PreprintPDF Available

Perceived economic risk (vs. health risk) motivates individual efforts to fight COVID-19: A multilevel analysis in 24 countries

Authors:
Preprints and early-stage research may not have been peer reviewed yet.
ResearchGate Logo

This preprint is featured on the COVID-19 research community page

View COVID-19 community

Abstract and Figures

This paper examines whether compliance with COVID-19 mitigation measures is motivated by wanting to save lives or save the economy (or both), and which implications this carries to fight the pandemic. National representative samples were collected from 24 countries (N=25,435). The main predictors were (i) perceived risk to contract coronavirus, (ii) perceived risk to suffer economic losses due to coronavirus, and (iii) their interaction effect. Individual and country-level variables were added as covariates in multilevel regression models. We examined compliance with various preventive health behaviors and support for strict containment policies. Results show that perceived economic risk consistently predicted mitigation behavior and policy support - and its effects were positive. Perceived health risk had mixed effects. Only two significant interactions between health and economic risk were identified – both positive. These results do not corroborate the view that people engage in health versus economy zero-sum thinking in the fight against COVID-19.
Content may be subject to copyright.
Perceived economic risk (vs. health risk) motivates individual efforts to fight COVID-19: 1
A multilevel analysis in 24 countries 2
3
4
Authors 5
Claudia F. Nisa1*, Jocelyn J. Bélanger1, Daiane G. Faller2, Jochen O. Mierau3, Maura Austin4, 6
Nicholas R. Buttrick4, Birga M. Schumpe1, Edyta M. Sasin1, Maximilian Agostini5, Ben 7
Gützkow5, Jannis Kreienkamp5, The PsyCorona Team, N. P. Leander5 8
9
10
Affiliations 11
1 New York University Abu Dhabi, Department of Psychology, UAE 12
2 New York University Abu Dhabi, Center for Global Sea Level Change, UAE 13
3 University of Groningen, Faculty of Economics and Business, Netherlands 14
4 University of Virginia, Department of Psychology, USA 15
5 University of Groningen, Department of Psychology, Netherlands 16
17
The PsyCorona Team (Consortium) is fully detailed at the end of the manuscript including authors’ 18
names and affiliations 19
20
* Corresponding Author: cfn1@nyu.edu. New York University Abu Dhabi, Department of 21
Psychology, Saadiyat Island, PO BOX 129188, UAE 22
23
24
2
Abstract 25
This paper examines whether compliance with COVID-19 mitigation measures is motivated by 26
wanting to save lives or save the economy (or both), and which implications this carries to fight 27
the pandemic. National representative samples were collected from 24 countries (N=25,435). The 28
main predictors were (i) perceived risk to contract coronavirus, (ii) perceived risk to suffer 29
economic losses due to coronavirus, and (iii) their interaction effect. Individual and country-level 30
variables were added as covariates in multilevel regression models. We examined compliance with 31
various preventive health behaviors and support for strict containment policies. Results show that 32
perceived economic risk consistently predicted mitigation behavior and policy support - and its 33
effects were positive. Perceived health risk had mixed effects. Only two significant interactions 34
between health and economic risk were identified – both positive. These results do not corroborate 35
the view that people engage in health versus economy zero-sum thinking in the fight against 36
COVID-19. 37
38
39
3
The SARS-Coronavirus-2 Disease (COVID-19) pandemic is primarily a public health crisis. 40
Preventive health behaviors such as avoiding crowded spaces and social isolating are crucial 41
mitigation measures requested from the population in order to fight the spread of the COVID-19
42
(1). However, these mitigation measures rapidly produced unintended effects, generating a 43
collateral economic crisis, in the form of rising unemployment claims, income losses, and a 44
generalized uncertainty about global markets (2, 3). This challenge can be conceptualized as a risk-45
risk tradeoff (4): actions undertaken to minimize or eliminate certain risks to human health have 46
the perverse effect of promoting others, equally or more problematic than the original risk. This 47
tradeoff, occurring on a global scale, is an exceptional feature of this pandemic. 48
Here, we focus on risk perceptions about the COVID-19. Risk perceptions have proved 49
crucial to understand individuals’ attitudes and behaviors in the face of threat (4), and how people 50
weigh costs versus benefits when tackling hazards (5). Research about risk perception is prolific, 51
but mostly focuses on a single, primary hazard causing the threat – e.g., a virus, a hurricane, floods. 52
The dynamics that may occur with secondary or collateral risks has been subjected to less scrutiny. 53
However, this is a crucial point to examine under the current situation. In the COVID-19 pandemic, 54
the primary risk is considered to be contracting the virus, and the economic risk created by the 55
mitigation measures (e.g., unemployment, income loss) is regarded as a secondary risk, which 56
should be tolerated in order to address the primary risk. However, this secondary risk (economic) 57
has taken proportions that rival with the primary risk (health), to the point that some people claim 58
to be against following mitigation measures out of concerns for the economy (6). Anecdotal 59
evidence and media narratives commonly frame these risks as conflicting forces. But then again, 60
the question about whether economic (vs. health) concerns motivate or discourage following 61
public health measures has not received an empirical answer thus far – notwithstanding the heated 62
debate (1-3, 6). 63
The goal of this paper is to determine how perceived health risk versus perceived economic 64
risk due to the coronavirus are associated with (a) compliance with preventive health behaviors, 65
including frequent hand washing, avoiding crowded spaces and social isolation, and (b) support 66
for strict containment policies, comprising support for mandatory vaccination when developed, 67
support for mandatory quarantine for those infected or exposed to coronavirus, and reporting 68
suspected COVID-19 cases. This study focuses on individual-level psychological and behavioral 69
processes, although the analysis will control for a variety of macroeconomic and healthcare system 70
4
variables, previously shown to influence health behavior and health outcomes (7-11). The analysis 71
involves 24 countries from five continents that cover various levels of economic development and 72
different temporal stages of the COVID-19 pandemic. 73
Some economists and other scholars maintain that there should be no trade-off between 74
health and the economy (12). However, people are often presented with the binary, mutually 75
exclusive choice – should priority be given to save lives or save the economy? (13) – and 76
preferences tend to favor saving lives, suggesting a higher priority attributed to contain the virus 77
than to boost the economy. The hypothesis deriving from this result would be that risk perceptions 78
about getting infected with the virus should predict how much people comply with protective 79
behaviors and support the containment policies. This is also in line with the concerns expressed 80
about a partisan divide (14): the virus is being framed with different levels of lethality to distinct 81
political audiences, and these different perceptions about the virus gravity are suggested to 82
influence compliance with mitigation measures. 83
However, posing the problem as a mutually exclusive choice (lives vs. livelihoods) may 84
not fully capture the complexity of this issue nor provide the most accurate perspective about the 85
intricacies between health and economic risks. Notably, most health mitigation measures need to 86
be followed and sustained in order to safely reopen the economy. This further increases the 87
