Content uploaded by Jocelyn Raude
Author content
All content in this area was uploaded by Jocelyn Raude on Mar 10, 2023
Content may be subject to copyright.
1
Health Protective Behaviours during the COVID-19 pandemic:
Risk Adaptation or Habituation?
Dylan Martin-Lapoirie1*, Kathleen McColl2, Karine Gallopel-Morvan2, Pierre
Arwidson3, Jocelyn Raude2
1. Centre d’Economie de la Sorbonne, CNRS, Université Paris 1 Panthéon-Sorbonne,
Paris, France.
2. EHESP, CNRS, Inserm, Arènes - UMR 6051, RSMS (Recherche sur les Services et
Management en Santé) - U 1309, Université de Rennes, Rennes, France.
3. Direction de la prévention de la santé, Santé Publique France, Saint-Maurice, France.
*Corresponding author at: Centre d’Economie de la Sorbonne, 106-112 boulevard de l’Hôpital,
75647 Paris Cedex 13. dylan.martin-lapoirie@univ-paris1.fr.
Highlights
- Health protective behaviours considerably varied from March 2020 to September 2021.
- Engagement in protection was positively associated with the COVID-19 death incidence.
- Time exerted however a negative effect on the adoption of protective behaviours.
- Risk adaptation and habituation explain the response to the health threat over time.
Acknowledgements
The authors thank the participants of workshop “Models, Human Behaviour and
Infectious Diseases” organised by the Pasteur Institute and of the First Euro Public Health +
Consortium Research Seminar organised by the University College Dublin for their helpful
comments. They are also grateful to the COVIPREV group (Enguerrand du Roscoät, Jean-
Michel Lecrique, Linda Lasbeur, Christophe Léon, Pierre Arwidson, Isabelle Bonmarin, and
Oriane Nassany) from the Department of Health Promotion and Prevention (Santé Publique
France) for their valuable support.
Funding
This work was supported by the European Union’s Horizon 2020 research and
innovation program “PERISCOPE: Pan European Response to the ImpactS of COvid-19 and
future Pandemics and Epidemics”, under the grant agreement No. 101016233, H2020-SC1-
PHE_CORONAVIRUS-2020-2-RTD.
2
Health Protective Behaviours during the COVID-19 pandemic:
Risk Adaptation or Habituation?
Abstract
Many epidemiological works show that human behaviours play a fundamental role in
the spread of infectious diseases. However, we still do not know much about how people modify
their Health Protective Behaviours (HPB), such as hygiene or social distancing measures, over
time in response to the health threat during an epidemic. In this study, we examined the role of
the epidemiological context in engagement in HPB through two possible mechanisms
highlighted by research in the field of decision-making under risk: risk adaptation and risk
habituation. These two different mechanisms were assumed to explain to a large extent the
temporal variations in the public’s responsiveness to the health threat during the COVID-19
pandemic. To test them, we used self-reported data collected through a series of 25 cross-
sectional surveys conducted in France among representative samples of the adult population,
from March 2020 to September 2021 (N=50,019). Interestingly, we found that both
mechanisms accounted relatively well for the temporal variation in the adoption of HPB, which
is remarkable given their different assumptions about the underlying social cognitive processes
involved in response to a health threat. These results suggest that strengthening the incentives
for people to adopt HPB is crucial in critical settings, and that public health interventions should
be designed to counter risk habituation effects over time.
Keywords: Risk adaptation, risk habituation, prevalence-elasticity, health protective
behaviours, COVID-19
3
1. Introduction
The COVID-19 pandemic highlighted the need to better understand the complex
interplay between human behaviours and the epidemiological environment in which people
live. However, the role of such context in the decisions that people make to prevent or control
the risk of infection during epidemics or pandemics remains largely unknown. Moreover, while
epidemiologists and biomodelers have developed increasingly sophisticated epidemiological
models to predict the spread of infectious diseases, these models are still insufficiently based
on individual human behaviours (Lorig et al., 2021; Verelst et al., 2016). Instead, the individual
decisions to adopt Health Protective Behaviours (HPB) for preventing or reducing health risks
are simply ignored or represented by an exogenous proportion of the population which is
considered as protected, independently from the evolution of the infection risk over time.
Although some epidemiological models attempted to endogenize the protection decision, like
economic-epidemiological models, the way to model the relationship between the adoption of
HPB and the epidemiological context remains a major issue. Therefore, it seems important to
investigate here the relationship between the adoption of HPB and contextual variables through
two different social cognitive mechanisms: risk adaptation and risk habituation.
1.1. Social and epidemiological context
Our paper examines individual decisions to engage in HPB during the COVID-19
pandemic in France from March 2020 to September 2021. Thus, the studied period covers four
epidemic waves. During the first wave in Spring 2020, people did not have prior experience
with the virus and they were not used to being protected. In addition, there were not enough
facemasks to protect the entire population. The French government implemented a strict
lockdown from 17 March to 10 May 2020 to reduce the spread of the epidemic. After this
lockdown, restrictions were progressively lifted. A second lockdown was imposed during the
4
second wave, from 30 October to 14 December 2020, in which people had to stay at home but
schools remained open. After the second wave, the incidence level was still high, which
motivated the French government to maintain restrictions through many curfews from 15
December 2020 to 19 June 2021. During this period, to control the third epidemic wave, a third
lockdown was implemented from 3 April to 2 May 2021, in which travel was allowed within a
10-km radius of one’s home. From June to September 2021, restrictions were lifted, although
France was experiencing the fourth epidemic wave. At the same time, from 26 December 2020,
people were allowed to be vaccinated according to their age. Since 15 June 2021, vaccination
has been available for all French adults. Furthermore, from 1 July 2021, access to leisure
activities and long distance public transport was subject to holding a health pass obtained
following vaccination or a negative COVID test.
In France, management of the pandemic crisis was essentially based on hospitalisation
rates, and the evolution of the reproduction number of the disease. Other countries, like the
United Kingdom, based their policies on the dynamics of human behaviour to avoid
demotivating people, but were accused of mismanaging the crisis (Abbasi, 2020). It is clear that
the responsiveness of human behaviour to the epidemiological context is a crucial element in
predicting the spread of the pandemic and in reducing the health threat. An epidemiological
context can be defined by a large range of variables related to the strength and severity of the
infection, such as the incidence of cases and deaths in the region, the contagiousness of the
disease, the proportion of immunised or naive patients, the hospitalisation or death rate, and so
on. In this research, we focused on the actual risk of being infected and the time of exposure to
the threat, as these two contextual variables are central to specific social cognitive mechanisms
- risk adaptation and risk habituation - which have been proposed in the literature to explain
temporal variation in an individual’s engagement in HPB (Johnson and Mayorga, 2020; Raude
et al., 2019; Loewenstein and Mather, 1990).
