reported between situations. For example, health authorities in China
mask use, which may be inaccurate (Jakubowski et al., advise people to wear masks in crowded indoor situations
2021). Second, it measured mask use and social norms with but not outdoors (National Health Commission of the
the same participants, which could inflate the relationship (in People’s Republic of China, 2020b; Table S1).
other words, common-method variance). Considering these factors, people in tight cultures may
What is more, we theorized that cultures differ in norm adhere to mask wearing more in tight situations, but not in
tightness over the long term and across domains, not just loose situations. To test this, in Study 1, we measured mask
about masks. Thus, we tested whether more people would use across four situations where mask norms might differ: (a)
wear masks in regions with tight norms in general (rather than public sidewalks, (b) shopping mall entrances, (c) subway
specifically about masks). If this is true, it allows for much station entrances, and (d) public parks. We tested whether
more predictive power, since we can rely on general knowl- people would wear masks more in tight situations such as
edge about cultural tightness, rather than having to measure entrances to subway stations and shopping malls compared
norms for each new domain at each new point in time. To with in loose situations such as streets and parks. We also
measure tightness, we use measurements taken before the tested our assumption that these contexts varied in tightness
COVID-19 pandemic. Finally, we use measures of norm using a prior survey in China that asked people to rate the
tightness at the cultural level, rather than the individual level, tightness of these contexts. In Study 3, we measured people’s
which avoids the problem of common-method variance. mask use in the same four situations as Study 1. Unlike Study
1, in Study 3, we conducted our observations in a period
when there were no mask mandates in China, people can
Norms Vary Across Contexts
freely choose to wear a mask or not. Thus, we can test the
In this study, besides testing whether people in areas with prolonged constraints of situations on people’s mask-wear-
tight norms wear masks more often, we also tested whether ing behaviors during the post-pandemic era.
4 Personality and Social Psychology Bulletin 00(0)

In this study, we test what cultural factor is most associ- 5. We preregistered Study 3 when the COVID-19
ated with people’ mask use during the “drag on” phase of the restrictions were lifted in China, thus highlighting
coronavirus, while the objective risk is low and mask poli- robust nature of our predictions.
cies vary little across regions. By minimizing these factors’
confounding, we can test the “pure” and long-lasting associ-
Theoretical Contributions
ations between cultural factors and people’s behaviors of
mask use. With identifying that people in provinces with The primary theoretical contribution of this study is testing the
tight norms wear masks more, our study provides insight into idea that tightness depends on the situation, rather than being
why tight cultures contained COVID-19 better (Gelfand just a static trait. Most studies on tightness use it as a cultural
et al., 2021; Schopf, 2022). trait that is stable over time (e.g., Gelfand et al., 2011; Uz,
2015). And to be fair, cultural tightness is a trait that lasts over
time. This fact is evident in studies that have used measures of
Individual Factors cultural tightness taken from years past to explain behavior
Finally, we tested whether mask use differed based on indi- years later (e.g., Gelfand et al., 2021; Talhelm et al., 2023).
vidual factors, such as gender, age, and whether people were But we argue that the trait concept of tightness is only part
alone. We measured these factors because there is evidence of the picture. Tightness is also a combination of a trait and a
that mask use differs across demographic factors. product of situations. To make the best predictions, we must
Gender: Several studies have found that women wear know both the tightness of a culture and the tightness of the
masks more than men (such as Mahalik et al., 2021). An particular situation. This will give us more predictive power
observational study found that women were 16% more likely than knowing the trait alone.
to wear masks than men in the early days of COVID-19 in This trait-by-situation theory of tightness can help explain
Taiwan (Chuang & Liu, 2020). Another observational study some potentially puzzling recent findings. For example, a
in southwest Iran found that more women (60.2%) wore recent study found that tightness–looseness predicted
masks than men (38.7%; Rahimi et al., 2021). One potential COVID-19 cases and deaths across cultures, but it failed to
explanation is that studies have found that women are more predict differences in the first 6 months of the pandemic
likely to conform to social norms than men (meta-analysis: (Table 4, Talhelm et al., 2023). And if we extend this theo-
Bond & Smith, 1996). retical insight to other cultural traits, we can explain other
Age: A study in China found that elderly people were less differences across time. For example, a recent study found
likely to wear masks in the early days of COVID-19 (English that relational mobility differences between cultures changed
et al., 2022). This is puzzling because older people are more as the pandemic unfolded (Freeman & Schug, 2021).
likely to die from COVID-19, but one explanation could be Differences were largest at the beginning of the pandemic
that older people are more resistant to change. For example, and shrank as the pandemic worsened. This extends theory
older people are less likely than younger people to change by melding cultural traits with changes based on situations.
their shopping habits when they move to new regions
(Bronnenberg et al., 2012). Study 1
Groups: In addition, a study in Taiwan found that fewer
people in groups wore masks than people who were alone Method
(Chuang & Liu, 2020). One explanation we offer is that peo-
Sites. We observed 23,551 people in seven Chinese cities:
ple in groups are more likely to be talking, and people often
Wuhan, Beijing, Shanghai, Lanzhou, Nanchang, Chengdu,
report that masks make it harder to communicate.
and Zhuhai (Figure 1). We chose these seven cities based on
several factors:
Methodological Contributions
1. Geography: We wanted to cover a diverse geography,
This study makes several methodological contributions over
including north, south, east, and west. We also wanted
prior studies:
to include locations besides China’s large coastal cit-
ies, which tend to be overrepresented in research.
1. This study uses objective measures (observations),
2. Distance to the central government: We wanted to
cross-referenced with self-report.
get cities that varied in their distance from the central
2. This study collects a large amount of data (more than government, which could influence the enforcement
22,000 observations). of COVID-19 rules.
3. These data cover a large geographic area, with sev- 3. Availability of observers: Because our study stretched
eral cities and several sites within cities. 5 weeks over four types of location in each city, we
4. The data are longitudinal, covering 5 weeks over mul- needed dedicated observers, which limited us from
tiple locations, as well as a follow-up 3 years later. observing a larger number of cities.
Wei et al. 5

Figure 1. Mask Observation Sites in Seven Cities.

We selected two districts with different levels of eco- cars. Observers did not code staff (such as security guards
nomic development in each city. In each district, we observed and janitors) because they might be required to wear masks
people at three locations: street in residential areas, public while on duty.
park, and shopping mall entrance. We also selected at least Observers recorded whether people were wearing a
one subway station entrance in each city. Figure 2 shows pic- mask and (a) gender, (b) estimated age range (10 or below,
tures of data collection sites. The Supplemental Materials 11–20, 21–40, 41–60, above 60), and (c) whether they
explain the geographic sampling rationales in more detail. were alone or in a group. We coded people as wearing
We observed mask use for 5 weeks from May 25, 2020 to masks only when they were wearing a mask properly (cov-
June 28, 2020. Each observer was asked to observe each ering the nose and mouth). For people not wearing masks,
location at least once a week. Depending on foot traffic, we also coded whether they were eating, drinking, or
observations at each location lasted half an hour to 1 hour. smoking.
Table 1 shows the percentage of people wearing masks at To test whether the observations were reliable, two
each location and district. observers coded 76 people independently at the same place
and time. Agreement met the standard for “almost perfect or
Observation Rules. The principal researchers determined perfect agreement” (Landis & Koch, 1977): proper mask use
observation standards and trained the observers using a stan- (k = .97, p < .001), gender appearance (k = .94, p < .001),
dard protocol. On the street, observers coded people who age (k = .84, p < .001), and whether people were alone or in
were walking but not people riding a bike or in a car. We a group (k = .97, p < .001).
excluded cars and bikes because it would be difficult to accu- All data and analysis scripts are openly available on the
rately observe people moving by quickly on bikes and in Open Science Framework (https://osf.io/mwxe8/?view_only
6 Personality and Social Psychology Bulletin 00(0)