relevance to understand this association because policy measures impose restrictions and isolation 88
on individuals and households, who are also business owners, employees and consumers. 89
We will specifically examine whether perceived health and economic risks interact to 90
predict these outcomes. These risks may act synergistically to increase compliance with mitigation 91
measures (positive interaction), or in contrast, these risks may clash, meaning that perceiving a 92
high risk for both health and the economy may lead to conflicting views about mitigation measures 93
(negative interaction). The fact remains that, thus far, it is unclear whether fighting COVID-19 is 94
perceived as a choice between saving lives and saving the economy (or both). Both hypotheses 95
have been raised in national political arenas around the globe. This analysis is critical to inform 96
risk communication strategies that aim to be effective in achieving the public health targets. 97
This research responds to calls (15) to understand the psychological factors underlying 98
individuals’ response to this pandemic, mindful that the only approaches presently available to 99
reduce the transmission of coronavirus are behavioral, non-pharmaceutical interventions (16) - 100
largely dependent on voluntary compliance. Our primary data was collected during a critical 101
5
moment in the pandemic (April and May 2020), to examine whether health policy analysis should 102
consider not just governance-level guidelines, but also individual-level decision making as a 103
relevant dimension to understand compliance with policy measures. Policy guidelines may be 104
curtailed if these fail to effectively communicate the relevant risks or, as our data will show, focus 105
its communication on the wrong risks. 106
107
Results 108
All measures are fully described in Supplementary Table 1. Summary descriptive statistics per 109
country regarding sociodemographic variables, individual and country level covariates are 110
presented in Supplementary Tables 2 to 5. We start by illustrating the main variables at the 111
individual level with a series of descriptive statistics that control for potential cross-cultural 112
differences in response sets (17-18) (procedure described in Materials and Methods). This 113
descriptive analysis is followed by multilevel regression models that account for individuals nested 114
within countries. 115
116
Global risk perceptions about health and the economy 117
Figure 1 below presents both the perceived likelihood to get infected with coronavirus and the 118
perceived likelihood to suffer economic losses due to the coronavirus. Globally, average ratings 119
suggest a low perceived risk to get infected with the virus (M=3.23 SD=1.43 95% CI 3.21-3.24; 120
mean significantly below 4 or scale mid-point t(25370)=-86.19 p<.001; median=3). Regarding 121
economic risk perception, average perceptions suggest a moderate perceived risk (M=4.35 122
SD=1.80 95% CI 4.33-4.37; mean significantly above 4 or scale mid-point t(25382)=30.91 123
p<.001). Perceived health risk and economic risk are moderately correlated (r=.31 p<.001) (full 124
country breakdown per risk perception in Supplementary Table 3). 125
126
INSERT FIGURE 1 ABOUT HERE 127
128
Figure 2 further shows that people worldwide expect to suffer economically more than in 129
terms of health (all paired samples t-tests p<.01; Mdiff=-1.12 SD=1.92 95% CI -1.15, -1.10; 130
median=-1). Perceiving a higher risk to suffer economic losses, than to get infected with the virus, 131
is also a pattern consistent across sociodemographic groups. Different population groups regarding 132
6
age, gender, education, financial and employment status, and political ideology, unanimously 133
report a higher perceived economic risk (vs. health risk) due to the coronavirus (all paired t-tests 134
p<.001). More precisely, perceptions about health and economic risks differ between groups (e.g., 135
people under 25 perceive a lower health risk compared to all other ages), but perceived economic 136
risk is reliably higher in pairwise comparisons within all subgroups (further information about 137
differences per perceived risk within each sociodemographic category are presented in 138
Supplementary Materials - Figure 3). 139
140
INSERT FIGURE 2 ABOUT HERE 141
142
Global compliance and support for mitigation measures 143
We now turn to the analysis of global compliance with preventive health behaviors and support 144
for strict containment policies (country breakdown per outcome is presented in Supplementary 145
Tables 4 and 5). Overall, compliance with crowd avoidance is high (83% agree or strongly agree 146
complying/ supporting this measure), followed by frequent hand washing (81%), and to a lower 147
degree, social isolation from family and friends other than household members (55%). Regarding 148
the support for strict containment policies, the most supported measure would be mandatory 149
quarantine for those that have or have been exposed to coronavirus (73%). Both mandatory 150
vaccination for coronavirus when available (56%), and reporting suspicious COVID-19 cases 151
(57%) would be less approved. Figure 3 shows the density plots for these six outcomes. 152
153
INSERT FIGURE 3 ABOUT HERE 154
155
Association between risk perception and mitigation measures 156
To examine how risk perceptions are associated with these six outcomes, we conducted several 157
multilevel regression models. Given the hierarchical nature of the data, with individuals nested 158
within countries, multilevel regression was used to adjust for the dependence in the data and 159
possible confounders (step-by-step analyses taken, and detailed parameters of the models are fully 160
described in Materials and Methods). All models controlled for COVID-19 case-fatality rate per 161
country (total COVID-19 deaths per million/ total COVID-19 cases per million). The models also 162
tested a quadratic term for health risk due to exploratory visual analyses suggesting curvilinear 163
7
relationships between health risk (but not economic risk) and several outcomes (see exploratory 164
plots in Supplementary Figure 4). Moreover, individual and country-level covariates were included 165
in the last step (Model 2), informed by previous research as potential predictors of health behavior 166
and health outcomes (7-11) (covariates are fully described in Materials and Methods, and 167
Supplementary Tables 1 and 2; regression coefficients for covariates per regression model 168
presented in Supplementary Tables 6 and 7). The multilevel regression models predicting 169
preventive health behaviors are displayed in Table 1 below, while models predicting support for 170
strict containment policies are presented in Table 2. 171
Intraclass correlation coefficients (ICC) are very small across all models, particularly when 172
individual and country covariates are introduced in Model 2. The proportion of variance explained 173
by country ranges between 2% and 12%, suggesting the relationship between perceived risks and 174
following mitigation measures is mostly explained by individual differences across countries. 175
Results show that perceived economic risk consistently predicts following the COVID-19 176
mitigation measures, and that this relationship is linear. The more people perceive themselves to 177
be at risk of suffering economic losses due to the coronavirus, the more people comply with all 178
preventive health behaviors and support strict compliance policies. Alternatively, individuals' 179
perceived risk of getting infected with the coronavirus had no association with the two most 180
followed preventive health behaviors: frequent hand washing (linear B=.03 p=.11; quadratic B=-181
.01 p=.07) and avoiding crowded spaces (linear B=.02 p=.06; quadratic B=-.01 p=.13). Positive 182
linear associations for health risk were identified with support for two strict compliance measures: 183
the more people perceived a personal health risk, the more they support mandatory vaccination 184
(B=.06 p<.001) and support mandatory quarantine (B=.03 p=.02). Moreover, curvilinear 185