5
1.2. Risk adaptation
During the last decades, a variety of definitions of the term ‘adaptation’ have been used
in the social and behavioural sciences. Initially developed by biologists, this term is now widely
employed in the fields of psychology, sociology, anthropology, and even geography. Here, we
will use the American Psychological Association’s definition, according to which adaptation
refers to “modification to suit different or changing circumstances. In this sense, the term often
refers to behaviour that enables an individual to adjust to the environment effectively” (APA,
2022)
1
. Consistent with this definition, the term generally refers in public health to a set of
measures that people take to reduce or eliminate a given health risk. More precisely, the concept
of risk adaptation embraces the idea that engagement in preventive behaviours depends on the
risk magnitude. Applied to infectious diseases, adaptation implies that individuals engage in
HPB depending on observable changes within the epidemiological context. If individuals have
an accurate perception of risk, then they are likely to be more cautious when the risk of being
infected increases (and conversely). In health economics, the adaptation to infectious risk
through HPB is captured and measured through a concept called “prevalence-elasticity”. This
concept was developed by Geoffard and Philipson (1996) to explain the temporal and spatial
variations in the adoption of HPB when faced with the risk of HIV infection. In a SI
(Susceptible-Infected) model representing the HIV epidemic, they analysed the incentives of
an individual to adopt HPB to avoid catching the disease. An individual is assumed to compare
the cost of protection (the loss in utility from being protected rather than exposed) with the
expected cost of risking infection (the expected future utility of being infected rather than
susceptible). Geoffard and Philipson demonstrated that the individual protective decision
related to AIDS/HIV depends on prevalence level. If the current level is higher than a
1
Available online: https://dictionary.apa.org/adaptation (accessed September 22, 2022).
6
prevalence threshold, then the individual prefers to be protected in the current period. In the
reverse case, he chooses to be exposed. This pioneering work pushed epidemiological models
to take account of risk adaptation, leading to “economic-epidemiological models” (Chan et al.,
2015; Reluga, 2010), some of which are related to the COVID-19 epidemic (Farboodi et al.,
2021; Makris, 2021).
2
It is now natural to model the probability of engaging in HPB as a positive
function of the prevalence level.
In their seminal study, using longitudinal data collected from surveys conducted in San
Francisco from 1983 to 1992, Geoffard and Philipson showed that men reduce their exposure
to the risk of being infected with HIV as the prevalence rate among the respondents increases.
In other words, they adopt prevalence-elastic HPB. Moreover, many other empirical studies
have since found such a behaviour towards the risk of HIV infection (Guillon and Thuilliez,
2015). Among these studies, the decision to engage in HPB depends on either the prevalence
level (Oster, 2012; Young, 2007) or the incidence level (Godlonton and Thornton, 2013;
Boucekkine et al., 2009). The existence of prevalence-elastic HPB was also shown for other
diseases. There is strong evidence that the use of mosquito nets is positively correlated with
Malaria prevalence in tropical countries (Picone et al., 2017; Seban et al., 2013). In other studies
investigating diseases for which vaccination exists (and immunisation is achieved), such as
measles, it was found that parents were more likely to vaccinate their children as prevalence
increases (Philipson, 1996). In the same way, more individuals were found to be vaccinated as
the annual duration of the influenza epidemic increased (Mullahy, 1999). Regarding influenza
vaccination, individual decisions were positively influenced by the influenza prevalence, not
only for the current year but also that of the previous year (Li et al., 2004).
2
See McAdams (2021) for a literature review on economic-epidemiological models.
7
To the best of our knowledge, Battiston and Gamba (2021) are the only researchers to
have found a positive relationship between HPB and COVID-19 prevalence. More precisely,
using temporal data collected from March to April 2020 during the strict lockdown in Italy,
they showed a negative correlation between the initial COVID-19 prevalence level and the basic
reproduction number, meaning that individuals adopted HPB in response to the high initial
infection risk. Other studies focusing on the relationship between HPB and perceived risk
offered mixed evidence. Indeed, some studies highlighted a positive and significant relationship
(Wambua et al., 2022; Qin et al., 2021; Schneider et al., 2021) whereas others showed no
significant effect of perceived risk on the adoption of HPB (Fullerton et al., 2021; Zickfeld et
al., 2020). In the current study, we will test whether the adoption of HPB changes over time
with the actual infection risk, which was captured through the death incidence level.
1.3. Risk habituation
Contrary to that of risk adaptation, the concept of risk habituation has not been
extensively investigated in the field of infectious diseases. In psychology, the term
“habituation” refers generally to the process of growing accustomed to a threatening situation
or stimulus. More precisely, it can be defined as “the diminished effectiveness of a stimulus in
eliciting a response, following repeated exposure to the stimulus” (APA, 2022)
3
. In public
health, risk habituation refers to the mechanism through which “people tend to underestimate
progressively or neglect risks as a health threat becomes increasingly familiar” (Raude et al.,
2019). The habituation effect thus provides an explanation for the reduced effect (or the absence
of effect) of infection risk on the adoption of HPB over time. Such an effect was suggested to
explain the lack of correlation between perceived risk and protective measures in the literature.
Based on longitudinal data, Lima (2004) showed that people living closer to an incinerator have
3
Available online: https://dictionary.apa.org/habituation (accessed September 22, 2022).
8
a lower estimation of risk and adopt less extreme attitudes, which suggests habituation effects
in terms of attitudes and risk perception. Previously, in the field of technological risk, such
effects were also found on the basis of cross-sectional data. For instance, van der Pligt et al.
(1986) highlighted a familiarity effect for residents living near a nuclear power station. Indeed,
they were more optimistic than people living further from the station.
In the current literature, a habituation effect has only been highlighted to explain
changes in perceptions of the risk of being infected by mosquito-borne diseases. Using a
longitudinal study of a large epidemic of a mosquito-borne disease, Raude et al. (2019) found
that the adoption of preventive behaviours varied positively with disease prevalence, validating
the hypothesis of risk adaptation. However, they also showed that participants’ perceived
personal risk of infection decreased over time. This latter phenomenon could be the result of a
risk habituation effect. Overall, the hypotheses of risk adaptation and risk habituation are
usually presented as competing to explain engagement in HPB, but they are not necessarily
incompatible. Indeed, individuals can adopt HPB as a response to an increase in disease
incidence and this effect can also decrease over time.
It should be noted that in the context of the COVID-19 pandemic, similar concepts to
that of risk habituation - such as “behavioural fatigue” or “pandemic fatigue” - have been
increasingly used and discussed. In particular, pandemic fatigue was cited for a long time by
some countries, like the United Kingdom, as the reason that people were unable to comply with
health recommendations such as self-isolation (Reicher and Drury, 2021; Abbasi, 2020). They
hypothesised that pandemic fatigue would have thus led the public authorities to delay the
implementation of lockdown policies. Pandemic fatigue has been defined by the World Health
Organization (2020) as “demotivation to follow recommended protective behaviours, emerging
gradually over time and affected by a number of emotions, experiences and perceptions”.
Nonetheless, this definition is not canonical and fatigue could be based on many explanations
9
(Michie et al., 2020). Moreover, there is still a debate as to whether the phenomenon really
exists. Indeed, some authors consider that pandemic fatigue could be simply a naive construct
or a policy contrivance (Harvey, 2020). This conclusion is supported by the engagement in
hygiene measures and social distancing that has remained high since the beginning of the
COVID-19 epidemic in some cross-sectional studies (Dixon et al., 2022; Smith et al., 2022).
However, others found a decreasing trend of engagement in social distancing, which could be
evidence in favour of such a phenomenon (Schaner et al., 2022; Franzen and Wöhner, 2021).
In particular, based on longitudinal data collected in 14 countries from March to December
2020, Petherick et al. (2021) showed that mask wearing was associated with a linear increase
in adherence to this behaviour, but physical distancing was associated with a decline over time,
which was increasingly low, with small rebounds in later months.
