IBD Salle 16
AMU - AMSE
5-9 boulevard Maurice Bourdet
Anushka Chawla : anushka.chawla[at]univ-amu.fr
Laura Sénécal : laura.senecal[at]univ-amu.fr
Carolina Ulloa Suarez : carolina.ulloa-suarez[at]univ-amu.fr
Having real time information about the state of the economy is crucial for any economic agent whether governments, firms or even households, to take rational decisions. One of the main economic aggregates reflecting the business cycle is the Gross Domestic Product (GDP). Quarterly estimates of GDP are released very late, until 45 days or more after the quarter. To get timely information, several indexes about the state of the economy was created such as the Stock-Watson index and the Eurocoin based on available administrative data and survey data. In this paper, we build a new business cycle index based on online news which can be used to nowcast GDP growth rate. For that purpose, we use almost 600.000 French news articles to build sentiment variables about different aspect of the economy. These sentiment variables are used in a Recurrent Neural Network model to predict the business cycle.