relationships with health risk were also found. Social isolation has a negative quadratic association 186
with health risk (B=-.02 p<.001). This suggests people increasingly comply with social isolation 187
up to when their perceived infection risk increases to a moderate level, but this compliance 188
decreases when health risk is perceived to be very high. In contrast, health risk has a positive 189
quadratic association with support for reporting suspect cases (B=.02 p<.001). This implies that 190
people show reduced support for reporting possible COVID-19 cases as their own personal risk 191
increases, but only up to the point of moderate risk. For high levels of perceived health risk, people 192
are more supportive of reporting suspect cases. Although the curvilinear patterns are idiosyncratic, 193
8
altogether they illustrate that increases in perceived health risk are not a reliable predictor of 194
compliance with mitigation measures. 195
196
INSERT TABLE 1 ABOUT HERE 197
198
199
INSERT TABLE 2 ABOUT HERE 200
201
Interaction effects were significant in two out of six cases, both positive for frequent hand 202
washing (B=.02 p<.001) and support for mandatory vaccination (B=.01 p=.02). Figure 4 below 203
plots these positive interactions. Without controlling for individual and country covariates, the 204
interaction between these risks is negative for social isolation (B=-.02 p<.01), but no longer reaches 205
significance in Model 2 (p=.07). 206
207
INSERT FIGURE 4 ABOUT HERE 208
209
We conducted a sensitivity analysis to examine the extent to which the results would hold 210
in the subgroup with the lowest economic risk perception: people with no or low financial 211
insecurity. It could be argued that the global pattern of results reflected a generalized modest 212
financial situation by the participants. Nonetheless, results restricted to people who perceive to be 213
financially comfortable largely corroborate the global results (Supplementary Table 8). In this 214
subgroup, the three most accepted measures were also positively predicted by economic risk (hand 215
washing B=.06 p=.05; avoid crowded spaces B=.06 p<.01; mandatory quarantine B=.07 p<.001). 216
This implies that, although people are financially secure, increases in their perceived economic 217
liability are associated with following these measures more. Similar results were also found for 218
mandatory vaccination, predicted by health risk as in the global results (B=.06 p<.001). Differences 219
were found in the two instances that exhibit more complex relationships between risk perception 220
and behavior: social isolation and reporting suspect cases. Social isolation was only predicted by 221
health risk (B=.04 p<.01) and risk perception was not associated with support for reporting suspect 222
COVID-19 cases. No interactions nor quadratic effects were found in this subgroup, suggesting 223
9
that financial security simplifies individuals’ psychological rapport with COVID-19 mitigation 224
measures. 225
226
Discussion 227
This work sheds an empirical light into one the most heated debates in the era of COVID-19. We 228
examined global risk perceptions regarding contracting the virus and suffering economic losses 229
due to the pandemic, and both their association with compliance and support for the mitigation 230
measures to fight COVID-19. The key takeaway is that, globally, people are not perceiving saving 231
lives and saving the economy as dueling goals. This work suggests that that the path forward 232
should not be to ignore the virus nor minimize its dangers to reopen the economy, nor to focus on 233
health vulnerabilities and lives lost to increase preventive health behaviors. Inversely, public 234
messaging may be more effective if delivering the message that COVID-19 mitigation measures 235
need to be followed to avoid (further) economic and job losses. This key takeaway derives from a 236
number of important results uncovered in this work. 237
First, on average, global risk perceptions are low to moderate. Despite the widespread 238
disarray created by the coronavirus, relentless media coverage, and the high volume of cases and 239
death toll, people perceive contracting the virus as an unlikely event. Across all countries 240
examined, the highest perceived likelihood to get infected with coronavirus only reached a fifty-241
fifty chance. Perceived as a more likely prospect is the risk of suffering economic consequences 242
due to the coronavirus. Average economic risk perceptions are moderate: people think that 243
experiencing economic losses is somewhat likely. The higher perceived economic risk (vs. the 244
health risk) from the coronavirus is a remarkably consistent pattern across all countries and social 245
groups regardless of age, gender, education, employment and financial status and political 246
ideology. These results suggest that risk perceptions seem to accurately reflect the objective 247
probabilities reported by international organizations regarding both risks. The probability to get 248
infected with the virus is considered to be low to moderate for the general population (19), whereas 249
the probability to suffer economic losses is nearly 50% for the global workforce (20). Therefore, 250
at the aggregate level, people seem to be correctly assessing their relative vulnerability regarding 251
these risks. 252
Second, perceived economic risk – and not health risk - is the main predictor of mitigation 253
behavior and policy support. Moreover, its effects are positive. According to our data, only 254
10
economic concerns positively predicted all outcomes. This association is unrelated to the fact that 255
economic risk is perceived to be higher; instead, it indicates that it is the variation in perceived 256
economic risk that is co-varying with changes in compliance and support for COVID-19 measures. 257
The more people perceive a personal risk to suffer economic losses due to the pandemic, the more 258
they frequently wash their hands, avoid crowds, socially isolate, support mandatory vaccination, 259
mandatory quarantine for those that have coronavirus or who have been exposed to the virus, and 260
support reporting suspected COVID-19 cases. Based on these results, the view (6) that some 261
people seem to be against following mitigation measures because of their concerns about the 262
economy is not supported as a mainstream perspective. 263
Perceived health risk exhibited mixed effects. The strongest associations with health risk 264
were support for mandatory vaccination and mandatory quarantine. Null effects were found for 265
the two most followed preventive health behaviors: frequent hand washing and avoiding crowded 266
spaces. Furthermore, results also showed quadratic effects of health risk on support for the strictest 267
measures such as social isolation and reporting suspect COVID-19 cases. Regarding social 268
isolation, if people perceive contracting the virus as very unlikely, the sacrifice to socially isolate 269
may not seem worth it. If personal virus infection risk increases too much, people don’t want to be 270
isolated from friends and family, possibly as a coping mechanism against rising anxiety and fear. 271
Regarding the support for reporting suspect cases, results imply that the burden of reporting 272
suspected COVID-19 cases would only be undertaken when people perceive themselves either at 273
a very low or very high health risk. That is, they would only support such a measure when they 274
think it could not happen to them, or when the fear of infection is so high that it justifies drastic 275
action. There is a precedent for people having conflicted psychological attitudes towards restrictive 276
policies, often more supported when it mostly affects others, but assessed negatively when it 277
affects themselves (21). This suggests that while strict policies are expected to better contain the 278
virus spread, more moderate measures may have higher public acceptability and less behavioral 279
backlash. 280
Third, few significant interactions between health risk and economic risk were identified, 281