4
In our study, rather than using the concept of fatigue, which is still subject to many
criticism, we thought it more relevant to investigate the mechanism of risk habituation. The
difference between these concepts is that risk habituation clearly rests on a precise definition,
as the term refers to an increasingly smaller response to a repeated threatening stimulus over
time, whereas the demotivation concept underlying the notion of pandemic fatigue appears to
be much more ambiguous. Although there exists evidence to support a temporally decreasing
trend in the adoption of preventive behaviours, the “fatigue” concept is not well grounded
theoretically, which prevents correctly capturing and measuring it. That is why we tested
instead the risk habituation hypothesis, which seems to provide a more credible and refutable
explanation of the demotivation to act phenomenon which was documented during the
pandemic. For this purpose, we will examine whether there exists a correlation between time
of exposure and the adoption of preventive behaviours, controlling for actual infection risk.
4
For mask wearing, it seems that the authors did not take account of the shortage in masks that occurred at the
beginning of the epidemic, which could naturally explain the linear increase over time that they observed.
10
Therefore, the main objective of this paper is to test the validity of the hypotheses of
risk adaptation and risk habituation in the context of the COVID-19 pandemic, by determining
to what extent an individual’s engagement in HPB in response to a health threat is driven by
these two important contextual factors.
2. Materials and methods
2.1. Participants and procedure
Our study of the attitudes towards the COVID-19 infection risk is based on cross-
sectional data, collected through 25 online surveys conducted from March 2020 to September
2021 in France. Samples, which comprise N=50,019 responses, were representative of adults
residing in France. Indeed, a stratified sampling method was adopted to recruit participants
based on sex, age, occupation, community size and region recorded during the 2016 national
general census conducted by the National Institute of Statistics and Economic Studies
(INSEE).
5
Ethical approval was granted by the University Hospital Institute “Mediterranee
Infection” Ethics Committee Marseille, France and the EHESP School of Public Health Office
for Personal Data Protection.
The period covered a strict lockdown, two mild lockdowns and many curfews. Table 1
provides the number of observations concerned by national restrictions in France since the
beginning of the epidemic.
[Insert Table 1]
2.2. Measures
5
Surveys were conducted by the BVA research institute (https://www.bvagroup.com/en/about-us/).
11
In this research, the relationship between engagement in HPB and death incidence level
was analysed. The latter corresponds to the number of deaths per day (7-day moving average).
We captured the COVID-19 infection risk by the death incidence level rather than the
prevalence level or the incidence level, which expresses numbers of COVID-19 cases, for two
reasons. Firstly, these variables were not reliable due to the underdetection of symptomatic
COVID-19 cases (Pullano et al., 2021; Shaman, 2021). Secondly, it is likely that individuals
were more affected by the number of serious cases rather than the number of symptomatic and
non-symptomatic cases (Wambua et al., 2022; Qin et al., 2021). The death incidence level is a
good proxy for the number of serious cases.
We measured engagement in HPB to reduce the risk of COVID-19 infection by asking
participants to report whether in the last days they had adopted the following behaviours:
“Avoid close contacts with other people” (social gatherings with people who do not live at
home), “Do not shake hands”, and “Wash hands often”. The first behaviour corresponds to a
social distancing measure while the others are related to hygiene measures. Participants
answered “Yes, systematically”, “Yes, often”, “Yes, sometimes”, or “No, never”. As explained
in a preliminary study (Raude et al., 2020), the responses revealed a ceiling effect in favour of
the upper limit of the scale. That is why we created a dichotomous variable for each behaviour,
coded as 1 for the “high compliance” response (“Yes, systematically”), and coded as 0 for the
other options merged into a “lower compliance” category. Percentages of participants who
reported engaging in HPB over time and death incidence level are displayed in Figure 1. For
each percentage, we calculated 95% confidence intervals. The mask wearing has been excluded
from our analysis because of the shortage at the beginning of the epidemic in France.
[Insert Figure 1]
12
As it was assumed that legal restrictions implemented in France since March 2020 may
moderate the effect of infection risk on social distancing behaviours, we included in our analysis
an index capturing the variations in restrictions over time. More precisely, the stringency index,
developed by the Oxford COVID-19 Government Response Tracker (OxCGRT), indicates the
strictness of ‘lockdown style’ policies, the purpose of which is primarily to restrict people’s
behaviour (Hale et al., 2022). It is an average score that falls between 0 and 1, comprising
containment and closure policy indicators and a public information campaign indicator.
In each of our surveys, we also collected sociocultural and demographic variables. A
wide range of items was included in the questionnaire, such as sex, age, education level (higher
or lower than the High School Diploma) and socioeconomic status (high: self-employed and
entrepreneurs, managerial and professional occupations, intermediate professions; low:
employees, workers; inactive: students, others). Moreover, participants were asked how they
perceived their financial situation (bad: “I cannot get there without debts”, “I hardly get there”;
good: “it is correct”, “it is okay”, “I am comfortable”). They also had to report whether they
had personally suffered from signs or symptoms indicating a possible COVID-19 infection
since February 2020 (“yes” or “no”). Finally, we asked participants to indicate the number of
rooms in their housing and the living area to determine whether their housing was overcrowded,
which might influence the number of social gatherings. Descriptive statistics of these
sociocultural and demographic variables are provided in Table 2.
[Insert Table 2]
2.3. Statistical analysis
Our study focuses on the relationship between the engagement in HPB we measured and
the epidemiological context represented by the infection risk (risk adaptation) and time (risk
13
habituation). As a first step, we used comparison tests to investigate whether the three
preventive behaviours were associated with these two variables. We proceeded by t-tests to
compare the mean death level according to the adoption of each HPB. Results are indicated in
Table 3. Furthermore, we compared the proportions of individuals engaging in each HPB over
time by Pearson’s chi-square tests.
As a second step, we focused on the social distancing behaviour, which is avoiding close
contacts with other people, because it exhibits more variance over time and varies more with
the death incidence level than the two hygiene behaviours, in particular handwashing. Table 4
provides the results of Pearson’s chi-square tests and t-test to assess the association between
each sociocultural and demographic variable and the engagement in social distancing. To
expand upon the determinants of the probability of adopting social distancing, as well as to
control for the influence of sociocultural and sociodemographic variables, we performed a
multivariate analysis. Table 5 reports the results of Probit regressions, in which the probability
of avoiding close contacts with other people was the dependent variable. In Model (1), in order
to estimate whether social distancing was incidence-elastic, we estimated a baseline
specification in which we included the death incidence level, the stringency index, and time
expressed by the number of days since 16 March 2020 (announcement by the French President
of the strict lockdown). In Model (2), we added sociocultural and demographic variables to
highlight potential differences in the degree of engagement in social distancing among
participants.
6
In Model (3), we considered six interaction terms between the death incidence
level and sociocultural and demographic variables to study whether the effect of these latter
variables depended on infection risk. Finally, in Model (4), we controlled for context by
6
These variables are: Sex equal to 1 if the participant if female; Age expressed in years; History of COVID-like
symptoms equal to 1 if the participant reports an experience of COVID-19 infection; Completed secondary school
equal to 1 if the participants holds the High School Diploma; Low equal to 1 if the participant’s socioeconomic
status is low; Inactive equal to 1 if the participant is inactive; Financial situation equal to 1 if the participant’s
financial situation is bad; and Housing overcrowding equal to 1 if the participant’s housing is overcrowded.
14
including survey fixed effects. They capture the effects of time-specific variables - the death
incidence level, the stringency index and naturally the survey day - which cannot be estimated
with this specification.