=86ffd698b341450dbc0bffde86015e30). The study was that people in more developed areas would wear masks
approved by the IRB of the first corresponding author’s more. More developed districts may have more resources
university. for enforcement, education, and easier access to masks.
Developed areas may also be much denser. We used district
Main Independent Variable GDP per capita in 2019 to represent each district’s economic
Norm Tightness. Norm tightness data come from a sur- development.
vey of 11,662 people across 31 provinces in China (Chua
et al., 2019). Unfortunately, that study only reported scores Other Cultural Factors. In addition to norm tightness, we
for provinces, not for any smaller units such as prefectures tested other cultural factors, including education, historical
or districts. The survey measured norm tightness using Gel- infectious disease, historical air pollution, collectivism, and
fand’s six-item scale (Gelfand et al., 2011). For example, par- relational mobility. To test the influence of historical disease
ticipants rate “In this province, people almost always comply rates, we collected three indexes: recent history of influenza,
with social norms” and “In this province, if people behave in historical pathogen prevalence, and recent history of SARS
an inappropriate way, others will strongly disapprove” on a cases. For collectivism, we tested relational collectivism and
scale from 1 (strongly disagree) to 6 (strongly agree). The group collectivism from the large-scale survey from Chua
scale had good reliability (alpha = .71) and significant vari- and colleagues (2019). The Supplemental Materials intro-
ability across provinces (p < .001; Chua et al., 2019). duced the rationales, measurements, and sources of all these
variables.
Control Variables To test the association of mask policies and people’s
Demographic Variables. In all analyses, we included peo- actual mask use, we used the coding of researchers at Oxford
ple’s gender, estimated age range, and whether they were (Oxford University, 2023). They coded policies on masks
alone or in a group. We also included a variable representing outside the home. The scale is from 0 (no policy) to 4
nighttime observations. We defined nighttime as after 18:00, (required outside the home at all times regardless of location
which is approximately the time of sunset during the sum- or presence of other people). We averaged the policies of
mer in China. To be conservative, we also tested observation each province over time for the time periods of our
time as a continuous variable (by hour) in the Supplemental observations.
Materials (Table S3).
Analysis. We analyzed the data using hierarchical (multi-
Average Temperature. Because we observed people dur- level) binomial models (logistical regression) with individu-
ing the summer, we tested whether hot temperatures would als nested within districts further nested in cities. We used
make people less likely to wear masks. We recorded the aver- formula of this type with the GLMER function in the pro-
age temperature of each observation day. gram R: Mask Use ~ Norm Tightness (or Other Cultural
Variables) + Control Variables + (1|City/District).
Air Pollution. One potential confound is that more people
might wear masks when air pollution is bad. To account
Results
for this, we collected the average fine particle air pollution
(PM2.5) on each observation day. We also tested average Tight Norms. More people wore masks in provinces with
historical air pollution because long-term exposure may tight norms at entrances to subways (B = 0.73, SE = 0.18,
affect regions separately from air pollution on any particu- z = 4.01, p < .001, d = 0.12) and shopping malls (B = 0.45,
lar day. SE = 0.21, z = 2.15, p = .032, d = 0.05), but not walking on
the street (B = 0.36, SE = 0.22, z = 1.63, p = .103, d = 0.04)
Population Density. COVID-19 may spread faster in areas or in parks (B = 0.32, SE = 0.33, z = 0.99, p = .321, d =
that are densely populated. This could cause more people to 0.03; Table 2). There was a significant interaction between
wear masks. We tested this by collecting population density norm tightness and tight situations (dummy coded as streets
for each district (log transformed). and parks = 0, subways and malls = 1; B = 0.14, SE = 0.03,
z = 5.02, p < .001, d = 0.07; Table 2).
COVID-19 Cases. Local COVID-19 cases might increase
people’s perception of risk and cause more people to wear Tighter and Looser Situations. People wore masks most often
masks. We collected the total local cases of each district by at entrances to subways (70%) and malls (62%). Fewer peo-
the day we started the observation. We used log COVID- ple wore masks on the street (52%) and in parks (35%; chi-
19 cases because case data was skewed (skewness = 2.58), square tests with Bonferroni correction for pairwise
exceeding a recommended cutoff of 2 for skewness (Curran comparisons). The same trend appeared in each city (Figure 3).
et al., 1996). These results showed that people’s mask-wearing behaviors
differed widely, even in the same city (and even in the same
Economic Development. There are a few reasons to expect district).
Wei et al. 7

Figure 2 Pictures of Observation Sites in Beijing, Zhuhai, Nanchang, and Lanzhou


8 Personality and Social Psychology Bulletin 00(0)

Table 1. Sample Size and Mask Use Proportions in Each Location of Each District.

Subway entrance Mall entrance Street Park

City District People % mask People % mask People % mask People % mask
Beijing Chaoyang 550 91.3 587 86.4 529 74.3 538 37.7
Haidian 92 79.3 522 79.1 655 66.6 490 59.6
Wuhan Jianghan / / 553 88.8 524 79.6 560 43.9
Wuchang 544 93.9 475 89.7 534 82.2 600 57.3
Shanghai Hongkou 375 79.5 403 60.0 566 63.1 196 40.8
Yangpu / / 378 67.2 279 60.6 349 56.4
Songjiang 103 82.5 100 62.0 101 37.6 100 18.0
Lanzhou Chengguan 516 62.8 511 35.6 496 42.1 504 26.0
Yuzhong 507 46.0 520 43.5 519 37.0 513 34.4
Nanchang Donghu 349 38.1 405 32.1 342 32.7 302 9.6
Honggutan Xinqu 387 63.6 364 47.5 253 24.1 / /
Chengdu Jinniu / / 405 63.2 405 25.7 407 8.6
Qingyang 404 65.8 405 61.0 404 47.0 406 26.6
Zhuhai Jinwan / / 500 26.6 500 21.2 499 7.0
Xiangzhou 449 74.6 500 71.6 499 40.7 549 37.5

Note. Because Zhuhai and the Yuzhong district of Lanzhou do not have a subway, we observed mask use in a bus station. Honggutan Xinqu in Nanchang is
a new district and does not have a park. (Parks are often built after all high-rise buildings have been completed.) We missed seven people’s data on mask
use and another 21 people’s data on locations, so the total sample size here is 23,523.