and when found, these were positive interactions. These risks do not appear to work as competing 282
forces, but mostly as independent main effects that positively contribute towards mitigation 283
behavior – with a stronger contribution from economic risk. In the case of the positive interactions 284
identified, health and economic risk collaborate to increase frequent hand washing and supporting 285
11
mandatory vaccination. We interpret this positive interaction as a sign that neither of these 286
measures affect economic activities, and both protect personal and public health. Our data did not 287
include willingness to wear face masks in public nor compliance with public social distancing, 288
although our results suggest that these could also be instances of a positive interaction between 289
health and economic risks. Both face masks and keeping a distance from others in public spaces 290
protect health while preserving the continuity of economic activities. No significant negative 291
interactions were identified, which could have been expected for measures that protect health at 292
the cost of reduced economic interactions i.e., mandatory quarantine and social isolation. 293
Therefore, overall, this paper does not suggest corroboration for the narrative that regular people 294
engage in the health vs. economy zero-sum thinking, often disseminated in journalistic, political 295
and business messaging. 296
Last, there were null effects from case-fatality rates, included in all models as a control 297
variable. The number of COVID-19 deaths and cases, and their ratio (case-fatality rate), are some 298
of the most frequently publicized pieces of information about the pandemic, yet seemingly 299
unrelated to following protective health behaviors and supporting containment measures, with or 300
without controlling for covariate factors. This may suggest the need to shift public health 301
messaging away from COVID-19 health statistics, and more towards economic statistics. 302
In conclusion, we show that economic concern is a better predictor of virus prevention 303
behavior and support for strict health policies to contain the virus, compared to the concern about 304
getting infected with coronavirus. In other words, some people may deny the seriousness of the 305
virus (14) but fewer are denying the economy is being affected. Hence, a focus on economic threats 306
is universally shared and can be a way to unify people around a common goal. This raises the 307
question of whether appealing to personal economic risk is a more effective way to motivate virus 308
mitigation behavior, rather than appealing to personal virus (health) risk. 309
Nonetheless, some limitations in this work should be addressed in future research. An 310
important point is that no causality can be attributed to risk perception in its effects on mitigation 311
behavior and policy support. Cross-sectional designs are liable to the possibility of reverse 312
causality, by which it would be following mitigation measures that decreases perceived (and 313
objective) risk. Although this is an open possibility, we argue that it is unlikely that frequent hand 314
or avoiding crowds would reduce perceived economic risk, but not health risk. Furthermore, the 315
logic of reverse causality would only apply to personal behaviors reducing personal risk, but less 316
12
so to how more positive attitudes towards potential containment policies decrease perceived risk. 317
We maintain that our version of causality is more parsimonious across all outcomes. Nonetheless, 318
other research designs (e.g., longitudinal studies, quasi-experimental designs examining survey 319
data in individuals affected by different lockdown measures) are needed to establish the direction 320
of this relationship more conclusively (22). 321
Another noteworthy point is that, given the large sample sizes involved, effects small in 322
magnitude were statistically significant results. This applies both to main effects and interaction 323
effects. Therefore, even though economic risk seems to be a better predictor of compliance and 324
support for mitigation measures, compared to health risk, both these factors offer a low 325
contribution to understand what drives people to follow COVID-19 measures. Nonetheless, small 326
effects can add up to substantial effects when scaled-up to the population level (23). For example, 327
even though smoking is one of the greatest behavioral risk factors for developing lung cancer or 328
heart disease, the 10-year absolute risk for a heavy smoker to develop lung cancer is only 0.3 329
percent and the risk of developing heart disease is only 0.9 percent (24). And yet, these small 330
effects have tremendous significance from a population perspective, with hundreds of thousands 331
of heavy smokers dying prematurely. Given that the COVID-19 pandemic literally has a global 332
reach, small effects matter. Therefore, risk communication strategies that potentially influence risk 333
perceptions about personal risk may add up to a substantial increase in compliance and support for 334
mitigation measures. 335
A concluding remark is that future research should explore further the role of country/ 336
culture characteristics (17-18) in modulating individual perceptions about the health and economic 337
risks posed by the COVID-19. Countries differ in the characteristics of their healthcare (e.g., no 338
access to free healthcare) and economic systems (e.g., high unemployment rate), and in their 339
overall organizational capacity to buffer the population from this challenge. Our analysis did not 340
dwell upon this subject, although our results from ICC and country-level covariate analysis suggest 341
that country differences play a small role. Nonetheless, a more in-depth cross-country analysis 342
may uncover the need for a cultural adjustment to risk communication. 343
344
Materials and Methods 345
Study Design and Data Collection 346
13
This cross-sectional study is part of the global Psycorona project (https://psycorona.org) which 347
focuses on how people feel and think about the coronavirus epidemic and its economic 348
consequences. This study complied with ethical regulations for research on human subjects and all 349
participants gave informed consent, as approved by the Institutional Review Board at New York 350
University Abu Dhabi (protocol HRPP-2020-42) and the Ethics Committee of Psychology at 351
Groningen University (protocol PSY-1920-S-0390). Personal identifiers were removed from all 352
sections of the manuscript, including supplementary information and public dataset. 353
Survey responses were collected through Qualtrics’ panel management service, except in 354
China where data was collected by WJX, following a similar methodological approach. The 355
company's methodology involves obtaining responses from invited internet users drawn from its 356
panel of over 90 million people worldwide. Data was collected in 24 countries: Argentina, 357
Australia, Brazil, Canada, China, France, Germany, Greece, Indonesia, Italy, Japan, Netherlands, 358
Philippines, Romania, Russia, Saudi Arabia, Serbia, South Africa, South Korea, Spain, Turkey, 359
United Kingdom, Ukraine, and the United States of America. These countries cover various levels 360
of economic development as well as different temporal stages of the COVID-19 pandemic. 361
National proportional quota samples were collected with a 3% margin of error and 95% confidence 362
level, representative of the country’s population in terms of gender and age (age representativeness 363
was less achieved in China, Greece, Saudi Arabia, Indonesia and the Philippines, where people 364
aged 55+ were less present in the survey). Data quality control was conducted by (i) examining IP 365
addresses to detect potential duplicate responders; and (ii) removing participants from the database 366
whose answers indicated random responses. Data was collected online between 10th April and May 367
11th 2020. 368
369
Measures 370
All measures are fully described in the Supplementary Table S1. The main predictors were the 371