As a last step, we measured the size of the relationship between the infection risk and
engagement in social distancing, controlling for sociocultural and demographic variables. Table
6 reports the Probit COVID-19 incidence elasticities associated with Model (2) in Table 5. All
data were treated and analysed using STATA 16.
3. Results
As shown in Figure 1, the engagement in HPB varies depending on whether the
behaviour is related to hygiene or social distancing. From March 2020 to September 2021, the
percentages of participants reporting systematically not shaking hands with other people and
washing hands often varied from a minimum of 58.8% (31 August-7 September 2021) to a
maximum of 92.8% (14-16 April 2020) and from a minimum of 58.6% (31 August-7 September
2021) to maximum of 76.5% (30 March-1 April 2020) respectively. By contrast, the percentage
of participants reporting systematically avoiding close contacts with other people varied from
a minimum of 32.8% (31 August-7 September 2021) to a maximum of 90.3% (30 March-1
April 2020). Thus, not shaking hands and handwashing, which correspond to hygiene measures,
exhibit a lower variance than the reduction of contacts with other people, which relates to social
distancing. Overall, contrary to hygiene measures, we do not observe in France that engagement
in physical distancing has remained high over time.
We examined the relationship between the adoption of HPB and the two aforementioned
contextual variables. We can identify in Figure 1 the different epidemic situations since March
2020. Our data covers the four first epidemic waves in France (2020 spring, 2020 autumn, 2021
winter and spring, 2021 summer). The death incidence level corresponding to different survey
15
times varied from a minimum of 12 (20-22 July 2020) to a maximum of 772 (14-16 April 2020)
deaths per day. In particular, it is interesting to note that an increase (reduction) in the death
incidence level corresponds to an increase (reduction) in the engagement in social distancing,
except for 5 surveys. To better visualise the relationship between the adoption of HPB and the
infection risk, we represented the mean adoption level in function of the number of deaths per
day in Figure 2. Each graph suggests an increasing relationship between both variables. We
confirmed the existence of such an association by comparing the mean death incidence level
according to engagement in HPB. As shown in Table 3, there is a significant positive
association between the death incidence level and the engagement in each HPB.
[Insert Figure 2]
[Insert Table 3]
Figure 1 also highlights the trends of adoption of a variety of preventive behaviours over
time. Indeed, between the first and the last survey wave, the percentages of participants
reporting systematically not shaking hands with other people and washing hands often
decreased by 33.59 points and 17.97 points respectively. At the same time, the percentages of
participants adopting social distancing decreased by 57.53 points. To test whether these
differences reflect a negative temporal trend, we looked at the differences in the proportions of
people engaging in preventive behaviours. The proportions are significantly different for each
behaviour (“Avoid close contacts with other people”: 𝜒²(24)=7,280.9, p<0.01; “Do not shake
hands”: 𝜒²(24)=2,641.2, p<0.01; “Wash hands often”: 𝜒²(24)=393.24, p<0.01), which suggests
a negative trend of time on the engagement in HPB.
As social distancing - i.e. avoiding close contacts with other people - is the HPB that
varies the most with the infection risk and over time, our multivariate analysis of the
relationship between the engagement in HPB and the epidemiological context focused on this
16
behaviour. As reported in Table 5, death incidence level is positively correlated with the
decision to engage in social distancing. This relationship is robust to the effects of legal
restrictions and the effects of sociocultural and demographic variables. It seems that participants
reacted to the infection risk by avoiding social gatherings, that is social distancing seems to be
an incidence-elastic behaviour.
[Insert Table 4]
[Insert Table 5]
To measure precisely how the engagement in social distancing varies with the death
incidence level, we also examined the average partial effect and the elasticities of engagement
with respect to the death incidence level derived from Model (2). We see in Table 6 that 100
additional deaths per day increases on average the probability of engagement in social
distancing by 2.4 percentage points. Furthermore, the elasticity at the average and the average
of elasticity are smaller than 1, meaning that the engagement in social distancing is weakly
elastic to the infection risk.
[Insert Table 6]
We also investigated the effect of time on the adoption of social distancing. Our
multivariate analysis in Table 5 confirms the negative effect of time on the probability of
adopting social distancing. This negative effect is robust, regardless of legal restrictions and
sociocultural and demographic variables. The mechanism of risk habituation states that people
are increasingly less sensitive to the stimulus with time. Here the stimulus is the infection risk,
which is captured by the death incidence level. That is why we analysed whether the effect of
time depends on the death incidence level. In this way, Model (3) included an interaction term
17
between time and death incidence level. This interaction term was negative and significant,
meaning that the negative effect of time slightly increases with the infection risk.
Furthermore, our results concern the effects of legal restrictions and sociocultural and
demographic variables on the probability of adopting social distancing. As indicated in Model
(2) in Table 5, the probability of adopting social distancing increases with the stringency index,
suggesting that people tended to comply with restrictions imposed by the French government.
Regarding sociocultural and demographic variables, Table 4 displays significant differences in
the engagement in social distancing. Model (2) in Table 5 allows us to confirm that being a
woman, being elderly, having a low socioeconomic status or being inactive, and living in
overcrowded housing were found to be associated with a higher degree of engagement in social
distancing. Moreover, contrary to the univariate analysis in Table 4, feeling poor was positively
correlated with the adoption of social distancing. By contrast, no significant effect on the
engagement in social distancing was found for having been infected by COVID-19 infection or
holding a High School Diploma.
Because individuals could react differently to the infection risk depending on their sex
or their socioeconomic status for example, we also examined the interactions between the death
incidence level and sociocultural and demographic variables in Model (3). First, it appears that
the interaction term between sex and incidence was positive and significant. Women were
typically more cautious than men in epidemic times, and this difference in reaction to the health
threat is increasing with the death incidence level. More counter-intuitively, Model (2) showed
that the wealthier participants were, the less cautious they were. In Model (3), the interaction
terms between socioeconomic status and the engagement in social distancing were negative and
significant, revealing that this relationship depended on the death incidence level. Thus, we
observed a convergence mechanism between socioeconomic status, in terms of social
distancing, as the death incidence level increased. In other words, wealthier people tended to
18
be more negligent than poorer people when the number of deaths per day was low, and they
became more cautious as the number of deaths per day increased.
4. Discussion
On the basis of the existing literature, we identified two possible mechanisms explaining
the adoption of HPB during an epidemic and tested them against the behaviours reported by the
adult population in a series of 25 cross-sectional surveys conducted during the two first years
of the COVID-19 pandemic in France. These mechanisms, that are risk adaptation and risk
habituation, may be considered competitive, or even mutually exclusive to explain human
behaviour in response to an emerging health threat. In other words, either the engagement in
HPB depends on risk magnitude, or it depends on time. To the best of our knowledge, our study
is the first to reveal the simultaneous existence of risk adaptation and habituation in the dynamic
of HPB during an epidemic. On the one hand, we showed that the engagement in social
distancing was a positive function of the COVID-19 infection risk, confirming the mechanism
of risk adaptation. Social distancing is an incidence-elastic preventive behaviour during an
epidemic but the elasticity is smaller than 1, meaning that the engagement in social distancing
is only slightly elastic to the death incidence level. Previous literature showed such incidence-
or prevalence-elastic behaviours in epidemics such as HIV, HPV, malaria, measles or influenza,
but not during the COVID-19 epidemic. This exception may be due to the implementation
around the world of exceptionally stringent policies affecting social behaviours in the attempt
to control the spread of the disease (Al-Zubaidy et al., 2021). On the other hand, we found a
negative effect of time on the engagement in social distancing. Moreover, this negative effect
seems to increase with the infection risk, suggesting that people get progressively accustomed
to the risk over time. This is therefore difficult to maintain barrier measures at the same high
level in the long term. Regarding infectious diseases, risk habituation is still an overlooked
19
process in the literature and we believe that it may represent a promising concept for future
research.