This finding is consistent with surveys of people’s percep- could be confounded with relaxing attitudes over time, but
tions about behavioral constraints in different contexts. Chua temperature remained significant after controlling for the
and colleagues (2019) surveyed people in China about norm day of observation (Table 2). For every 1 degree Celsius
tightness in four situations that mapped onto our observa- warmer, people’s mask use decreased 8% at subways, 4% by
tions: malls, streets, parks, and public transportation. They shopping malls, 6% on streets, and 6% in parks. These data
asked about buses, which we think should be similar to sub- suggest that people find masks more uncomfortable during
ways. People reported the tightest norms on buses (M = hot weather.
4.46, SD = 0.49), followed by shopping malls (M = 3.92, Where local COVID-19 cases were high, more people
SD = 0.57) and streets (M = 3.75, SD = 0.72). Parks had the wore masks in tight contexts (subways: B = 0.71, SE = 0.23,
loosest norms (M = 3.44, SD = 0.67).1 z = 3.02, p = .003, d = 0.09; shopping malls: B = 0.68,
SE = 0.25, z = 2.69, p = .007, d = 0.07; Table 2). Trends
Mask Use Across Demographics. Older people were less likely were similar in loose contexts, but not significant. These
to wear masks across all situations combined (p < .001). results suggest that people responded to local past risk, but
This is unexpected from the perspective of risk because older the effect of risk was greater in tight contexts.
people are at greater risk for COVID-19. However, it fits
with prior observational data in China (English et al., 2022). Robustness to Alternative Predictors. We tested the robustness
More women wore masks (60.5%) than men (46.5%). of norm tightness associated with the observed mask use. We
This gender difference was significant in all four contexts (ps also present analyses of other cultural factors without norm
< .001; Table 2). This finding fits with observations of mask tightness in the models.
use in Taiwan (Chuang & Liu, 2020). Disease prevalence did not strongly predict mask use.
In all four contexts, people who were alone were more Historical pathogen prevalence and SARS cases had no signifi-
likely to wear masks (59.5%) than people who were with cant relationship with mask use in any context (Table S6).
other people (48.2%, ps < .001; Table 2). At first glance, this Historical influenza predicted more mask use only in parks (B
result may seem counterintuitive because the virus spreads = 0.12, SE = 0.06, z = 2.13, p = .033, d = 0.06; Table S6).
between people. However, people may often be walking with After controlling for historical influenza, norm tightness
family or close friends and thinking that they are already predicted mask use at subways (B = 0.71, SE = 0.18, z =
exposed to these people. People may also take off their masks 3.87, p < .001, d = 0.12; Table S7) and shopping malls (B =
to make it easier to talk. 0.44, SE = 0.22, z = 1.99, p = .047, d = 0.05; Table S7).
Norm tightness predicted mask use at subways after control-
Environmental Factors. On hot days, fewer people wore masks ling for historical pathogen prevalence (B = 0.50, SE = 0.22,
(ps < .05 across all four contexts, Table 2). Temperature z = 2.31, p = .021, d = 0.07; Table S7) and SARS cases (B =
Wei et al. 9

Table 2. Norm Tightness Predicts Mask Use in Subways and Malls But Not in Streets or Parks

Subway entrance Mall entrance Street Park All All

Variable Tight Tight Loose Loose / /


Female 0.460*** 0.626*** 0.887*** 0.815*** 0.704*** 0.704***
(0.076) (0.059) (0.058) (0.061) (0.030) (0.030)
Age −0.145** 0.056 0.101*** −0.050 −0.055*** −0.056***
(0.050) (0.033) (0.030) (0.026) (0.015) (0.015)
Alone 0.407*** 0.318*** 0.426*** 0.273*** 0.486*** 0.488***
(0.080) (0.061) (0.059) (0.066) (0.031) (0.031)
Daytime −0.241 −0.036 0.191 0.663*** 0.239*** 0.271***
(0.144) (0.139) (0.129) (0.156) (0.060) (0.060)
Air pollution −0.004 −0.005* 0.004 0.004 −0.004* −0.002
(0.004) (0.003) (0.003) (0.003) (0.001) (0.002)
Temperature −0.083*** −0.038* 0.061*** −0.063*** −0.049*** −0.069***
(Celsius) (0.017) (0.016) (0.014) (0.017) (0.007) (0.009)
Population density 0.331 0.636* 0.779*** 0.642 0.630* 0.650*
(0.270) (0.274) (0.212) (0.370) (0.297) (0.278)
COVID-19 cases 0.710** 0.675** 0.515 0.339 0.378 0.424
(0.235) (0.251) (0.270) (0.338) (0.251) (0.234)
GDP per capita 0.004 0.003 −0.008 −0.020 0.008 0.009
(0.027) (0.034) (0.026) (0.055) (0.033) (0.032)
Norm tightness 0.727*** 0.453* 0.357 0.324 0.235 0.279
(0.182) (0.211) (0.219) (0.326) (0.212) (0.196)
Observation day 0.009***
(0.002)
Tight situations 0.526*** 0.525***
(0.102) (0.102)
Norm Tightness × 0.138*** 0.138***
Tight Situations (0.027) (0.027)
Observations 4,162 6,463 6,423 5,905 22,953 22,953
Districts 11 15 15 14 15 15
Cities 7 7 7 7 7 7

Note. Models are multilevel with individuals nested in districts in cities. Values are unstandardized regression coefficients, with standard errors in
parentheses. Daytime is defined as before 18:00. Because college graduates per 100 million and GDP per capita are highly correlated, we did not add
GDP per capita in the models. Tight situations are subways and shopping malls. COVID-19 = coronavirus disease; GDP = gross domestic product.
Results of norm tightness and tight situations are displayed in bold.
*p < .05. **p < .01. ***p < .001.

0.71, SE = 0.21, z = 3.33, p < .001, d = 0.10; Table S7), Sections 5 and 6. Except relational collectivism, none of the
while the results were weaker for shopping malls (Table S7). other variables had significant correlations with mask use.
Historical air pollution was positively correlated with mask
use. However, it was only significant for mask use at shopping Discussion
malls (B = 0.05, SE = 0.02, z = 2.03, p = .043, d = 0.05;
Table S7). When controlling for historical air pollution, norm Using objective observations of people’s actual mask use,
tightness still predicted mask use at subways (B = 0.72, SE = Study 1 found differences across situations and regions.
0.16, z = 4.51, p < .001, d = 0.14) and malls (B = 0.53, SE More people wore masks in provinces with tight norms.
= 0.16, z = 3.30, p < .001, d = 0.08; Table S7). Tightness also seemed to be a property of contexts. More
We also tested relational collectivism, group collectivism, people wore masks in tight contexts (malls and subways)
relational mobility, and mask policies (Table S4). We only than loose contexts (parks and streets).
found that relational collectivism was correlated with mask One limitation of Study 1 is that in-person coding is time-
use at subways and shopping malls. However, they were both consuming. This limits the number of cities we could mea-
in the opposite direction of what theory would predict (Lu sure, and it limits the statistical power to test for regional
et al., 2021). We discuss these results in detail in Supplemental differences. To address this shortcoming, we used large-scale
10 Personality and Social Psychology Bulletin 00(0)

Figure 3. Mask Use Varied Widely Across Tight (Orange) and Loose (Blue) Situations.
Note. Bars are ±1 SEM (Standard Error of the Mean). There is no subway in Zhuhai or the Yuzhong district of Lanzhou, so we observed a bus station.