perceived likelihood to get infected with coronavirus, the perceived likelihood to suffer economic 372
consequences due to the coronavirus, and their interaction effect. A total of six outcomes were 373
predicted. The primary outcomes were compliance with preventive health behaviors, including 374
frequent hand washing, avoiding crowded spaces and social isolation (i.e., no contact with friends 375
and family other than household members). The secondary outcomes were the support for strict 376
health measures, namely support for mandatory coronavirus vaccination (when developed), 377
14
mandatory quarantine for those that have coronavirus and those that have been exposed to the 378
virus, and reporting of suspected coronavirus cases. We chose to examine these items individually, 379
as informative in their own right. However, single item measures do not allow for an internal 380
consistency analysis. Nonetheless, a reliability analysis of the six items (outcomes) – as a measure 381
of overall acceptability of public health measures - reveals a good internal consistency (α=.77). 382
Several individual and country-level predictors - previously shown to be associated with 383
preventive health behavior and health outcomes (7-11) - were added as covariates in multilevel 384
regression models. Individual-level covariates were (i) direct exposure to someone (self, family, 385
friends) in their personal network infected with COVID-19; (ii) knowledge about the COVID-19, 386
(iii) knowledge about the economic consequences of the COVID-19; (iv) the quality of the public 387
messages received, (v) community norms about mitigation measures, and (vi) sociodemographic 388
variables (age, gender, education, employment and financial status, and political ideology). As part 389
of a larger research project PsyCorona, there were other psychological measures collected that 390
were ultimately not selected as covariates due to low theoretical justification. The full set of 391
psychological measures collected can be found here https://psycorona.org/about/. Country-level 392
covariates included (i) total population of the country (in millions), (ii) gross domestic product 393
(GDP) per capita (in current $US), (iii) unemployment rate (as % of the labor force), (iv) old age 394
dependency ratio (%), (v) Gini Index, (vi) general health expenditure (as %GDP), (viii) private 395
health expenditure (as % health expenditure), (viii) out-of-pocket health payments (as % health 396
expenditure), (ix) number of hospital beds (per 1000 people). 397
398
Sample 399
Summary statistics for each country regarding sociodemographic variables and individual and 400
country level covariates are presented in the Supplementary Table 2. At the aggregate level, the 401
sample was gender balanced (51% women), with 52% up to 44 years of age, and 48% aged 45 to 402
old age (range 18-85). Most participants were educated up to completed high school (59%), and 403
the remaining with a completed higher education (19% with Bachelor degree and 13% with 404
postgraduate studies). Most participants were employed (57%, either part- or full-time), and about 405
a third (35%) reported difficulties paying for their expenses. Politically, 40% self-categorize as 406
left leaning, whereas 50% self-categorize as right leaning (about 10% other/ no political 407
15
preference). The analysis includes participants who have already contracted the virus (n=142) 408
and/ or who have already lost their jobs (n=1295). 409
410
Statistical Analysis 411
To the best of our knowledge, previous literature about multiple risks or risk interaction was slim 412
to confidently propose or guide in hypotheses formulation. Thus, we opted for not formalizing nor 413
pre-registering any hypotheses (25). We conducted exploratory analyses examining the relative 414
association between health and economic risks and multiple outcomes related to following 415
mitigation measures. This analysis controlled for several covariates at the individual and country 416
level, theoretically justified (7-11). 417
Descriptive differences between countries and between risk perceptions were examined 418
with analysis of variance (ANOVA), LSD and Tukey HSD post hoc tests, and paired samples t-419
tests. We classified correlations (r) and betas as small if between 0.05 and 0.19, moderate between 420
0.20 and 0.49, and large if above 0.50, as characteristic in the social sciences (26). 421
Different response sets between countries were controlled for by standardizing health and 422
economic risk for the cross-country comparisons in the descriptive statistics. Raw scores on risk 423
perception and the six outcomes were averaged to create a within-subject response average. This 424
average was then subtracted from the raw scores of perceived health risk and perceived economic 425
risk to generate standardized scores for these two variables (17-18). Given that this procedure did 426
not change the average results and country comparisons, we presented the raw score for a better 427
interpretability by the reader. We, nevertheless, present the standardized health and economic risk 428
standard scores per country in Supplementary Figures 1 and 2.
429
We estimated the Intraclass Correlation Coefficient (ICC) to describe the correlation 430
among observations within the countries. The ICC is also equivalent to the variance partition 431
coefficient, which can be interpreted as the proportion of variation that is due to a variation 432
between countries (27). 433
We also applied hierarchical models (27) to understand the effects of controlling for 434
person-level predictors taking into account the random variations across nations. In preparation to 435
run these models, we eliminated the missing values from the entire dataset (n=592), considering 436
all the variables together. If a subject had missing values, all variables from the subject were 437
eliminated. We detected the multivariate outliers using Mahalanobis’ distance and chi-square 438
16
distribution (ɑ=0.95) with a total of 2282 eliminated. The total sample used in the models was 439
N=22561, constant across models. The predictors from the individual-level were group-mean-440
centering by country (and scaling is done by dividing the (centered) columns of x by their standard 441
deviations). Country-level variables used grand-mean-centering, given these have a single value 442
for each country. The models were implemented using R and the package lme4 (29-30). To predict 443
each of the six outcomes, a total of three nested models were selected from a range of 15 models 444
(using ANOVA approach for between model comparison).
445
The selected models vary in increasing complexity. All models controlled for COVID-19 446
case-fatality rate: total COVID-19 deaths per million/ total COVID-19 cases per million. Model 447
0: Model 0 or empty model provided unadjusted rates for the behavior response (outcome) that 448
accounted for clustering. Model 1: included the individual-level variables perception of risk of 449
infection and economic loss (and their interaction), a quadratic term for health risk (and the 450
interaction with economic risk) as predictors for fixed effects and perception of risk of infection 451
and economic loss as random intercept within the country. The use of the random statement 452
measures the variance in the effects of risk of infection and economic loss on behavior responses 453
across countries. Interaction was not used as a random effect because it led to a non-convergence. 454
Model 2: Same as model one plus the individual-level and country-level covariates described in 455
the section Measures above. Model 2 added these covariates as fixed effects. Political ideology 456
was not included in the multilevel regression analysis due to the high number of missing values in 457
most countries and a complete absence of replies in China. This decision was due to wanting to 458
keep the sample size across all models (N=22561). Nonetheless, regression models including 459
political ideology were conducted as a sensitivity analysis and results held across the models.
