• Publicationsclose
• Research dataclose
• Other research productsclose
• FR close
• Hal-Diderot close
• Mémoires en Sciences de l'Information et de la Communication close
• Hyper Article en Ligne close
• Hyper Article en Ligne - Sciences de l'Homme et de la Société close
• COVID-19 close
• Digital Humanities and Cultural Heritage close

Abstract Using data from 5474 banks located in 23 OECD countries over the period 2019Q2–2022Q1, we study the influence of COVID-19 on bank profitability (before and during the COVID-19 vaccination period). Our results show a negative impact of the COVID-19 pandemic on bank profitability, especially at the onset of the health crisis. In addition, we find that vaccination against COVID-19 had a positive effect on bank profitability, not yet sufficient to compensate for the losses generated at the beginning of the pandemic. Finally, we show that these effects depend on the characteristics of banks (notably size and capital) before vaccination and on the severity of the crisis across countries. Overall, we provide the first evidence of the influence of vaccination on bank behavior in terms of profitability.

ABSTRACTThe virus neutralization test (VNT) is the reference for the assessment of the functional ability of neutralizing antibodies (NAb) to block SARS-CoV-2 entry into cells. New competitive immunoassays measuring antibodies preventing interaction between the spike protein and its cellular receptor are proposed as surrogate VNT (sVNT). We tested three commercial sVNT (a qualitative immunochromatographic test and two quantitative immunoassays named YHLO and TECO) together with a conventional anti-spike IgG assay (bioMérieux) in comparison with an in-house plaque reduction neutralization test (PRNT50) using the original 19A strain and different variants of concern (VOC), on a panel of 306 sera from naturally-infected or vaccinated patients. The qualitative test was rapidly discarded because of poor sensitivity and specificity. Areas under the curve of YHLO and TECO assays were, respectively, 85.83 and 84.07 (p-value >0.05) using a positivity threshold of 20 for PRNT50, and 95.63 and 90.35 (p-value =0.02) using a threshold of 80. However, the performances of YHLO and bioMérieux were very close for both thresholds, demonstrating the absence of added value of sVNT compared to a conventional assay for the evaluation of the presence of NAb in seropositive subjects. In addition, the PRNT50 assay showed a reduction of NAb titers towards different VOC in comparison to the 19A strain that could not be appreciated by the commercial tests. Despite the good correlation between the anti-spike antibody titer and the titer of NAb by PRNT50, our results highlight the difficulty to distinguish true NAb among the anti-RBD antibodies with commercial user-friendly immunoassays.

As the pandemic progressed, disinformation, fake news and conspiracy spread through many parts of society. However, the disinformation spreading through social media is, according to the literature, one of the causes of increased COVID-19 vaccine hesitancy. In this context, the analysis of social media is particularly important, but the large amount of data exchanged on social networks requires specific methods. This is why machine learning and natural language processing (NLP) models are increasingly applied to social media data.Objective: The aim of this study is to examine the capability of the CamemBERT French language model to faithfully predict elaborated categories, with the knowledge that tweets about vaccination are often ambiguous, sarcastic or irrelevant to the studied topic.Methods: A total of 901,908 unique French tweets related to vaccination published between July 12, 2021, and August 11, 2021, were extracted using the Twitter API v2. Approximately 2,000 randomly selected tweets were labeled with two types of categorization: (1) arguments for ("pros") or against ("cons") vaccination (sanitary measures included) and (2) the type of content of tweets ("scientific", "political", "social", or "vaccination status"). The CamemBERT model was fine-tuned and tested for the classification of French tweets. The model performance was assessed by computing the F1-score, and confusion matrices were obtained. Results: The accuracy of the applied machine learning reached up to 70.6% for the first classification ("pros" and "cons" tweets) and up to 90.0% for the second classification ("scientific" and "political" tweets). Furthermore, a tweet was 1.86 times more likely to be incorrectly classified by the model if it contained fewer than 170 characters (odds ratio = 1.86; 1.20 < 95% confidence interval < 2.86).Conclusions: The accuracy is affected by the classification chosen and the topic of the message examined. When the vaccine debate is jostled by contested political decisions, tweet content becomes so heterogeneous that the accuracy of the models drops for less differentiated classes. However, our tests showed that it is possible to improve the accuracy of the model by selecting tweets using a new method based on tweet size. Les réseaux sociaux participent activement à la diffusion de la désinformation sur la covid-19 et, selon de nombreuses études, auraient causés l’augmentation de la réticence vis-à-vis des vaccins anti-covid. Dans ce contexte, l’analyse des réseaux sociaux est des plus importants en matière de santé publique. Toutefois, au regard du grand volume de données échangées chaque jour par les internautes, elle nécessite des méthodes spécifiques. C’est pourquoi les chercheurs ont de plus en plus souvent recours aux modèles d’apprentissage automatisé et au traitement du langage naturel (NLP) en particulier. L’objectif de la présente étude est d’examiner la capacité du modèle CamemBERT, pré-entraîné sur la langue française, à catégoriser automatiquement les tweets traitant de la vaccination alors qu’ils sont souvent ambigus, sarcastique ou sans rapport avec le sujet.Les résultats obtenus, sur 2 000 tweets francophones, montrent que la précision de l'apprentissage automatique atteint jusqu'à 70,6 % pour la première classification (tweets « pour » et « contre ») et jusqu'à 90,0 % pour la seconde (tweets « scientifiques » et « politiques »). De plus, un tweet a 1,86 fois plus de chances d'être mal classé par le modèle s'il contient moins de 170 caractères que s'il en contient plus de 170 (odd ratio = 1,86 ; 1,20 < intervalle de confiance à 95 % < 2,86).En conclusion, la précision du modèle est affectée par la classification choisie et le sujet du message examiné. Lorsque le débat sur le vaccin est bousculé par des décisions politiques contestées, les tweets deviennent si hétérogènes que la précision des modèles chute sur les classes les moins différenciantes. Toutefois, nos tests ont également montré qu'il serait possible d'améliorer la précision du modèle en sélectionnant les tweets à l'aide d'une nouvelle méthode basée sur la taille des tweets.