survey data covering 284 prefectures in 30 provinces over a Study 2


longer window of time. While Study 1 measures mask use at
the situational level and focuses on the interactions of norm Method
tightness in general and situational restrictions, Study 2 mea- The international market research company YouGov sur-
sures mask use at the societal level. With much more broad veyed 15,985 people from April 2, 2020 to September 25,
sample sites, it provides more solid evidence on the associa- 2020, which corresponds to what we call the “drag on”
tion of general norm tightness and mask use. phase. The survey covered 284 prefectures in 30
Wei et al. 11

provinces (not including Tibet, Hong Kong, Macau, or Other Cultural Factors. We tested the same alternative cul-
Taiwan). Prefectures are similar to U.S. counties, and tural factors as in Study 1.
they are the smallest geographic unit the survey
identified. Analysis. To analyze the data, we used hierarchical linear
The survey included a question about masks: “How often models with people nested in prefectures, further nested in
have you worn a face mask outside your home (e.g., when on provinces. This takes into account the fact that individual
public transport, going to a supermarket, going to a main respondents are not truly independent data points. We used
road)?” (1 = Not at all, 2 = Rarely, 3 = Sometimes, 4 = the LMER function in the program R to run models in this
Frequently, 5 = Always). format: Mask Use ~ Norm Tightness (or Other Cultural Vari-
ables) + Control Variables + (1|Province/Prefecture).
Independent Variable
Norm Tightness. We used the same provincial norm tight-
Results
ness data as in Study 1.
Consistent with Study 1, more people in provinces with tight
Control Variables norms reported wearing masks (B = 0.09, SE = 0.03, t =
Demographic Variables. The YouGov survey asked par- 2.77, p = .011, d = 1.13; Table 3, Models 1; Figure 4).
ticipants’ gender, age, and socioeconomic status. Participants Tightness continued to predict mask use after controlling for
reported their social status on a scale from 0 (the worst possible prefecture population density and COVID-19 cases (B =
life) to 10 (the best possible life). A total of 3,002 people did 0.07, SE = 0.03, t = 2.18, p = .039, d = 0.88; Table 3,
not report status. Because there were fewer than 30 responses Model 2). In Model 8, we found similar results after using
in Ningxia and Qinghai provinces, we excluded them (norm norm tightness values that reduced collinearity with GDP (B
tightness still predicted mask use when including all provinces; = 0.12, SE = 0.04, t = 3.54, p = .002, d = 1.59).
Table S11). This left 12,949 responses in the final analysis.
Demographic Differences. Demographic differences were
Time. Because mask use could change over time, we also similar to the observational findings. Women reported wear-
controlled for the week participants took the survey. ing masks more than men (B = 0.16, SE = 0.02, t = 9.92, p
< .001, d = 0.17; Table 3, Model 1). Older people also
Population Density. We used log prefecture-level popula- reported using masks less (B = −0.006, SE = 0.001, t =
tion density in 2015 from the China City Statistical Yearbook −7.92, p < .001, d = −0.14; Table 3, Model 1). People with
2016. higher socioeconomic status wore masks more (B = 0.023,
SE = 0.004, t = 5.99, p < .001, d = 0.11; Table 3, Model 1).
COVID-19 Cases. We collected data on the total COVID- People’s mask use decreased over time (B = −0.029, SE =
19 cases (including local cases and imported cases from 0.002, t = −13.99, p < .001, d = −0.25; Table 3, Model 1).
inbound travelers) from April 2020 to September 2020 of Overall, the self-report mask data gave a similar picture to
each prefecture. Because 219 prefectures reported 0 cases, the observational data. However, we could not compare the
the data was highly skewed (9.28) and kurtotic (108.79). exact percentage of mask use because the self-report survey
This is a problem for statistical tests that assume a normal used imprecise estimates, such as “sometimes.”
distribution. To correct for this, we first computed each pre-
fecture’s COVID-19 cases per capita, then added a small Robustness to Alternative Predictors. People reported more
number (10−7), and log transformed it. This transforma- mask use in prefectures with more COVID-19 cases (B =
tion lowered the skewness (1.44) and kurtosis (0.16). This 0.008, SE = 0.003, t = 3.35, p = .001, d = 0.54; Table 3,
reduced skewness below the suggested cutoff of 2 (Curran Model 2). People also reported wearing masks more in pre-
et al., 1996). fectures with more college graduates (B = 0.420, SE =
0.114, t = 3.67, p < .001, d = 0.59; Table 3, Model 3). After
Economic Development. Chua and colleagues (2019) controlling for education, tightness continued to predict
found that more developed Chinese provinces have tighter mask use (B = 0.07, SE = 0.03, t = 2.27, p = .032, d =
norms. Thus, norm tightness is collinear with provincial 0.90; Table 3, Model 3). People in provinces with higher his-
GDP per capita. To test whether the results were robust to torical air pollution reported wearing masks more (B =
collinearity, we ran Models 1 to 7 with the original tightness 0.005, SE = 0.002, t = 2.61, p = .017, d = 1.19; Table 3,
data and Model 8 with tightness adjusted to remove collin- Model 7). Norm tightness continued to predict self-reported
earity with GDP. To reduce collinearity, we ran a regression mask use after controlling for historical air pollution (B =
using GDP to predict norm tightness and saved the unstan- 0.07, SE = 0.03, t = 2.45, p = .026, d = 1.20; Table 3,
dardized residual. We used the residual to represent tightness Model 7).
after controlling for GDP. Provincial GDP per capita is from Mask use was not significantly related to historical influ-
2019, from the China Statistical Yearbook 2020. enza cases (p = .334, Table S8, Model 1), historical
12 Personality and Social Psychology Bulletin 00(0)

Table 3. People in Provinces With Tighter Norms Reported Wearing Masks More Often.

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8


Female 0.157*** 0.156*** 0.155*** 0.156*** 0.156*** 0.156*** 0.155*** 0.156***
(0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) (0.016)
Age −0.006*** −0.006*** −0.006*** −0.006*** −0.006*** −0.006*** −0.006*** −0.006***
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Socioeconomic 0.023*** 0.023*** 0.022*** 0.023*** 0.023*** 0.022*** 0.022*** 0.023***
status (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004)
Week −0.029*** −0.029*** −0.029*** −0.029*** −0.029*** −0.029*** −0.029*** −0.029***
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Pref. 0.036 0.016 0.045 0.032 0.038 −0.015 0.036
population (0.064) (0.063) (0.064) (0.064) (0.065) (0.062) (0.059)
density
COVID-19 0.008** 0.003 0.008** 0.009*** 0.008** 0.010*** 0.009***
cases
(0.003) (0.003) (0.003) (0.003) (0.003) (0.002) (0.002)
Coll. graduates 0.420***
per 100 million (0.114)
Historical 0.007
influenza (0.006)
Historical −0.009
pathogen (0.010)
prevalence
SARS cases 0.003
(0.012)
Historical air 0.005*
pollution (0.002)
Norm 0.122**
tightness GDP- (0.035)
adjusted
Norm 0.090* 0.070* 0.072* 0.075* 0.068* 0.067 0.067*
tightness
(0.033) (0.032) (0.032) (0.032) (0.032) (0.036) (0.027)
Survey 12,949 12,910 12,910 12,910 12,910 12,910 12,910 12,910
respondents
Prefectures 279 274 274 274 274 274 274 274
Provinces 28 28 28 28 28 28 28 28

Note. Models are hierarchical linear models with individuals nested in prefectures in provinces. Values are unstandardized regression coefficients, with
standard errors in parentheses. Week represents the week the participant took the survey (from Week 1 to 16). COVID-19 = coronavirus disease;
GDP = gross domestic product; SARS = severe acute respiratory syndromes. Results of norm tightness are displayed in bold.
*
p < .05. **p < .01. ***p < .001.