Our study also contributes to the behavioural literature on epidemics by exploring the
influence of sociocultural and demographic variables on the decision to engage in preventive
behaviours. During the COVID-19 epidemic, as in previous epidemics (Bish and Michie, 2010;
Seale et al., 2020), numerous studies showed that being a woman, being elderly or having a
high level of formal education was positively associated with a higher probability of adopting
preventive behaviours (Papageorge et al., 2021; Smith et al., 2021; Wright and Fancourt, 2021;
Faasse and Newby, 2020). Consistent with this literature, we found that being a woman and
being elderly was indeed positively correlated with the probability of engaging in social
distancing. Moreover, women’s HPB positively depended on the death incidence level,
reflecting the fact that women are commonly more risk-averse than men (Croson and Gneezy,
2009). By contrast, having some college education did not have any significant effect on this
behaviour. Surprisingly, we also found that a high socioeconomic status or perceived wealth
was associated with a less protective behaviour. This relationship also depended on the death
incidence level. The negligent behaviour associated with a high economic status reduced the
infection risk. In other words, while more socio-economically advantaged people are less likely
to engage in social distancing than those who are more disadvantaged, a convergence trend was
observed among socioeconomic status when the infection risk increased.
Our study was motivated by the need to better understand how to model the individual
interactions between HPB and the epidemiological context. Indeed, most epidemiological
modellers still assume that the decision to engage in protection is exogenous, i.e., independent
of the epidemiological context, or follows a fixed behavioural pattern (Lorig et al., 2021;
Verelst et al., 2016). Based on our results, we recommend that models account for the adoption
of HPB as a positive function of the infection risk and a negative function of time. To date, only
20
a few models, such as the economic-epidemiological models, have integrated infection risk and
time as parameters. In these models, the expected cost of being infected increases with the
infection risk, which makes the individual more likely to engage in protection, and future
benefits and costs of being exposed or protected are discounted at a rate that represents the
preference for the present. However, it should be noted that economic modelling is not the only
one to include these variables, since agent-based models (Lorig et al., 2021), and more recently
transmission models (Weitz et al., 2020), also integrate the infection risk and time in the agents’
decision rule. Our results confirm the importance of developing these various types of
modelling in future research (Bedson et al., 2021; Ferguson, 2007).
Our results may also help inform policy-makers on how to control the spread of future
epidemics. On the one hand, risk adaptation means that people tend to adjust their HPB as a
function of the spread of the disease. As shown in our study, people seem to use external cues
to action based on epidemiological data to deal with the threat during an epidemic. In this
context, when the epidemic situation is becoming critical, it could be relevant for policy-makers
to support external cues, by implementing non-pharmaceutical interventions such as nudges
which promote social distancing and hygiene measures. For instance, messages or posters were
widely used during the pandemic and proved effective (Dai et al., 2021 ; Lunn et al., 2020). We
believe that their use should be reinforced during the peak of the epidemic. On the other hand,
while risk adaptation is a personal mechanism that is complementary to public action, risk
habituation effects threaten the control of the spread of the epidemic, as people are less worried
about contracting the disease and decreasingly responsive to the infection risk over time. In this
matter, the challenge for policy-makers could be to deter habituation behaviours; here again it
may be done by using behavioural nudges, i.e. messages disseminated over time.
Numerous studies focus on the effects of nudges on protective behaviours. In a
systematic review, Epton et al. (2022) were interested in the effectiveness of interventions and
21
behaviour change techniques which promote adherence to physical distancing. They showed
that future punishment, with government fines, are ineffective. By contrast, providing feedback
on behaviour, through proximity buzzers, providing information about health consequences,
with loss-framed messages on posters, and restructuring the physical environment, via
directional systems, are effective techniques to promote physical distancing. Other techniques
are still little tested in the existing studies and deserve more attention. Among them, social
comparison and the provision of information about others’ approval are techniques already used
to nudge desirable behaviours in other contexts (Allcot and Kessler, 2019; Bicchieri and Xiao,
2009). There is some evidence that such nudges influenced engagement in social distancing
during the COVID-19 pandemic (Martínez et al., 2021). Future experimental studies are
welcomed to analyse how these nudges, based on social norms, might amplify the mechanism
of risk adaptation whilst mitigating that of risk habituation.
4.1. Limitations
Although our study identifies significant effects of death incidence level and time on the
adoption of social distancing, important mechanisms other than risk adaptation and risk
habituation could contribute to explain these effects. Moreover, risk adaptation is captured in
our analysis by the positive and significant effect of the death incidence level on the probability
of engaging in social distancing. We assumed that this effect was caused by the fact that
infection risk motivates people to take protective measures against COVID-19. Nonetheless,
other factors could explain the variation of preventive behaviour with the infection risk. The
main one is the above-mentioned constraint derived from social norms during an epidemic.
Indeed, regarding the COVID-19 epidemic, perceived social norms were found to drive
intentions (Scholz and Freund, 2021) and engagement in preventive behaviours (Hensel et al.,
2022). As the death incidence level increases, the weight of social norms could encourage
22
people to engage in social distancing due to social pressure, but not due to fear of being infected.
Other factors motivating an individual’s responsiveness to the infection risk could be the desire
to protect others or the desire to avoid future restrictions imposed by the government. In the
absence of measures of these variables, it is not possible to favour one explanation over another.
In the same vein, our results are suggestive of a risk habituation effect, but we were not
able to prove directly that people became accustomed to the infection risk over time. Indeed,
we found a negative and significant effect of time on the probability of adopting social
distancing. However, this negative effect could be due to a decrease in risk perception over
time. Moreover, contrary to the hypothesis of risk adaptation, testing the hypothesis of risk
habituation requires collecting longitudinal data among individuals. The problem with repeated
cross-sectional data is that they do not capture how the preferences of individuals change over
time. Our study compares the responsiveness to the infection risk of many individuals at
different periods, but risk habituation is an individual mechanism that requires comparing the
responsiveness of the same individual at different periods. In the literature, only a few studies
of the COVID-19 epidemic use longitudinal data (Schaner et al., 2022; Petherick et al, 2021;
Qin et al., 2021) and these studies generally cover only the first year of the pandemic, which
prevents studying habituation to the infection risk over a sufficiently long period.
5. Conclusion
In this study, we investigated the relationship between the engagement in HPB and the
epidemiological context. The objective was to determine whether the mechanisms of risk
adaptation and risk habituation contributed to explaining engagement in HPB during the
COVID-19 epidemic. Using 25 cross-sectional surveys conducted in France from March 2020
to September 2021, we found the infection risk was positively associated with the probability
of engaging in HPB, consistent with a mechanism of risk adaptation. The engagement in HPB
23
was also found to decrease over time, which suggests a phenomenon of risk habituation. These
results should encourage epidemiological models to consider the decision to engage in HPB as
a positive function of the infection risk and a negative function of time. The existence of risk
habituation should also be subject to special attention by policy-makers in the control of the
spread of the epidemic. Finally, even though this study is the first to simultaneously highlight
the existence of risk adaptation and risk habituation in an individual's HPB, the identification
of these two mechanisms could be improved. In particular, we believe that future studies should
be based on longitudinal data covering many epidemic stages and should control for social and
cognitive variables that interfere with the decision to engage in HPB.
