Don par téléphone

Contacter la fondation de l’Université Laval au 418-656-3292


Don par la poste

Imprimer cette fiche remplissez-la et postez-la avec votre paiement à l’adresse suivante :

Faculté des sciences sociales
Pavillon Charles-De Koninck
1030, av. des Sciences-Humaines
Local 3456-E
Québec (Québec) G1V 0A6

Événements

Midi de science politique

Le Département de science politique, la Chaire de recherche sur la démocratie et les institutions parlementaires, le Centre d'analyse des politiques publiques, le Groupe de recherche en communication politique et l'APEUL vous invitent à assister à la conférence de Mickael Temporão, Corentin Vande Kerckhove et Yannick Dufresne.

La conférence s'intitule : « Idéologies non-dites : Validation d'indicateurs issus des médias sociaux ».

Data is ubiquitous and publicly-available on social media such as Twitter. However, uncertainties persist about the potential of these data to measure valid political attitudes and ideologies. The difficulty to validate estimates based on social media data explains in large parts why some doubt of the scientific potential of these new types of data sources. Most validation procedures used in the literature focus on elites' data that can be associated to specific ideologies. Yet, research show that citizens do not share elites' ideological consistency. There is therefore no direct ways of validating citizens' ideologies estimated with social media. To address this issue, this paper uses a survey of 62,633 social media users to validate network and textual indicators of ideologies with conventional survey indicators. Hence, this large-scale survey offers a unique opportunity to empirically demonstrate the value of publicly-available social media data for public opinion research.

In order to show the reliability of social media ideological indicators, this paper first shows the association existing between partisan elites' ideological positions estimated with their publicly-available network and textual data. The strong correlations existing between these two indicators and the elites' partisan ideologies suggest that these indicators capture a common underlying ideological dimension. A similar unidimensional structure can be estimated with citizens' social media network and textual information. However, contrary to partisan elites, most citizens do not display publicly their partisan ideologies. We must thus rely on conventional survey data. Our results show strong correlations between conventional survey ideological indicators and those estimated with network and textual social media data. These findings show that valid instruments can be estimated with real-time publicly-available data that have the potential to enhance fundamentally our understandings of public opinion dynamics.

Au plaisir de vous y voir!

Date et heure
Jeudi 31 mars 2016 de 11 h 45 à 13 h

Lieu
Université Laval, pavillon De Koninck, salle 3470