pathogen prevalence (p = .334, Table S8, Model 2), or SARS specific to tightness and not factors that are often correlated with
cases (p = .297, Table S8, Model 3). Norm tightness tightness, such as collectivism and relational mobility.
remained predictive after controlling for historical influenza
(B = 0.07, SE = 0.03, t = 2.30, p = .031, d = 0.94; Table 3, Discussion
Model 4), historical pathogen prevalence (B = 0.07, SE =
0.03, t = 2.09, p = .048, d = 0.87; Table 3, Model 5), and Overall, the self-report survey data mirrored the observa-
SARS cases (marginally significant; B = 0.07, SE = 0.04, t tional data. Among individual factors, women, younger peo-
= 1.84, p = .080, d = 0.80; Table 3, Model 6). ple, and people with higher social status reported wearing
Self-report mask use was not significantly related to rela- masks more. One contribution is that Study 2 had a larger
tional collectivism, group collectivism, relational mobility, or sample, covering many more provinces and prefectures. This
mask policies (Table S5 and S12). These results suggest that the larger geographic sample replicated the finding that masks
association between norm tightness and self-report mask use is were more common in provinces with tight norms.
Wei et al. 13

Study 3
Method
We preregistered Study 3 on the OSF at the start of data col-
lection (https://osf.io/mk3xc/?view_only=b99800f736da4c6
9b7d21e43fa90bc8a). We preregistered two hypotheses:
More people will wear masks in tight areas than in loose
areas, and more women will wear masks than men. We also
preregistered the observation cities and locations and the
analysis methods we were going to use.
We observed mask use in five Chinese cities, Beijing,
Hangzhou, Shanghai, Wuhan, Hong Kong, and one
Portuguese city: Lisbon. We chose one city outside China
because we wanted to gain a sense of China’s mask use com-
pared with a culture that did not experience the strict zero-
Figure 4 People Report Wearing Masks More Often in COVID policy. China and Portugal have similar norm
Provinces with Tighter Norms tightness scores (China: 7.9; Portugal: 7.8) in Gelfand and
Note. The correlation remained significant after excluding the outlier colleagues’ (2011) study. Portugal lifted its mask mandate
(Guizhou) with the least tightness, r = .42, p = .028.
and recommendations starting on October 1, 2022 (Oxford
University, 2023). This provided us an opportunity to test if
Norm tightness predicted beyond other logical factors. situational differences in mask use would still last half a year
For example, population density and local COVID-19 cases after the lifting of mask mandates.
are logical risk factors for COVID-19. Yet, norm tightness Like in Study 1, in each city, we observed people at four
continued to explain variation in mask use after accounting locations: subway station entrances, shopping mall entrances,
for these risk factors. streets in residential areas, and public parks. The observa-
One interesting finding was that collectivism did not pre- tional rules were the same as in Study 1, except that each
dict higher mask use in Study 1 or Study 2. This is inconsis- observer planned to record around 50 observations at each
tent with an earlier study (Lu et al., 2021). One possibility is site.
that the differences in collectivism across cultures in the ear- We observed 1,287 people from March 9, 2023 to March
lier study are different from the differences in collectivism 19, 2023 (Table S14). We used chi-square tests to examine
here within China. However, we hesitate to draw strong con- people’s mask use differences between the four locations. All
clusions comparing the two studies because the two studies p values are Bonferroni corrected for pairwise comparisons.
use different measures of collectivism. We also used hierarchical linear models with people nested
Regions’ history of infectious disease is also a logical in cities to test the preregistered hypotheses.
explanation for mask use. For example, SARS hit provinces
like Beijing hard in 2003, whereas other provinces experi-
enced zero cases (Talhelm et al., 2023). Yet, norm tightness
Results
was a stronger predictor than regions’ history of SARS and Putting all data from all six cities together, we found similar
the flu. results as Study 1: more people wore masks at entrances to
In Study 3, we built upon the earlier studies by testing subways (B = 1.58, SE = 0.24, z = 6.49, p < .001, d = 0.36;
whether differences remained after China ended its COVID-19 Table 4, Model 1) and entrances to malls (B = 0.91, SE =
restrictions. We preregistered this study and measured people’s 0.22, z = 4.07, p < .001, d = 0.23; Table 4, Model 1). Across
mask use in March, 2023. This came 4 months after China all five Chinese cities we observed, more people wore masks
abruptly lifted its COVID restrictions in December, 2022. at subways than in parks (Figure 5, Table S15).
The timing of Study 3 is important because it gives Differences were large. For example, in Hangzhou, mask
insight into the question of formal mask policies and use was more than double at the subway (80%) compared
enforcement. Did people wear masks in subways and malls with streets (34%) and parks (22%). This suggests that mask
simply because of formal policies, or is there a larger sense use continued to differ between tight and loose contexts,
of situational tightness in these types of places that extend even months after China lifted COVID rules.
beyond formal rules?
Study 3 can rule out the effect of having formal policies in Mask Use Lower in Portugal. Mask use was far lower in Portu-
place. Four months after China ended its zero-COVID policy, gal. In Lisbon, we observed only four out of 215 people
people had free choice to wear masks or not. And they had wearing masks. Although more people wore masks in tight
had 4 months to adjust to life without COVID restrictions. contexts (2.8%) than loose contexts (0.9%) in Portugal, the
14 Personality and Social Psychology Bulletin 00(0)