References
Abbasi, K., 2020. Behavioural fatigue: a flawed idea central to a flawed pandemic response.
BMJ 370:m3093. https://doi.org/10.1136/bmj.m3093.
Al-Zubaidy, N., Crespo, R., Jones, S., Drikvandi, R., Gould, L., Leis, M., et al., 2021. Using
big data analytics to explore the relationship between government stringency and preventative
social behaviour during the COVID-19 pandemic in the United Kingdom.
medRxiv:2021.07.09.21260246. https://doi.org/10.1101/2021.07.09.21260246.
Allcott, H., Kessler, J.B., 2019. The welfare effects of nudges: a case study of energy use social
comparisons. American Economic Journal: Applied Economics 11(1), 236-276.
https://doi.org/10.1257/app.20170328.
Battiston, P., Gamba, S., 2021. COVID-19: R0 is lower where outbreak is larger. Health Policy
125(2), 141-147. https://doi.org/10.1016/j.healthpol.2020.10.017.
24
Bedson, J., Skrip, L.A., Pedi, D., Abramowitz, S., Carter, S., Jalloh, M.F., et al., 2021. A review
and agenda for integrated disease models including social and behavioural factors. Nature
Human Behaviour 5, 834-846. https://doi.org/10.1038/s41562-021-01136-2.
Bicchieri, C., Xiao, E., 2009. Do the right thing: but only if others do so. Journal of Behavioral
Decision Making 22(2), 191-208. https://doi.org/10.1002/bdm.621.
Bish, A., Michie, S., 2010. Demographic and attitudinal determinants of protective behaviours
during a pandemic: a review. British Journal of Health Psychology 15(4), 797-824.
https://doi.org/10.1348/135910710x485826.
Boucekkine, R., Desbordes, R., Latzer, H., 2009. How do epidemics induce behavioral
changes? Journal of Economic Growth 14, 233-264. https://doi.org/10.1007/s10887-009-9042-
1.
Chan, T.Y., Hamilton, B.H., Papageorge, N.W., 2015. Health, risky behaviour and the value of
medical innovation for infectious disease. Review of Economic Studies 83(4), 1465-1510.
https://doi.org/10.1093/restud/rdv053.
Croson, R., Gneezy, U., 2009. Gender differences in preferences. Journal of Economic
Literature 47(2), 448-474. https://doi.org/10.1257/jel.47.2.448.
Dai, H., Saccardo, S., Han, M.A., Roh, L., Raja, N., Vangala, S., et al., 2021. Behavioural
nudges increase COVID-19 vaccinations. Nature 597, 404-409.
https://doi.org/10.1038/s41586-021-03843-2.
Dixon, D., Den Daas, C., Hubbard, G., Johnston, M., 2022. Using behavioural theory to
understand adherence to behaviours that reduce transmission of COVID-19; evidence from the
25
CHARIS representative national study. British Journal of Health Psychology 27(1), 116-135.
https://doi.org/10.1111/bjhp.12533.
Epton, T., Ghio, D., Ballard, L.M., Allen, S.F., Kassianos, A.P., Hewitt, R., et al., 2022.
Interventions to promote physical distancing behaviour during infectious disease pandemics or
epidemics: a systematic review. Social Science and Medicine 303:114946.
https://doi.org/10.1016/j.socscimed.2022.114946.
Faasse, K., Newby, J., 2020. Public perceptions of COVID-19 in Australia: perceived risk,
knowledge, health-protective behaviors, and vaccine intentions. Frontiers in Psychology
11:551004. https://doi.org/10.3389/fpsyg.2020.551004.
Farboodi, M., Jarosch, G., Shimer, R., 2021. Internal and external effects of social distancing
in a pandemic. Journal of Economic Theory 196:105293.
https://doi.org/10.1016/j.jet.2021.105293.
Ferguson, N., 2007. Capturing human behaviour. Nature 446, 733-733.
https://doi.org/10.1038/446733a.
Franzen, A., Wöhner, F., 2021. Fatigue during the COVID-19 pandemic: evidence of social
distancing adherence from a panel study of young adults in Switzerland. PLoS ONE 16(12):
e0261276. https://doi.org/10.1371/journal.pone.0261276.
Fullerton, M.K., Rabb, N., Mamidipaka, S., Ungar, L., Sloman, S.A., 2021. Evidence against
risk as a motivating driver of COVID-19 preventive behaviors in the United States. Journal of
Health Psychology. https://doi.org/10.1177%2F13591053211024726.
Geoffard, P.-Y., Philipson, T., 1996. Rational epidemics and their public control. International
Economic Review 37(3), 603-624. https://doi.org/10.2307/2527443.
26
Godlonton, S., Thornton, R.L., 2013. Learning from others’ HIV testing: updating beliefs and
responding to risk. American Economic Review: Papers and Proceedings 103(3), 439-444.
https://doi.org/10.1257/aer.103.3.439.
Guillon, M., Kergall, P., 2020. Attitudes and opinions on quarantine and support for a contact-
tracing application in France during the COVID-19 outbreak. Public Health 188, 21-31.
https://doi.org/10.1016/j.puhe.2020.08.026.
Hale, T., Anania, J., de Mello, B.A., Angrist, N., Barnes, R., Boby, T., et al., 2022. Variation
in government responses to COVID-19 (version 13.0). Blavatnik School of Government
Working Paper BSG-WP-2020/032.
Harvey, N., 2020. Behavioral fatigue: real phenomenon, naïve construct, or policy contrivance?
Frontiers in Psychology 11:589892. https://doi.org/10.3389/fpsyg.2020.589892.
Hensel, L., Witte, M., Caria, A.S., Fetzer, T., Fiorin, S., Götz, F.M., et al., 2022. Global
behaviors, perceptions, and the emergence of social norms at the onset of the COVID-19
pandemic. Journal of Economic Behavior and Organization 193, 473-496.
https://doi.org/10.1016/j.jebo.2021.11.015.
Johnson, B.B., Mayorga, M., 2020. Temporal shifts in Americans’ risk perceptions of the Zika
outbreak. Human and Ecological Risk Assessment: An International Journal 27(7), 1242-1257.
https://doi.org/10.1080/10807039.2020.1820852.
Li, Y.-C., Norton, E.C., Dow, W.H., 2004. Influenza and pneumococcal vaccination demand
responses to changes in infectious disease mortality. Health Services Research 39(4), 905-926.
https://doi.org/10.1111%2Fj.1475-6773.2004.00264.x.
27
Lima, M.L., 2004. On the influence of risk perception on mental health: living near an
incinerator. Journal of Environmental Psychology 24(1), 71-84. https://doi.org/10.1016/S0272-
4944(03)00026-4.
Loewenstein, G., Mather, J., 1990. Dynamic processes in risk perception. Journal of Risk and
Uncertainty 3(2), 155-175. https://doi.org/10.1007/BF00056370.
Lorig, F., Johansson, E., Davidsson, P., 2021. Agent-based social simulation of the Covid-19
pandemic: a systematic review. Journal of Artificial Societies and Social Simulation 24(3), 1-
26. http://doi.org/10.18564/jasss.4601.