International audience; In this paper, we examine the changes in the dependence structure of global stock markets amid the outbreak of COVID-19. We divide 56 stock markets into developed, emerging, and frontier markets and study their daily price data from 15 October 2019 to 17 August 2020 using the canonical vine (C-vine) copula approach. We observed significant changes in the dependence structure, the selection of the pair copula families, and the associated parameter estimates in the tree. In developed markets, during the COVID-19, the dependence of markets shifted from the Netherlands to France while the root node position of Hungary replaced the market of Poland. The frontier markets showed strong dependence signs with Mauritius before COVID-19 and Slovenia during the outbreak. Our findings are of interest to regulators and practitioners, particularly in monitoring the value at risk of portfolios and adopting appropriate strategies in light of the varying dynamics of stock markets during extreme events, such as COVID-19.

Background: Senegal reported the first COVID-19 case on March 2, 2020. A nationwide cross-sectional epidemiological survey was conducted to capture the true extent of COVID-19 exposure. Methods: Multi-stage random cluster sampling of households was carried out between October 24 and November 26, 2020, at the end of the first wave of COVID-19 transmission. Anti-SARS-CoV-2 antibodies (IgG and/or IgM) were screened using three distinct ELISA assays. Adjusted prevalence for the survey design were calculated for each test separately, and thereafter combined. Crude, adjusted prevalence based on tests performances and weighted prevalence by sex-age strata were estimated to assess the seroprevalence. Findings: Of the 1,463 participants included in this study, 58·8% were women and the mean age of participants was 29·2 years (range 0·25–82·0). The national seroprevalence was estimated at 28 . 4% (95% CI: 26·1-30·8). There was substantial regional variability. Four regions recorded the highest seroprevalence: Ziguinchor (56·7%), Sedhiou (48·0%), Dakar (44·0%) and Kaolack (32·7%) whereas, Louga (11·1%) and Matam (11·2%), located in the Center-North, were less impacted in our analysis. All age groups were impacted and the prevalence of SARS-CoV-2 was comparable in symptomatic and asymptomatic groups. We estimated 4,744,392 SARS-CoV-2 (95% CI: 4,360,164 – 5,145,327) potential infected in Senegal compared to 16,089 COVID-19 RT-PCR laboratory-confirmed cases reported at the time of the survey. Interpretation: These results provide an estimate of SARS-CoV-2 virus dissemination in the Senegalese population. Preventive and control measures need to be reinforced in the country and especially in the south border regions. Funding Information: This work was supported by US Centers for Disease Control and Prevention (CDC), the Senegalese Ministry of Health, the Senegalese National Statistics and Demography Agency (ANSD), the WHO Unity program and the COVID-19 Task-force of the International Pasteur Institute Network (IPIN, REPAIR project). Declaration of Interests: We declare no competing interests. Ethics Approval Statement: All participants have consented to participate in the study. For people younger than 18 years, a legal representative provided informed consent. The study was approved by the Senegalese National Ethics Committee for Research in Health (reference number N°0176/MSAS/DPRS/CNERS, 10 October 2020).

