July 12, 2021
David Ardia, Associate Professor in the Department of Decision Sciences, received the Best Paper Award for 2018-2019 from the International Journal of Forecasting.
The article is entitled Questioning the news about economic growth: Sparse forecasting using thousands of news-based sentiment values. It was co-authored with Kris Boudt, Professor at Ghent University, Vrije Universiteit Brussel and Vrije Universiteit Amsterdam, and Keven Bluteau, Postdoctoral Researcher at Ghent University and HEC Montréal.
The award was officially presented to the three researchers at the 2021 symposium of the International Institute of Forecasters. Based on a very competitive process, the selection of the winning entry was made from among all the articles published in the journal during 2018 and 2019.
The research discussed in the article aimed to forecast US industrial production by analyzing texts published in the American media. Thus, in addition to the habitual usage of macroeconomic or financial variables, the researchers added thousands of textual sentiment variables to their model. For each topic covered in a news article at a specific moment in time, a sentiment was automatically extracted and assigned a score – either negative or positive.
The researchers then used machine learning techniques to determine which of these variables were most relevant for forecasting industrial production in the US. They noticed that the variables extracted from texts significantly improved the forecasts. In short, their research has helped develop algorithms to read media texts and improve forecasting models.
Interesting point: this artificial intelligence model is an open access tool. For example, an investment fund could use it to improve the asset yield projections in its portfolio.
The team would also like to develop a prediction model for the Quebec economy, like the one developed for the award-winning article. Incidentally, another research project was carried out recently, using news articles drawn from Quebec media archives. The aim, in this instance, was to measure economic policy-related uncertainties for Quebec and Canada.
David Ardia is an Associate Professor in the Department of Decision Sciences at HEC Montréal. Trained in quantitative methods for finance, he has a keen interest in asset allocation, risk management and text mining. He is a founding member of the Sentometrics organization, and a regular member of GERAD (Group for Research in Decision Analysis), Fin-ML and Quantact.
David Ardia spent four years in the financial industry in Switzerland and was named “Swiss Risk Manager of the Year” in 2018.