What can we learn from Covid-19 pandemic's impact on human behaviour? The case of France's lockdown

The unprecedented COVID-19 pandemic and the resulting large-scale lockdowns offer a unique opportunity to study human behavior under extreme circumstances. While quarantines have been studied before, the scale and duration of lockdowns in 2020 were unparalleled. The pandemic presented a dual crisis: a biological one from the virus itself and a social crisis stemming from individual perceptions and reactions. Understanding individual decisions during this period is crucial for improving future pandemic management strategies. Past research suggests that protective measures adherence is influenced by factors like gender, age, anxiety levels, and trust in government. Fear of infection, self-perceived vulnerability, and risk of severe illness also play significant roles. However, risk perception is often biased, with individuals tending to underestimate their own risk. Government communication is a key modulator of risk perception and subsequent behavior. This study uses a large-scale online survey conducted during the French lockdown to identify individual characteristics and perceptions that influenced adherence to government measures.

Previous research on epidemics and quarantines provides valuable context. Studies have shown that protective measures are often better followed by women, older individuals, people with high anxiety, and those who trust the government. Fear of infection and perceived vulnerability significantly increase compliance with recommendations. Interestingly, self-estimated risk tends to focus on personal vulnerability rather than the risk of infecting others. Furthermore, government messaging has been shown to significantly modulate risk perception and protective behaviors. Existing studies highlight that acceptance of measures hinges more on perception of the situation than pre-existing dispositions.

An online survey was conducted from March 24th to May 10th, 2020, encompassing the French lockdown period. The survey, distributed through various networks (professional, personal, associations, political parties, companies, newspapers), targeted adults residing in metropolitan France. Data from individuals under 18, those not living in metropolitan France, and those residing outside France during the lockdown were excluded. The survey utilized LimeSurvey and Google Forms to collect data on demographics, occupation, mobility, living habits, social perceptions, residency, lockdown perceptions, feelings, and intentions for post-pandemic behavior changes. Data analysis employed R software, utilizing various statistical methods: Chi-squared tests for 2x2 contingency tables, ANOVA for numerical variables, partial binomial multiple regressions for bivariate variables, and partial multinomial logistic regression for categorical variables. Partial models controlled for potential biases by including multiple predictors. Cohen's d effect sizes were calculated. The Bonferroni correction addressed multiple comparisons. K-means clustering identified distinct participant profiles based on lockdown perceptions and feelings. Text mining analyzed free-text responses to identify word frequencies and thematic differences across clusters.

The study involved 12,064 participants. Analysis revealed that 11.7% changed their usual residence in the days leading up to the lockdown, primarily to be closer to family or friends. Logistic regression showed that residence changes before lockdown were predicted by factors such as familial organization (higher for those living alone or in shared housing without children), professional situation (higher for students, retired people, and unemployed individuals), age (negative correlation), residency type (higher for city apartment dwellers), perceived salary (higher for those reporting a comfortable life), and gender (higher for women). K-means clustering identified three distinct participant profiles based on lockdown perception: Cluster 1 (30.8%) showed positive lockdown experiences and pandemic concern; Cluster 2 (33.8%) was pessimistic, stressed, less compliant, and dissatisfied with government measures; Cluster 3 (35.4%) was less affected and exhibited lower compliance, believing others were more compliant. Multinomial logistic regression revealed that these clusters differed significantly across several factors such as perceived salary, gender, education level, residency type, home composition, region of residency, and socio-professional categories. Text mining revealed distinct discourse patterns among the clusters. Cluster 1 focused on new activities and positive aspects; Cluster 2 expressed negative feelings; and Cluster 3 focused on continued activities in a quieter environment. Post-lockdown, differences emerged in willingness to change various aspects of life, particularly regarding social interactions and personal habits. The study further suggests that lockdown compliance was inversely related to unfavorable living environments (socioeconomic inequality), high fear combined with distrust in government measures, and low risk perception based on perceived high compliance of others.

The findings highlight significant pre-lockdown mobility, with over 10% of participants changing residences, potentially contributing to virus spread. This supports prior research showing that population movement control reduces transmission. Three distinct profiles reveal how diverse perceptions shaped lockdown compliance. Cluster 2, characterized by socioeconomic disadvantages, fear, and distrust, exhibited lower compliance, aligning with research highlighting the vulnerability of low socioeconomic groups. Cluster 3's low compliance stems from a perception of low personal risk and the belief that others were compliant, illustrating the role of social comparison and risk perception biases. The findings underscore the importance of addressing pre-existing inequalities and tailoring communication strategies to specific subgroups to improve adherence to public health recommendations during crises. Fear, while potentially beneficial for compliance, can also negatively impact decision-making under uncertainty.

This study offers valuable insights into human behavior during a large-scale lockdown. The identification of three distinct behavioral profiles based on perceptions of risk and trust emphasizes the need for nuanced government communication. Future research should investigate the long-term effects of the pandemic and lockdown on individual behavior and mental health. Targeted interventions addressing socioeconomic inequalities and fostering trust in public health measures are crucial to effectively manage future health crises.

The study's limitations include an unbalanced sample with overrepresentation of highly educated and executive-level participants, the lack of data from the first week of the lockdown, and the absence of information on COVID-19 infection status among participants or their close contacts. These limitations should be considered when interpreting the results.