460
As we used linear mixed models, the variables were checked for normal distribution, scaled 461
in relation to the mean, and extreme outliers were excluded. Models used the Nelder-Mead 462
optimization algorithm for derivative-free optimization. All reported P values are two-sided. 463
464
465
466
17
References 467
1. McKee, M., & Stuckler, D. If the world fails to protect the economy, COVID-19 will damage 468
health not just now but also in the future. Nat Med (2020). https://doi.org/10.1038/s41591-020-469
0863-y 470
2. Cutler D. How Will COVID-19 Affect the Health Care Economy? JAMA 323, 2237-8 (2020). 471
3. Ichino, Andrea, Carlo A. Favero, & Aldo Rustichini. Restarting the economy while saving 472
lives under Covid-19. (2020) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3594296## 473
4. Graham, J.D., and J.B. Wiener. Risk vs. Risk: Tradeoffs in protecting public health and the 474
environment. Cambridge, MA: Harvard University Press (1995). 475
5. Gregorian Jr RS, Gasik A, Kwong WJ, Voeller S, & Kavanagh S. Importance of side effects in 476
opioid treatment: a trade-off analysis with patients and physicians. J. Pain 11, 1095-108 (2010). 477
6. NY Times. Coronavirus-Economy debate (retrieved April 10th 2020) 478
https://www.nytimes.com/2020/04/10/magazine/coronavirus-economy-debate.html 479
7. Wakefield MA, Loken B, & Hornik RC. Use of mass media campaigns to change health 480
behaviour. The Lancet 376, 1261-71 (2010). 481
8 Vinck P, Pham PN, Bindu KK, Bedford J, & Nilles EJ. Institutional trust and misinformation 482
in the response to the 2018–19 Ebola outbreak in North Kivu, DR Congo: a population-based 483
survey. The Lancet Inf Diseas 19, 529-36 (2019). 484
9. Zingg W, Holmes A, Dettenkofer M, Goetting T, Secci F, Clack L, Allegranzi B, Magiorakos 485
AP, & Pittet D. Hospital organisation, management, and structure for prevention of health-care-486
associated infection: a systematic review and expert consensus. The Lancet Inf Diseas 15, 212-24 487
(2015). 488
10. Dooley D, Fielding J, Levi L. Health and unemployment. Annual Rev Public Heal 17, 449-65 489
(1996). 490
11. Retzlaff-Roberts D, Chang CF, Rubin RM. Technical efficiency in the use of health care 491
resources: a comparison of OECD countries. Health Pol 69, 55-72 (2004). 492
12. Stock J. Reopening the Coronavirus-Closed Economy. Tech. rep., Hutchins Center Working 493
Paper; 2020 https://www.brookings.edu/wp-content/uploads/2020/05/WP60-Stock_final.pdf 494
13. Washington Post. Despite widespread economic toll, most Americans still favor controlling 495
outbreak over restarting economy (Retrieved June 1st 2020). 496
https://www.washingtonpost.com/politics/despite-widespread-economic-toll-most-americans-497
18
still-favor-controlling-outbreak-over-restarting-economy-post-abc-poll-498
finds/2020/06/01/3e052ec0-a27b-11ea-81bb-c2f70f01034b_story.html 499
14. Allcott H, Boxell L, Conway J, Gentzkow M, Thaler M, Yang DY. Polarization and public 500
health: Partisan differences in social distancing during the Coronavirus pandemic. NBER 501
Working Paper (2020) http://web.stanford.edu/~gentzkow/research/social_distancing.pdf 502
15. Lipsitch M, Swerdlow DL, & Finelli L. Defining the epidemiology of Covid-19—studies 503
needed. New England J Med 382, 1194-6 (2020). 504
16. Haushofer J, & Metcalf CJ. Which interventions work best in a pandemic? Science 5, 505
368(6495), 1063-5 (2020). 506
17. Gelfand, M. J., Raver, J. L., Nishii, L., Leslie, L. M., Lun, J., Lim, B. C., & Aycan, Z. 507
Differences between tight and loose cultures: A 33-nation study. Science 332, 1100-1104 (2011). 508
18. Chiu, C. Y., Gelfand, M. J., Yamagishi, T., Shteynberg, G., & Wan, C. Intersubjective 509
culture: The role of intersubjective perceptions in cross-cultural research. Persp Psychol Sci 5, 510
482-493 (2010). 511
19. European Centre for Disease Prevention and Control. Risk assessment on COVID-19, 11 512
June 2020 https://www.ecdc.europa.eu/en/current-risk-assessment-novel-coronavirus-situation 513
20. International Labor Organization. As job losses escalate, nearly half of global workforce at 514
risk of losing livelihoods (Retrieved April 29 2020) https://www.ilo.org/global/about-the-515
ilo/newsroom/news/WCMS_743036/lang--en/index.htm 516
21. Zippay AL. Psychiatric residences: Notification, NIMBY, and neighborhood relations. Psych 517
Serv 58, 109-13 (2007). 518
22. Nisa, C. F., Bélanger, J. J., Schumpe, B. M. & Faller, D. G. Meta-analysis of randomised 519
controlled trials testing behavioural interventions to promote household action on climate 520
change. Nat. Commun. 10, 4545 (2019). 521
23. Prentice, D. & Dale M. When small effects are impressive. Psychol Bull 112, 160 (1992). 522
24. Jeffery, R. Risk behavior and health: Contrasting individual and population perspectives. Am 523
Psychol, 44, 1194-1202 (1989). 524
25. Asendorpf J., Conner M., De Fruyt F., De Houwer J., Denissen J., Fiedler K., Fiedler S., 525
Funder D., Kliegl R., Nosek B., & Perugini M. Recommendations for Increasing Replicability in 526
Psychology. Europ. J. Pers. 27, 108-19 (2013). 527
19
26. Fritz, C.O., Morris, P.E. & Richler, J.J. Effect size estimates: current use, calculations, and 528
interpretation. J Exp Psyc: General, 141, 2 (2012). 529
27. Austin PC, & Merlo J. Intermediate and advanced topics in multilevel logistic regression 530
analysis. Stat Med 36, 3257-77 (2017). 531
28. Raudenbush SW, Bryk AS, Cheong YF, Congdon R, & Du Toit M. HLM 6: Hierarchical 532
linear and nonlinear modeling. Scientific Software International. Inc., Lincolnwood, IL. (2000). 533
29. Bates D, Maechler M, Bolker B, Walker S, Christensen RH, Singmann H, Dai B, 534
Grothendieck G, Green P, & Bolker MB. Package ‘lme4’. Convergence 12, 2 (2015). 535
30. Finch WH, Bolin J. E., & Kelley, K. (2014). Multilevel Modeling Using R. CRC Press. 536
537
538
539
20
PsyCorona Team 540
541
Georgios Abakoumkin6, Vjollca Ahmedi7, Handan Akkas8, Carlos A. Almenara9, Anton 542
Kurapov10, Mohsin Atta11, Sabahat Cigdem Bagci12, Sima Basel1, Phatthanakit Chobthamkit13, 543
Hoon-Seok Choi14, Mioara Cristea15, Sára Csaba16, Kaja Damnjanovic17, Ivan Danyliuk10, 544
Arobindu Dash18, Daniela Di Santo19, Karen M. Douglas13, Violeta Enea20, Gavan Fitzsimons21, 545
Alexandra Gheorghiu20, Ángel Gómez22, Qing Han23, Mai Helmy24, Bertus Jeronimus5, Ding-Yu 546
Jiang25, Veljko Jovanovic26, Željka Kamenov27, Anna Kende28, Shian-Ling Keng29, Jamilah H. B. 547