Table 4. Even Months After China Lifted COVID Restrictions, Chinese cities. Supplemental Section 13 reports the results
More People Wore Masks in Tight Context (March 2023). in more detail.
Variable All cities China
Two-Phase Comparisons of Beijing, Shanghai, and Wuhan. In
Female 0.892*** 0.915*** another exploratory analysis (not preregistered), we com-
(0.132) (0.133) pared changes in mask use between 2020 and 2023 in Bei-
Age 0.080 0.056 jing, Shanghai, and Wuhan. We observed these three cities at
(0.076) (0.077) both time points. Beijing’s mask use dropped from 71% in
Time −0.064 −0.040 2020 to 64% in 2023, t(229.04) = 2.15, p = .032, d = 0.28.
(0.064) (0.060)
Shanghai’s mask use dropped from 61% to 38%, t(3,169) =
Air pollution −0.016 −0.010
6.55, p < .001, d = 0.23, and Wuhan’s mask use dropped
(0.008) (0.007)
from 76% to 45%, t(253.52) = 9.35, p < .001, d = 1.17. It
Temperature (Celsius) 0.025 0.017
is interesting to note that mask use dropped the most (31%)
(0.023) (0.022)
Subway 1.581*** 1.544***
in the city with the loosest norms.
(0.244) (0.235)
Shopping mall 0.912*** 0.834*** Discussion
(0.224) (0.222)
Street 0.255 0.233 The data show that people’s mask use was still attuned to
(0.219) (0.217) tight versus loose contexts, even months after China lifted its
Observations 1,287 1,072 zero-COVID policy. This finding rules out the idea that the
Cities 6 5 higher mask use in tight contexts was a product of tight poli-
cies and enforcement alone. By March, 2023, China had
Note. Models are multilevel with individuals nested in cities. Values are lifted its mask policies and enforcement. Yet people contin-
unstandardized regression coefficients, with standard errors in parentheses.
Model 1 includes all six cities that we preregistered. Model 2 limits the ued wearing masks more often in tight contexts. These data
analysis to China (excluding Lisbon). Subway, shopping mall, and street are illustrate the usefulness combining both long-term tightness
coded as 1 = yes, 0 = no. Parks are the baseline for comparisons. Since and the tightness of situations.
the observer in Hong Kong did not record whether people were alone
or in a group on the street and at the entrance to subway or entrance to
shopping mall, we did not include this variable in the main models. Results General Discussion
were similar in analyses controlling for this variable (Supplemental Section
12, Table S16). COVID-19 = coronavirus disease. Results of subway and The results of these studies illuminate consistent high
shopping mall are displayed in bold.
*p < .05. **p < .01. ***p < .001.
mask use, even during a time of low COVID cases. Most
subways and malls were still above 50% mask use, even
after 4 weeks of COVID case counts that were very low by
differences were tiny and nowhere near significant. This global standards. In March 2023, the subways in all five
result indicates a huge difference in mask use between China Chinese cities were still above 50% mask use despite no
and abroad. requirements or enforcement. Overall, these data fit with
earlier findings that China is a culture with tight norms
More Women Wear Masks Than Men. Consistent with the (Gelfand et al., 2011).
findings of Studies 1 and 2, more women wore masks than Yet the data showed this is not uniform across China. For
men (B = 0.89, SE = 0.13, z = 6.73, p < .001, d = 0.38; example, mask use averaged 71% in Beijing versus 41% in
Table 4, Model 1). This is also consistent with the finding Lanzhou from late May 2020 to June 2020. These results fit
that women’s risk perception is higher than men (Brown with earlier research finding regional differences in norms
et al., 2021). across China (Chua et al., 2019). This adds to the idea that
China is not a single culture. There are meaningful cultural dif-
Tight Norms and Tight Situations. Although we did not pre- ferences within China (Chua et al., 2019; Talhelm et al., 2014).
register this analysis, we did an exploratory test of the The results also fit with the idea that norm tightness is a
interaction between regional norm tightness and situa- stable cultural trait. The tightness data were measured years
tional tightness. We found that more people wore masks in before COVID-19, yet it still predicted mask use in 2023. This
tighter provinces (B = 0.62, SE = 0.21, z = 2.98, p = suggests that cultural differences in tightness are enduring,
.003, d = 0.20, across all situations) and tighter situations even in the face of a disease that did not exist before 2019.
(B = 2.54, SE = 1.06, z = 2.39, p = .017, d = 0.16), but
the interaction was not significant (B = −0.31, SE = 0.25,
z = −1.23, p = .219, d = −0.08; Table S17). However,
Tight Regions and Tight Contexts
these results should be treated with caution because we The main theoretical contribution of this study is that it
only have comparable regional tightness data for four demonstrates how norm tightness is both a stable cultural
Wei et al. 15

Figure 5 Mask Use Varied Widely Across Tight (Orange) and Loose (Blue) Situations (March, 2023)

trait and a product of situations. Although tightness– predict mask use in parks, for example. But it did predict
looseness theory recognized differences between situa- mask use in the subway.
tions since the early days (Gelfand et al., 2011), most This helps flesh out Gelfand’s idea of tight–loose ambi-
studies (including our own) use tightness as a static dif- dexterity—the ability to switch between tight and loose
ference between groups. This is true of studies that use (Gelfand, 2021a, 2021b). The situation is one determinant of
tightness to predict cultural differences in crime and drug people switching between tight and loose. Our data suggest
use (Gelfand, 2018), creativity (Chua et al., 2019), that people in China were switching between tight norms on
COVID-19 spread (English et al., 2022; Talhelm et al., the subway to loose norms on the sidewalk.
2023), and conformity in how people respond to surveys The 2023 data found that these tight–loose differences per-
(Uz, 2015). sisted even in the absence of formal rules. In December, 2022,
In contrast, this study documents how the same behavior China lifted its COVID restrictions, and people had free choice
(wearing a mask in public) depends on cultural tightness to wear masks or not. Despite this, months later, people were
only in the right situation. Tightness did not significantly still masking more in tight situations than loose situations.
16 Personality and Social Psychology Bulletin 00(0)

Limitations and Future Directions Practical Implications


One limitation of this study is that data cannot prove causal- The results of this study might be useful for planning public
ity because we did not manipulate norm tightness. Future health responses. The data may give clues about where it
studies can use experiments or interventions to pin down would be wise to spend public health and enforcement dol-
causal effects of norm tightness on mask use (Dimant et al., lars. For example, if we know what contexts have tight
2022; Gelfand et al., 2022). social norms to begin with, we can reasonably predict that
Another limitation is that we measured compliance with compliance will be higher there. Governments and schools
only one social norm—wearing masks. COVID-19 preven- might be better off if they put effort into raising awareness
tion included other norms, such as staying home while sick, campaigns, monitoring, and enforcement in weaker
testing before seeing people, and washing hands. One distin- contexts.
guishing feature about masks is that they are easy to monitor. The results of Study 3 (done after the restrictions were
It is easy to see whether people are wearing masks, but it is lifted) also suggest practical implications for public health.
harder to see whether a parent is failing to say that their kids The results suggest that cultural differences persist even after
were exposed to COVID-19 before taking them to school. governments lift formal rules. This can be useful in helping
There is evidence that people take advantage of the fact governments forecast disease outcomes after changing pol-
that some COVID behaviors are not easy to verify. In a sur- icy. For example, when China lifted its COVID restrictions,
vey, one fourth of U.S. parents reported breaking the rules it experienced a massive wave of illness and death.
for their children, such as by not telling other people that Policymakers could have used tightness data as a part of
their child was exposed to COVID-19 (Levy et al., 2023). forecasts for where to stockpile antiviral treatments in antici-
Future studies can test whether norm tightness applies as pation of the coming wave of infections.
well to these hard-to-monitor behaviors. Finally, cultural factors like norm tightness could partly
The third limitation is that there may be other potential explain why Johns Hopkins’s pandemic preparedness rank-
mechanisms for people to keep wearing masks in subways ings were so far from actual COVID-19 outcomes just a year
and shopping malls after the lift of mask mandates. For later (Nuclear Threat Initiative, 2019). For example, Johns
example, the “inertia” of habitual behaviors may be driving Hopkins ranked the United States and the Netherlands at the
people to wear masks more in subways and malls. Or, even top of the world in pandemic preparedness. But these two
though the mask mandates were lifted, they may still have a countries had some of the highest COVID-19 cases per cap-
lasting effect on people’ mask use. Future studies can test ita (Talhelm et al., 2023).
these potential reasons. It is probably not a coincidence that both the United States
The fourth limitation is generalizability. We do not know and the Netherlands have loose norms (Gelfand et al., 2011).
whether this finding is generalizable beyond China. To test While Johns Hopkins took into account many hard assets
this, future studies can collect data on mask use or other such as hospital beds and research facilities, it did not account
health-protective behaviors in other countries. for “soft” assets such as social norms. This study suggests
one concrete metric that researchers can start using to take
Strategic Loosening Might Combat Pandemic culture into account before the next pandemic.
Fatigue Declaration of Conflicting Interests
Although tight norms meant more masks in these data, norm The author(s) declared no potential conflicts of interest with
tightness is not simply good or bad. Instead, there are trade- respect to the research, authorship, and/or publication of this
offs between order and openness. Tight cultures tend to have article.
more order—better public health outcomes, less crime, and
less drug use (Gelfand et al., 2020). In contrast, loose cul- Funding
tures tend to be better at innovating and protecting individual The author(s) disclosed receipt of the following financial support
rights. for the research, authorship, and/or publication of this article: This
Gelfand (2021a) proposed that cultures would benefit project was partially supported with funding from the Susan and
from “tight-loose ambidexterity” by tightening norms during Richard Kiphart Center for Global Health and Social Development
pandemics. Our data suggest another layer—that cultures at the University of Chicago’s Crown Family School of Social
can also have ambidexterity across different contexts at the Work, Policy, and Practice.
same time.
This ambidexterity might help societies fight pandemic ORCID iDs
fatigue. Wearing masks can be a burden. But by allowing Liuqing Wei https://orcid.org/0000-0001-6488-7454
people the freedom to go without masks in low-risk contexts Alexander Scott English https://orcid.org/0000-0002-0142-
like parks, people may have more patience with wearing 146X
masks in high-risk contexts like subway cars. Thomas Talhelm https://orcid.org/0000-0002-0954-5758
Wei et al. 17

