Can we predict the quality of the economic growth of the country simply by reading the newspapers? Three researchers linked to the Insee-Clément Bortoli, Stephanie Combes and Thomas Renault – responded in the affirmative in an interesting study, published Thursday in the journal Economics and statistics of the Insee. Classically, the statistics of the gross domestic Product (GDP) in French are published quarterly, with a deadline of thirty days at the end of each quarter. Between these dates, to assess the quality of growth, researchers can rely on the surveys.
The economists, Insee has decided to refine those estimates planchant on the newspaper articles available on the Internet that deal with the economic news. They began with a database of one million articles published in the daily newspaper The World between 1990 and 2017. First they have refined this database to retain only the articles dealing with the economic situation in hex, or 200 000 papers. In order to process these contents, using the techniques of Big data, the economists were then drawn up “a list of terms connoted positively or negatively from an economic point of view”. These terms were identified in each section of the base which is to assign a “score sense”.
“feeling media”
“Thus, it is possible to synthesize the information contained in the database in the form of a single numerical indicator, which we call sentiment in the media”, summarize the authors. Once this indicator was built, economists have compared the results obtained with the other indicators available. They conclude “that a model which combines sentiment in the media and surveys brings, for some horizons of prediction, an accuracy significantly superior to that of a model auto-regressive [treatment of a single observed variable at different periods of time, editor’s NOTE] increased only surveys.”
in other words, considered alone, the business climate in the Insee remains more reliable than the new sentiment indicator in the media, but this brings many clarifications and complementary information, and allows this title to improve the forecasts of GDP.