Face à la pandémie du COVID-19, la science de l’information se trouve confrontée à un certain nombre de nouvelles problématiques, liées aux crises d’information, auxquelles elle n’était pas préparée. Qui dit crise mondiale, crise économique, crise énergétique, changement climatique, etc... dit crise d’information, crise caractérisée par une augmentation drastique de la quantité d’informations, par une épidémie d’information i.e. une infodémie et par de nouvelles et nombreuses mauvaises conduites des pratiques informationnelles. La crise sanitaire liée à la pandémie du COVID-19 n’échappe pas à la règle: infodémie de bonnes informations médicales et sanitaires mais aussi infodémie de mauvaises informations, bonnes pratiques informationnelles mais aussi nombreuses inconduites. Habituée au registre de la vérité scientifique, la science de l’information doit affronter le registre de la post-vérité et du complotisme qui semble caractériser nos sociétés actuelles. Information-soupçon, information-dénonciation, information- mensonge, c’est, pour la science de l’information, un nouvel horizon guère réjouissant qu’elle devrait pouvoir maîtriser sans problème en mobilisant son arsenal de concepts, de méthodes, de lois, de modèles et de théories bien établis en conservant et en développant l’horizon réjouissant de l’information-vérité. In the face of the COVID-19 pandemic, information science is facing a number of new challenges related to information crises for which it was not prepared. Who says global crisis, economic crisis, energy crisis, climate change, etc ... says information crisis, crisis characterized by a drastic increase in the amount of information, by an information epidemic ie an infodemic and by news and numerous misconduct of information practices. The health crisis linked to the COVID-19 pandemic is no exception to the rule: infodemic of good medical and health information but also infodemic of bad information, good informational practices but also numerous misconduct. Accustomed to the register of scientific truth, information science must confront the register of post-truth and conspiracy that seems to characterize our current societies. Information-suspicion, information-denunciation, information-lies, it is, for the science of information, a hardly encouraging new horizon that it should be able to master without problem by mobilizing its arsenal of concepts, methods, laws, of well- established models and theories by preserving and developing the joyous horizon of information-truth.

La 15e journée doctorale de l’École doctorale d’archéologie de l’université Paris 1 Panthéon Sorbonne (ED 112) avait pour thème l’alimentation et était intitulée : « À table ! De l’approvisionnement au dernier repas. Regards croisés sur l’archéologie de l’alimentation ». Initialement prévue le 27 mai 2020 dans les locaux de l’INHA, il avait été impossible de maintenir cette journée d’étude en raison de la crise sanitaire liée à la Covid-19. Délocalisée dans la sphère du virtuel, elle s’est dé...

COVID-19 outbreaks at nursing homes during the recent pandemic, which received ample media coverage, may have lasting negative impacts on individuals’ perceptions regarding ursing homes. We argue that this could have sizable and persistent implications for savings and long-term care policies. We first develop a theoretical model predicting that higher nurs- ing home aversion should induce higher savings and stronger support for policies subsidizing home care. We further document, based on a survey on Canadians in their 50s and 60s, that higher nursing home aversion is widespread: 72% of respondents are less inclined to enter a nursing home because of the pandemic. Consistent with our model, we find that the latter are much more likely to have higher intended savings for older age because of the pandemic. We also find that they are more likely to strongly support home care subsidies.

Italy was hit harshly by the Covid-19 pandemic, registering more than 35,000 Covid-19 deaths between February and July 2020. During this first wave of the epidemic, the virus spread unequally across the country, with northern regions witnessing more cases and deaths. We investigate demographic and socio-economic factors contributing to the diverse regional impact of the virus during the first wave. Using generalized additive mixed models, we find that Covid-19 mortality at regional level is negatively associated with the degree of intergenerational co-residence, number of intensive care unit beds per capita, and delay in the outbreak of the epidemic. Conversely, we do not find strong associations for several variables highlighted in recent literature, such as population density or the share of the population who are older or have at least one chronic disease. Our results underscore the importance of context-specific analysis for the study of a pandemic.

This paper examines the empirical relationship between individuals’ cognitive and non-cognitive abilities and COVID-19 compliance behaviors using cross-country data from the Survey of Health, Ageing and Retirement in Europe (SHARE). We find that both cognitive and non-cognitive skills predict responsible health behaviors during the COVID-19 crisis. Episodic memory is the most important cognitive skill, while conscientiousness and neuroticism are the most significant personality traits. There is also some evidence of a role for an internal locus of control in compliance. usc Refereed/Peer-reviewed