Abdul Khaiyom30, Edona Berisha Kida7, Tra Thi Thanh Kieu31, Yasin Koc5, Kamila Kovyazina32, 548
Inna Kozytska10, Joshua Krause5, Arie W. Kruglanski33, Maja Kutlaca34, Nóra Anna Lantos28, 549
Edward Lemay33, Cokorda Bagus Jaya Lesmana35, Winnifred R. Louis36, Adrian Lueders37, Najma 550
Malik11, Anton Martinez38, Kira McCabe39, Mirra Noor Milla40, Idris Mohammed41, Erica 551
Molinario33, Manuel Moyano42, Hayat Muhammad43, Silvana Mula19, Hamdi Muluk40, Solomiia 552
Myroniuk5, Reza Najafi44, Boglárka Nyúl28, Paul A. O’Keefe29, Jose Javier Olivas Osuna22, 553
Evgeny N. Osin45, Joonha Park46, Gennaro Pica19, Antonio Pierro19, Jonas Rees47, Anne Margit 554
Reitsema5, Elena Resta19, Marika Rullo48, Michelle K. Ryan5, 49, Adil Samekin50, Pekka Santtila51, 555
Heyla A Selim52, Michael Vicente Stanton53, Samiah Sultana5, Robbie M. Sutton13, Eleftheria 556
Tseliou54, Akira Utsugi55, Jolien Anne van Breen56, Caspar J. Van Lissa57, Kees Van Veen5, 557
Michelle R. vanDellen58, Alexandra Vázquez22, Robin Wollast37, Victoria Wai-lan Yeung59, 558
Somayeh Zand44, Iris Lav Žeželj17, Bang Zheng60, Andreas Zick47, Claudia Zúñiga61 559
560
561 6 University of Thessaly, Laboratory of Psychology, Department of Early Childhood Education, 562
Greece 563
7 Pristine University, Pedagogy, Kosovo 564
8 Ankara Science University, Organizational Behavior, Turkey 565
9 Universidad Peruana de Ciencias Aplicadas, Faculty of Health Science, Peru 566
10Taras Shevchenko National University of Kyiv, Department of Psychology, Ukraine 567
11University of Sargodha, Department of Psychology, Pakistan 568
12Sabancı University, Department of Psychology, Turkey 569
13University of Kent, Department of Psychology, UK 570
14Sungkyunkwan University, Department of Psychology, Korea 571
21
15Heriot Watt University, Department of Psychology, UK 572
16ELTE Eötvös Loránd University, Doctoral School of Psychology, Hungary 573
17University of Belgrade, Department of Psychology, Serbia 574
18International University of Business Agriculture & Technology (IUBAT), Department of 575
Psychology, Bangladesh 576
19University "La Sapienza" Rome, Department of Social and Developmental Psychology, Italy 577
20Alexandru Ioan Cuza University, Romania 578
21Duke University, Marketing and Psychology, USA 579
22Universidad Nacional de Educación a Distancia, Social and Organizational Psychology, Spain 580
23University of Bristol, The School of Psychological Science, UK 581
24Menoufia University, Department of Psychology, Egypt 582
25National Chung-Cheng, University Department of Psychology, Taiwan 583
26University of Novi Sad, Department of Psychology, Serbia 584
27University of Zagreb, Faculty of Humanities and Social Sciences, Croatia 585
28Eötvös Loránd University, Department of Social Psychology, Hungary 586
29Yale-NUS College, Division of Social Science, Singapore 587
30 International Islamic University, Department of Psychology, Malaysia 588
31Tra Thi Thanh Kieu, HCMC University of Education, Department of Psychology, Vietnam 589
32 Independent Researcher, Kazakhstan 590
33University of Maryland, Department of Psychology, USA 591
34Durham University, University of Osnabrück, Department of Psychology, UK 592
35Udayana University, Department of Psychiatry, Indonesia 593
36University of Queensland, School of Psychology, Australia 594
37Université Blaise Pascal, Laboratoire de Psychologie Sociale et Cognitive, France 595
38University of Sheffield, Department of Psychology, Argentina/UK 596
39Vanderbilt University, Psychology and Human Development, USA 597
40Universitas Indonesia, Department of Psychology, Indonesia 598
41Usmanu Danfodiyo University Sokoto, Mass Communication, Nigeria 599
42University of Cordoba, Department of Psychology, Spain 600
43University of Peshawar, Department of Psychology, Pakistan 601
44Islamic Azad University of Rasht, Department of Psychology, Iran 602
22
45National Research University Higher School of Economics, Department of Psychology, Russia 603
46NUCB Business School, Graduate School of Management, Japan 604
47University of Bielefeld, Research Institute Social Cohesion, Institute for Interdisciplinary 605
Research on Conflict and Violence, and Department of Social Psychology, Germany 606
48University of Siena Department of Educational, Humanities and Intercultural Communication, 607
Italy 608
49University of Exeter, Psychology, UK 609
50 S. Toraighyrov Pavlodar State University, Department of Psychology and Pedagogy, 610
Kazakhstan 611
51New York University Shanghai, Department of Psychology, China 612
52 King Saud University, Department of Psychology, Saudi Arabia 613
53 California State University, East Bay, Health Sciences, USA 614
54University of Thessaly, Laboratory of Psychology, Department of Early Childhood Education, 615
Greece 616
55Nagoya University, Graduate School of Humanities, Japan 617
56University of Exeter, Department of Psychology, UK 618
57Utrecht University, Department of Methodology & Statistics, Netherlands 619
58 University of Georgia, Department of Psychology, USA 620
59Lingnan University, Department of Psychology, Hong Kong 621
60 Imperial College London, Ageing Epidemiology Research Unit, School of Public Health, 622
Faculty of Medicine, UK 623
61Universidad de Chile, Department of Psychology, Chile 624
625
626
627
628
629
23
Acknowledgments 630
631
Funding: This work was funded by New York University Abu Dhabi through Vice-Chancellor 632
Support for COVID-19 research in the Department of Psychology. This funding was attributed to 633
JJB. This work also received funding from Groningen University via the Ubbo Emmius Fund. This 634
funding was attributed to NPL. 635
636
Competing interests: The authors declare no competing interests. 637
638
Data and materials availability: The dataset and an example of code for R lme4 are publicly 639
available at the Open Science Framework https://osf.io/xvyna/ 640
641
Ethics Approval. This study complied with ethical regulations for research on human subjects 642
and all participants gave informed consent, as approved by the Institutional Review Board at New 643
York University Abu Dhabi (protocol HRPP-2020-42) and the Ethics Committee of Psychology 644
at Groningen University (protocol PSY-1920-S-0390). 645
646
Author contributions: CFN was responsible for conceptualization, writing both initial draft and 647
review & editing, and supervision. JJB was involved in writing review & editing, supervision, 648