Supplemental Material English, A. S., Talhelm, T., Tong, R., Li, X., Su, Y., & Lu, J.
(2022). Historical rice farming explains faster mask use during
Supplemental material is available online with this article.
early days of China’s COVID-19 outbreak. Current Research
in Ecological and Social Psychology, 3, 100034. https://doi.
Note
org/10.1016/j.cresp.2022.100034
1. Chua and colleagues (2019) did not report these results in their Freeman, J. D., & Schug, J. (2021). Freedom to stay-at-home?
paper, so we computed the scores in four situations using their Countries higher in relational mobility showed decreased
raw data. Repeated-measures ANOVA with Bonferroni correc- geographic mobility at the onset of the COVID-19 pan-
tion showed that the pairwise comparisons of behavioral con- demic. Frontiers in Psychology, 12, Article 3999. https://doi.
straints between four situations were all statistically significant. org/10.3389/fpsyg.2021.648042
Freidin, E., Acera Martini, L., Senci, C. M., Duarte, C., & Carballo,
References F. (2021). Field observations and survey evidence to assess
Badillo-Goicoechea, E., Chang, T. H., Kim, E., LaRocca, S., predictors of mask wearing across different outdoor activities
Morris, K., Deng, X., Chiu, S., Bradford, A., Garcia, A., Kern, in an Argentine city during the COVID-19 pandemic. Applied
C., Cobb, C., Kreuter, F., & Stuart, E. A. (2021). Global trends Psychology: Health and Well-Being, 14(1), 81–100. https://
and predictors of face mask usage during the COVID-19 pan- doi.org/10.1111/aphw.12292
demic. BMC Public Health, 21(1), Article 2099. https://doi. Gelfand, M. J. (2018). Rule makers, rule breakers: How tight and
org/10.1186/s12889-021-12175-9 loose cultures wire our world. Scribner.
Bond, R., & Smith, P. B. (1996). Culture and conformity: A meta- Gelfand, M. J. (2021a). Cultural evolutionary mismatches in response
analysis of studies using Asch’s (1952b, 1956) line judgment to collective threat. Current Directions in Psychological Science,
task. Psychological Bulletin, 119(1), 111–137. https://doi. 30(5), 401–409. https://doi.org/10.1177/09637214211025032
org/10.1037/0033-2909.119.1.111 Gelfand, M. J. (2021b, August 23). A failure of fear: Why certain
Bronnenberg, B. J., Dubé, J. P. H., & Gentzkow, M. (2012). The nations flunked the COVID-19 threat test. Behavioral Scientist.
evolution of brand preferences: Evidence from consumer https://behavioralscientist.org/a-failure-of-fear-why-certain-
migration. The American Economic Review, 102(6), 2472– nations-flunked-the-covid-19-threat-test-tight-loose-cultures/
2508. https://doi.org/10.1257/aer.102.6.2472 Gelfand, M. J., Caluori, N., Jackson, J. C., & Taylor, M. K. (2020).
Brown, G. D., Largey, A., & McMullan, C. (2021). The impact of The cultural evolutionary trade-off of ritualistic synchrony.
gender on risk perception: Implications for EU member states’ Philosophical Transactions of the Royal Society B: Biological
national risk assessment processes. International Journal of Sciences, 375(1805), 20190432. https://doi.org/10.1098/
Disaster Risk Reduction, 63, 102452. https://doi.org/10.1016/j. rstb.2019.0432
ijdrr.2021.102452 Gelfand, M. J., Jackson, J. C., Pan, X., Nau, D., Pieper, D., Denison,
Bruine de Bruin, W., & Bennett, D. (2020). Relationships between ini- E., Dagher, M., Van Lange, P. A. M., Chiu, C. Y., & Wang, M.
tial COVID-19 risk perceptions and protective health behaviors: (2021). The relationship between cultural tightness–looseness
A national survey. American Journal of Preventive Medicine, and COVID-19 cases and deaths: A global analysis. The Lancet
59(2), 157–167. https://doi.org/10.1016/j.amepre.2020.05.001 Planetary Health, 5(3), e135–e144. https://doi.org/10.1016/
Chinese Government. (2023, January 14). https://www.gov.cn/xin- S2542-5196(20)30301-6
wen/gwylflkjz231/index.htm Gelfand, M. J., Li, R., Stamkou, E., Pieper, D., Denison, E.,
Chua, R. Y. J., Huang, K. G., & Jin, M. (2019). Mapping cultural Fernandez, J., Choi, V., Cha, t, man, J., Jackson, J., & Dimant,
tightness and its links to innovation, urbanization, and happi- E. (2022). Persuading republicans and democrats to comply
ness across 31 provinces in China. Proceedings of the National with mask wearing: An intervention tournament. Journal of
Academy of Sciences of the United States of America, 116(14), Experimental Social Psychology, 101, 104299. https://doi.org/
6720–6725. https://doi.org/10.1073/pnas.1815723116 https://doi.org/10.1016/j.jesp.2022.104299
Chuang, Y., & Liu, J. C. E. (2020). Who wears a mask? Gender Gelfand, M. J., Raver, J. L., Nishii, L., Leslie, L. M., Lun, J., Lim,
differences in risk behaviors in the COVID-19 early days in B. C., Duan, L., Almaliach, A., Ang, S., Arnadottir, J., Aycan,
Taiwan. Economics Bulletin, 40(4), 2619–2627. Z., Boehnke, K., Boski, P., Cabecinhas, R., Chan, D., Chhokar,
Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of J., D’Amato, A., Ferrer, M., Fischlmayr, I. C., & Yamaguchi,
test statistics to nonnormality and specification error in confir- S. (2011). Differences between tight and loose cultures: A
matory factor analysis. Psychological Methods, 1(1), 16–29. 33-nation study. Science, 332(6033), 1100–1104. https://doi.
https://doi.org/10.1037/1082-989X.1.1.16 org/10.1126/science.1197754
Dimant, E., Clemente, E. G., Pieper, D., Dreber, A., Gelfand, Haischer, M. H., Beilfuss, R., Hart, M. R., Opielinski, L., Wrucke,
M., Hallsworth, M., Holzwarth, A., & Tantia, P. (2022). D., Zirgaitis, G., Uhrich, T. D., & Hunter, S. K. (2020). Who
Politicizing mask-wearing: Predicting the success of behav- is wearing a mask? Gender-, age-, and location-related dif-
ioral interventions among republicans and democrats in the ferences during the COVID-19 pandemic. PLOS ONE,
U.S. Scientific Reports, 12(1), 7575. https://doi.org/10.1038/ 15(10), Article e0240785. https://doi.org/10.1371/journal.
s41598-022-10524-1 pone.0240785
English, A. S., & Li, X. (2021). Mask use depends on the individual, Harrington, J. R., & Gelfand, M. J. (2014). Tightness-looseness
situation, and location—Even without COVID-19 transmission: across the 50 United States. Proceedings of the National
An observational study in Shanghai. Frontiers in Psychology, Academy of Sciences of the United States of America, 111(22),
12, Article 754102. https://doi.org/10.3389/fpsyg.2021.754102 7990–7995. https://doi.org/10.1073/pnas.1317937111
18 Personality and Social Psychology Bulletin 00(0)