Lunn, P.D., Timmons, S., Belton, C.A., Barjaková, M., Julienne, H., Lavin, G., 2020.
Motivating social distancing during the COVID-19 pandemic: an online experiment. Social
Science and Medicine 265:113478. https://doi.org/10.1016/j.socscimed.2020.113478.
Makris, M., 2021. Covid and social distancing with a heterogenous population. Economic
Theory. https://doi.org/10.1007/s00199-021-01377-2.
Martínez, D., Parilli, C., Scartascini, C., Simpser, A., 2021. Let’s (not) get together! The role
of social norms on social distancing during COVID-19. PLoS ONE 16(3):e0247454.
McAdams, D., 2021. The blossoming of economic epidemiology. Annual Review of
Economics 13, 539-570. https://doi.org/10.1146/annurev-economics-082120-122900.
Michie, S., West, R., Harvey, N., 2020. The concept of “fatigue” in tackling covid-19. BMJ
371:m4171. https://doi.org/10.1136/bmj.m4171.
Mullahy, J., 1999. It'll only hurt a second? Microeconomic determinants of who gets flu shots.
Health Economics 8(1), 9-24. https://doi.org/10.1002/(sici)1099-1050(199902)8:1%3C9::aid-
hec396%3E3.0.co;2-x.
28
Oster, E., 2012. HIV and sexual behavior change: why not Africa? Journal of Health Economics
31(1), 35-49. https://doi.org/10.1016/j.jhealeco.2011.12.006.
Papageorge, N.W., Zahn, M.V., Belot, M., van den Broek-Altenburg, E., Choi, S., Jamison,
J.C., et al., 2021. Socio-demographic factors associated with self-protecting behavior during
the Covid-19 pandemic. Journal of Population Economics 34(2), 691-738.
https://doi.org/10.1007/s00148-020-00818-x.
Petherick, A., Goldszmidt, R., Andrade, E.B., Furst, R., Hale, T., Pott, A., et al., 2021. A
worldwide assessment of changes in adherence to COVID-19 protective behaviours and
hypothesized pandemic fatigue. Nature Human Behaviour 5(9), 1145-1160.
https://doi.org/10.1038/s41562-021-01181-x.
Philipson, T., 1996. Private vaccination and public health: an empirical examination for U.S.
measles. The Journal of Human Resources 31(3), 611-630. https://doi.org/10.2307/146268.
Picone, G., Kibler, R., Apouey, B.H., 2017. Malaria prevalence, indoor residual spraying, and
insecticide treated net usage in Sub-Saharan Africa. Journal of African Development 19(2), 19-
32. https://doi.org/10.5325/jafrideve.19.2.0019.
Pullano, G., Di Domenico, L., Sabbatini, C.E., Valdano, E., Turbelin, C., Debin, M., et al.,
2021. Underdetection of cases of COVID-19 in France threatens epidemic control. Nature 590,
134-139. https://doi.org/10.1038/s41586-020-03095-6.
Qin, H., Sanders, C., Prasetyo, Y., Syukron, M., Prentice, E., 2021. Exploring the dynamic
relationships between risk perception and behavior in response to the Coronavirus Disease 2019
(COVID-19) outbreak. Social Science and Medicine 285:114267.
https://doi.org/10.1016/j.socscimed.2021.114267.
29
Raude, J., Lecrique, J.-M., Lasbeur, L., Leon, C., Guignard, R., du Roscoät, E., et al., 2020.
Determinants of preventive behaviors in response to the COVID-19 pandemic in France:
comparing the sociocultural, psychosocial, and social cognitive explanations. Frontiers in
Psychology 11:584500. https://doi.org/10.3389/fpsyg.2020.584500.
Raude, J., McColl, K., Flamand, C., Apostolidis, T., 2019. Understanding health behaviour
changes in response to outbreaks: findings from a longitudinal study of a large epidemic of
mosquito-borne disease. Social Science and Medicine 230, 184-193.
https://doi.org/10.1016/j.socscimed.2019.04.009.
Reicher, S., Drury, J., 2021. Pandemic fatigue? How adherence to covid-19 regulations has
been misrepresented and why it matters. BMJ 372:n137. https://doi.org/10.1136/bmj.n137.
Reluga, T.C., 2010. Game theory of social distancing in response to an epidemic. PLoS
Computational Biology 6(5):e1000793. https://doi.org/10.1371/journal.pcbi.1000793.
Schaner, S., Theys, N., Angrisani, M., Banerjee, J., Khobragade, P.Y., Petrosyan, S., et al.,
2022. Adherence to COVID-19 protective behaviours in India from May to December 2020:
evidence from a nationally representative longitudinal survey. BMJ Open 12:e058065.
https://doi.org/10.1136/bmjopen-2021-058065.
Schneider, C.R., Dryhurst, S., Kerr, J., Freeman, A.L.J., Recchia, G., Spiegelhalter, D., et al.,
2021. COVID-19 risk perception: a longitudinal analysis of its predictors and associations with
health protective behaviours in the United Kingdom. Journal of Risk Research 24(3-4), 294-
313. https://doi.org/10.1080/13669877.2021.1890637.
Scholz, U., Freund, A.M., 2021. Determinants of protective behaviours during a nationwide
lockdown in the wake of the COVID-19 pandemic. British Journal of Health Psychology 26(3),
935-957. https://doi.org/10.1111/bjhp.12513.
30
Seale, H., Dyer, C.E.F., Abdi, I., Rahman, K.M., Sun, Y., Qureshi, M.O., et al., 2020. Improving
the impact of non-pharmaceutical interventions during COVID-19: examining the factors that
influence engagement and the impact on individuals. BMC Infectious Diseases 20:607.
https://doi.org/10.1186/s12879-020-05340-9.
Seban, J., Thuilliez, J., Herbreteau, V., 2013. Possession of bed nets in Haut-Katanga (DRC):
prevalence-elastic behaviour or performance of health care system delivery? Health and Place
24, 275-285. https://doi.org/10.1016/j.healthplace.2013.09.001.
Shaman, J., 2021. An estimation of undetected COVID cases in France. Nature 590, 38-39.
https://doi.org/10.1038/d41586-020-03513-9.
Smith, L.E., Potts, H.W.W., Amlôt, R., Fear, N.T., Michie, S., Rubin, G.J., 2021. Adherence to
the test, trace, and isolate system in the UK: results from 37 nationally representative surveys.
BMJ 372:n608. http://doi.org/10.1136/bmj.n608.
Smith, L.E., Potts, H.W.W., Amlôt, R., Fear, N.T., Michie, S., Rubin, G.J., 2022. Engagement
with protective behaviours in the UK during the COVID-19 pandemic: a series of cross-
sectional surveys (the COVID-19 rapid survey of adherence to interventions and responses
[CORSAIR] study). BMC Public Health 22:475. https://doi.org/10.1186/s12889-022-12777-x.
van der Pligt, J., Eiser, J.R., Spears. R., 1986. Attitudes toward nuclear energy: familiarity and
salience. Environment and Behavior 18(1), 75-93.
Verelst, F., Willem, L., Beutels, P., 2016. Behavioural change models for infectious disease
transmission: a systematic review (2010–2015). Journal of the Royal Society Interface
13:20160820. https://doi.org/10.1098/rsif.2016.0820.
31
Wambua, J., Hermans, L., Coletti, P., Verelst, F., Willem, L., Jarvis, C.I., et al., 2022. The
influence of risk perceptions on close contact frequency during the SARS‑CoV‑2 pandemic.