project administration and funding acquisition. DGF was the lead responsible for formal analysis. 649
JOM was involved in validation. MauA was the lead responsible for visualization. NRB was 650
involved in validation and formal analysis. BMS was involved in writing review & editing. EMS 651
was involved in validation and formal analysis. MaxA was responsible for data curation and 652
project management. BG was responsible for data curation and project management. JK was 653
responsible for data curation and project management. The PsyCorona team comprises a bundle 654
of authors presented in Supplementary Materials. At this stage, we present this team as a single 655
entity because the contribution of all authors in this group was similar. These authors equally 656
contributed to methodology, software, investigation and resources. NPL was involved in writing 657
review & editing, supervision, project administration and funding acquisition. 658
659
24
Figures 660
661
662
Fig. 1. Perceived health risk versus perceived economic risk due to the coronavirus. Note: 663
Raw Scores, Error Bars 95% CI. Standardized scores correcting for cross-cultural response 664
sets returned the same country comparative hierarchy per risk. Standardized scores are 665
presented in Supplementary Figures 1 and 2. 666
667
668
25
669
670
Fig. 2. Mean difference between perceived health risk and perceived economic risk. Note: 671
Standardized Mean Difference, Error Bars 95% CI 672
673
674
675
26
A 676
677 678
679
B 680
681
!682
Fig. 3. Density plots for compliance with preventive health behaviors (upper figure A) and 683
support for containment policies (lower figure B). 684
27
685
!686
!!!!!A!687
!688
!!!!!!B!689
690
Fig. 4. Positive interaction between health and economic perceived risks in their association 691
with frequent hand wash (upper figure A) and support for mandatory vaccination 692
(lower figure B). 693
694
695
Tables 696
697
Table 1. Multilevel Regression Modeling: Preventive Health Behaviors 698
699
Hand Washing
Avoid Crowds
Social Isolation
0
1
2
0
1
2
0
1
2
Intercept
.02 (.06)
.02 (.05)
.00 (.03)
-.01 (.06)
.00 (.06)
.01 (.03)
-.04 (.08)
-.08 (.23)
-.04 (.06)
Control: Case-Fatality Rate
-.02 (.04)
-.02 (.04)
.00 (.02)
.01 (.04)
.00 (.04)
-.01 (.02)
.04 (.06)
.11 (.06)
.06 (.04)
Health Risk (HR)
.01 (.02)
.03 (.02)
.00 (.02)
.02 (.01)
.01 (.01)
.01 (.01)
Economic Risk (ER)
.11*** (.01)
.11*** (.01)
.09*** (.01)
.10*** (.01)
.05*** (.01)
.06*** (.01)
HR X ER
.01 (.01)
.02** (.01)
.00 (.01)
.01 (.01)
-.02** (.01)
-.01 (.01)
Health Risk2(HR2)
-.01 (.01)
-.01 (.00)
.00 (.01)
.01 (.00)
-.03*** (.01)
-.02** (.01)
HR2 X ER
.00 (.00)
-.01 (.00)
.01 (.00)
.00 (.00)
.00 (.00)
.01 (.00)
Adjusted ICC
.04
.05
.02
.04
.05
.02
.09
.05
.03
Note: reporting unstandardized coefficients, standard errors in parentheses. *p<.05, ** p<.01, ***p<.001. All predictors are presented in the Methods section and detailed in 700
Table S1. All models controlled for COVID-19 case-fatality rate: total COVID-19 deaths per million/ total COVID-19 cases per million. Model 2 adjusted for individual and 701
country level covariates as follows. Individual level covariates: (i) direct exposure to someone in their personal network (self, family, friends) infected with COVID-19; (ii) 702
perceived knowledge about the COVID-19, (iii) perceived knowledge about the economic consequences of the COVID-19; (iv) the perceived quality of the public messages 703
received, (v) community norms about mitigation measures, and (vi) sociodemographic variables (age, gender, education, employment and financial status). Country-level 704
covariates included (i) total population of the country (in millions), (ii) gross domestic product (GDP) per capita (in current $US), (iii) unemployment rate estimates for 2020 705
(as % of the labor force), (iv) old age dependency ratio (%), (v) Gini Index, (vi) general health expenditure (as %GDP), (viii) private health expenditure (as % health 706
expenditure), (viii) out-of-pocket health payments (as % health expenditure), (ix) number of hospital beds (per 1000 people). 707
708
709
710
711
712
713
714
715
716
717
718
719
720
29
Table 2. Multilevel Regression Modeling: Support for Strict Containment Measures 721
722
Mandatory Vaccination
Mandatory Quarantine
Report Suspected Cases
0
1
2
0
1
2
0
1
2
Intercept
-.09 (.08)
-.05 (.07)
-.04 (.06)
.00 (.07)
.00 (.06)
.01 (.03)
-.04 (.08)
-.08 (.23)
-.04 (.06)
Control: Case-Fatality Rate
.09 (.06)
.04 (.04)
.04 (.04)
.00 (.05)
.00 (.04)
-.01 (.02)
.04 (.06)
.11 (.06)
.06 (.04)
Health Risk (HR)
.05*** (.01)
.06*** (.01)
.00 (.02)
.02 (.01)
.01 (.01)
.01 (.01)
Economic Risk (ER)
.03* (.02)
.04** (.14)
.09*** (.01)
.10*** (.01)
.05*** (.01)
.06*** (.01)
HR X ER
.01 (.01)
.01* (.01)
.00 (.01)
.01 (.01)
-.02** (.01)
-.01 (.01)
Health Risk2(HR2)
.00 (.01)
.00 (.01)
.00 (.01)
.01 (.00)
-.03*** (.01)
-.02** (.01)
HR2 X ER
.00 (.00)
.00 (.00)
.01 (.00)
.00 (.00)
.00 (.00)
.01 (.00)
Adjusted ICC
.04
.08
.05
.04
.05
.02
.09
.05
.03
Note: reporting unstandardized coefficients, standard errors in parentheses. *p<.05, ** p<.01, ***p<.001. All predictors are presented in the Methods section and detailed in 723
Table S1. All models controlled for COVID-19 case-fatality rate: total COVID-19 deaths per million/ total COVID-19 cases per million. Model 2 adjusted for individual and 724
country level covariates as follows. Individual level covariates: (i) direct exposure to someone in their personal network (self, family, friends) infected with COVID-19; (ii) 725
perceived knowledge about the COVID-19, (iii) perceived knowledge about the economic consequences of the COVID-19; (iv) the perceived quality of the public messages 726
received, (v) community norms about mitigation measures, and (vi) sociodemographic variables (age, gender, education, employment and financial status). Country-level 727
covariates included (i) total population of the country (in millions), (ii) gross domestic product (GDP) per capita (in current $US), (iii) unemployment rate estimates for 2020 728
(as % of the labor force), (iv) old age dependency ratio (%), (v) Gini Index, (vi) general health expenditure (as %GDP), (viii) private health expenditure (as % health 729
expenditure), (viii) out-of-pocket health payments (as % health expenditure), (ix) number of hospital beds (per 1000 people). 730
731
732
733
734
735
736
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.