Jadil, Y., & Ouzir, M. (2021). Exploring the predictors of health- Qian, H., Miao, T., Liu, L., Zheng, X., Luo, D., & Li, Y. (2021).
protective behavior during the COVID-19 pandemic: A multi- Indoor transmission of SARS-CoV-2. Indoor Air, 31(3), 639–
country comparison. Environmental Research, 199, 111376. 645. https://doi.org/10.1111/ina.12766
https://doi.org/10.1016/j.envres.2021.111376 Rahimi, Z., Shirali, G. A., Araban, M., Mohammadi, M. J., &
Jakubowski, A., Egger, D., Nekesa, C., Lowe, L., Walker, M., & Cheraghian, B. (2021). Mask use among pedestrians during
Miguel, E. (2021). Self-reported mask wearing greatly exceeds the COVID-19 pandemic in Southwest Iran: An observational
directly observed use: Urgent need for policy intervention in study on 10,440 people. BMC Public Health, 21, Article 133.
Kenya. medRxiv. https://doi.org/10.1186/s12889-020-10152-2
Kitayama, S. (2002). Culture and basic psychological processes— Realo, A., Linnamägi, K., & Gelfand, M. J. (2015). The cultural
Toward a system view of culture: Comment on Oyserman et al. dimension of tightness-looseness: An analysis of situational
(2002). Psychological Bulletin, 128(1), 89–96. https://doi. constraint in Estonia and Greece. International Journal
org/10.1037//0033-2909.128.1.89 of Psychology, 50(3), 193–204. https://doi.org/10.1002/
Landis, J. R., & Koch, G. G. (1977). The measurement of observer ijop.12097
agreement for categorical data. Biometrics, 33(1), 159–174. Rieger, M. (2020). To wear or not to wear? Factors influencing
https://doi.org/10.2307/2529310 wearing face masks in Germany during the COVID-19 pan-
Levy, A. G., Thorpe, A., Scherer, L. D., Scherer, A. M., Butler, demic. Social Health and Behavior, 3(2), 50–54. https://doi.
J. M., Shoemaker, H., & Fagerlin, A. (2023). Parental non- org/10.4103/shb.shb_23_20
adherence to health policy recommendations for prevention Rubaltelli, E., Tedaldi, E., Orabona, N., & Scrimin, S. (2020).
of COVID-19 transmission among children. JAMA Network Environmental and psychological variables influencing reac-
Open, 6(3), e231587. https://doi.org/10.1001/jamanet- tions to the COVID-19 outbreak. British Journal of Health
workopen.2023.1587 Psychology, 25(4), 1020–1038. https://doi.org/10.1111/
Lopes, M. (2022, April 1). Even after Singapore lifts restrictions, bjhp.12473
residents stick to their masks. The Washington Post. https:// Savadori, L., & Lauriola, M. (2022). Risk perceptions and COVID-
www.washingtonpost.com/world/2022/04/01/singapore-ends- 19 protective behaviors: A two-wave longitudinal study of epi-
outdoor-mask-mandate-covid/ demic and post-epidemic periods. Social Science & Medicine,
Lu, J. G., Jin, P., & English, A. S. (2021). Collectivism predicts 301, 114949. https://doi.org/10.1016/j.socscimed.2022.114949
mask use during COVID-19. Proceedings of the National Schneider, C. R., Dryhurst, S., Kerr, J., Freeman, A. L. J., Recchia,
Academy of Sciences of the United States of America, 118(23), G., Spiegelhalter, D., & van der Linden, S. (2021). COVID-19
e2021793118. https://doi.org/10.1073/pnas.2021793118/-/ risk perception: A longitudinal analysis of its predictors and
DCSupplemental associations with health protective behaviours in the United
Mahalik, J. R., Di Bianca, M., & Harris, M. P. (2021). Men’s atti- Kingdom. Journal of Risk Research, 24, 294–313. https://doi.
tudes toward mask-wearing during COVID-19: Understanding org/10.1080/13669877.2021.1890637
the complexities of mask-ulinity. Journal of Health Psychology, Schopf, J. C. (2022). Why tightness alone is not enough: The vary-
27(5), 1187–1204. https://doi.org/10.1177/1359105321990793 ing anti-pathogenic effects of rational values and cultural
Mahdavian, F., Warren, G. W., Evensen, D., & Bouder, F. E. tightness at different phases of the COVID-19 pandemic.
(2022). The relationship between barriers and drivers of Journal of Health Psychology, 27(13), 2936–2950. https://doi.
COVID-19 protective behaviors in Germany and the UK. org/10.1177/13591053211073861
International Journal of Public Health, 67, 1604970. https:// Talhelm, T., & English, A. S. (2020). Historically rice-farming
doi.org/10.3389/ijph.2022.1604970 societies have tighter social norms in China and worldwide.
National Health Commission of the People’s Republic of China. Proceedings of the National Academy of Sciences of the
(2020a). http://www.nhc.gov.cn/jkj/s3577/202003/0a472cc09 United States of America, 117(33), 19816–19824. https://doi.
e744144883db6a74fe6e760.shtml org/10.1073/pnas.1909909117/-/DCSupplemental
National Health Commission of the People’s Republic of China. Talhelm, T., Lee, C. S., English, A. S., & Wang, S. (2023). How
(2020b). http://www.nhc.gov.cn/xcs/yqtb/202005/59f0545ad2 rice fights pandemics: Nature–crop–human interactions shaped
7249c78d25ecee508369d4.shtml COVID-19 outcomes. Personality and Social Psychology Bulletin,
Nuclear Threat Initiative. (2019). 2019 Global Health Security Index. 49, 1567–1586. https://doi.org/10.1177/01461672221107209
https://www.ghsindex.org/wp-content/uploads/2019/10/2019- Talhelm, T., Zhang, X., Oishi, S., Shimin, C., Duan, D., Lan, X.,
Global-Health-Security-Index.pdf & Kitayama, S. (2014). Large-scale psychological differ-
Oxford University. (2023). COVID-19 Government Response ences within China explained by rice versus wheat agriculture.
Tracker. https://github.com/OxCGRT/covid-policy-tracker Science, 344(6184), 603–608. https://doi.org/10.1126/sci-
Price, R. H., & Bouffard, D. L. (1974). Behavioral appropriateness ence.1246850
and situational constraint as dimensions of social behavior. Uz, I. (2015). The index of cultural tightness and looseness among
Journal of Personality and Social Psychology, 30(4), 579–586. 68 countries. Journal of Cross-Cultural Psychology, 46(3),
https://doi.org/10.1037/h0037037 319–335. https://doi.org/10.1177/0022022114563611

You might also like