Scientific Reports 12:5192. https://doi.org/10.1038/s41598-022-09037-8.
Weitz, J.S., Park, S.W., Eksin, C. Dushoff, J., 2020. Awareness-driven behavior changes can
shift the shape of epidemics away from peaks and toward plateaus, shoulders, and oscillations.
PNAS 117(51), 32764-32771. https://doi.org/10.1073/pnas.2009911117.
World Health Organization, 2020. Pandemic fatigue: reinvigorating the public to prevent
COVID-19: policy framework for supporting pandemic prevention and management: revised
version November 2020. World Health Organization. Regional Office for Europe. No.
WHO/EURO: 2020-1573-41324-56242.
Wright, L., Fancourt, D., 2021. Do predictors of adherence to pandemic guidelines change over
time? A panel study of 22,000 UK adults during the COVID-19 pandemic. Preventive Medicine
153:106713. https://doi.org/10.1016/j.ypmed.2021.106713.
Young, A., 2007. In sorrow to bring forth children: fertility amidst the plague of HIV. Journal
of Economic Growth 12, 283-327. https://doi.org/10.1007/s10887-007-9021-3.
Zickfeld, J.H., Schubert, T.W., Herting, A.K., Grahe, J., Faasse, K.,8 2020. Correlates of health-
protective behavior during the initial days of the COVID-19 outbreak in Norway. Frontiers in
Psychology 11 :564083. https://doi.org/10.3389/fpsyg.2020.564083.
32
Table 1: Legal restrictions in France since March 2020.
National restriction
Length
Observations
Description
First lockdown
17/03/20-10/05/20
10,013
(5 survey waves)
Strict lockdown in which
people had to stay at home.
Second lockdown
30/10/20-14/12/20
4,000
(2 survey waves)
Mild lockdown in which
people had to stay at home
but schools were open.
Curfew to 6 p.m.
16/01/21-19/03/21
6,001
(3 survey waves)
Obligation to stay at home
between 6 p.m. and 6 a.m.
Curfew to 7 p.m.
20/03/21-18/05/21
2,000
(1 survey wave)
Obligation to stay at home
between 9 p.m. and 6 a.m.
Third lockdown
03/04/21-02/05/21
2,001
(1 survey wave)
Mild lockdown in which
travel was allowed within a
10-km radius from his
home.
Note: Other legal restrictions were imposed at the national level but no data were collected
during these periods.
33
Table 2: Descriptive statistics.
Number of participants
50,019
Sex
Male
46.1%
Female
53.9%
Mean age
49.23 (16.37)
History of COVID-like symptoms
Yes
17.8%
No
82.2%
Completed secondary school
Yes
71%
No
29%
Socioeconomic status
High
48.9%
Low
38.9%
Inactive
12.2%
Financial situation
Good
81.3%
Bad
18.7%
Housing overcrowding
Yes
7.6%
No
92.4%
Note: Standard error in parentheses.
34
Table 3: Association between the death incidence attributable to COVID-19 and the
engagement in three HPB.
Mean death incidence level
Avoid close contacts
with other people
Do not shake hands
Wash hands often
Yes
291.08
253.34
241.51
No
169.98
169.80
221.44
Difference
𝛥= 121.10***
𝛥= 83.54***
𝛥= 20.07***
Note: t-test for comparing the mean death incidence level according to engagement in the
HPB. *** p<0.01, ** p<0.05, * p<0.1
35
Table 4: Association between each sociocultural and demographic variable and the
engagement in social distancing.
Percentage/mean level of individuals
engaging in social distancing
Association
All participants
53.8%
Sex
𝜒²(1) = 41.57***
Male
52.2%
Female
55.1%
Age
51.11 (16.00)
𝛥= 4.11***
History of COVID-like
symptoms
𝜒² (1) = 29.67***
Yes
51.2%
No
54.3%
Completed secondary school
𝜒² (1) = 21.20***
Yes
53.1%
No
55.4%
Socioeconomic status
𝜒² (1) = 25.06***
High
54.7%
Low
53.4%
Inactive
51.3%
Financial situation
𝜒² (1) = 1.23
Good
53.7%
Bad
54.3%
Housing overcrowding
𝜒² (1) = 4.64**
Yes
52.1%
No
53.9%
Note: Standard error in parentheses. t-test for comparing the mean age according to the
engagement in social distancing. Pearson’s chi-square tests for comparing the proportions of
individuals engaging in social distancing according to each sociocultural and demographic
variable. *** p<0.01, ** p<0.05, * p<0.1
36
Table 5: Effect of incidence and sociocultural and demographic factors on the probability
of avoiding close contacts with other people.
Probit estimation
(1)
(2)
(3)
(4)
Incidence
0.000657***
0.000685***
0.00112***
(4.00e-05)
(4.04e-05)
(0.000147)
Stringency index
0.0194***
0.0199***
0.0196***
(0.000614)
(0.000619)
(0.000628)
Time
-0.00103***
-0.00105***
-0.000322***
(4.18e-05)
(4.23e-05)
(5.80e-05)
Incidence × Time
-4.04e-06***
(2.30e-07)
Sex
0.171***
0.0933***
0.176***
(0.0124)
(0.0185)
(0.0125)
Incidence × Sex
0.000367***
(6.52e-05)
Age
0.0136***
0.0130***
0.0140***
(0.000424)
(0.000617)
(0.000428)
Incidence × Age
3.54e-06*
(2.13e-06)
History of COVID-like symptoms
0.00961
0.0102
0.00850
(0.0156)
(0.0156)
(0.0157)
Completed secondary school
0.000514
0.00273
0.00235
(0.0145)
(0.0145)
(0.0146)
Socioeconomic status
Low
0.0362**
0.118***
0.0324**
(0.0142)
(0.0204)
(0.0143)
Inactive
0.118***
0.228***
0.122***
(0.0215)
(0.0312)
(0.0217)
Incidence × Low
-0.000392***
(6.96e-05)
Incidence × Inactive
-0.000504***
(0.000107)
Financial situation
-0.0460***
-0.0838***
-0.0488***
(0.0157)
(0.0231)
(0.0158)
Incidence × Financial situation
0.000156*
(8.02e-05)
Housing overcrowding
0.0929***
0.0960***
0.101***
(0.0234)
(0.0234)
(0.0237)
Constant
-1.127***
-1.916***
-1.966***
0.556***
(0.0405)
(0.0526)
(0.0619)
(0.0507)
Survey fixed effects
No
No
No
Yes
Observations
50,019
50,019
50,019
50,019
R-squared
10.07
11.84
12.45
13.28
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
37
Table 6: Probit COVID-19 incidence elasticities.
Dependent variable: Avoid close contacts with other people
Average partial effect
Elasticity at the average
Average of elasticity
Incidence
0.00024***
0.1166***
0.0889***
(1.39e-05)
(0.0069)
(0.0050)
Note: Estimates derive from the model (2) in the Table 5, including stringency index, time, sex,
age, history of COVID-like symptoms, education level, socioeconomic status, financial
situation and housing overcrowding. Standard errors are in parentheses. *** p<0.01, ** p<0.05,
* p<0.1
38
Figure 1: Percentages of participants engaging in HPB and death incidence level over
time. Error bars are 95% confidence intervals.
39
Figure 2: Association between death incidence and the engagement in each HPB.
