Belgium Monthly Index
Download Data Annotated Chart
We are pleased to host three new EPU indices for Belgium developed by Samuel Borms, Kris Boudt, Jeroen Van Pelt, and Andres Algaba. Their approach is laid out in the research note "The Economic Policy Uncertainty Index for Flanders, Wallonia and Belgium", available here.
EPU indices are computed separately for the two main languages of the country. The index based on the Dutch language includes 8 newspapers (De Tijd, De Standaard, De Morgen, Het Laatste Nieuws, Het Nieuwsblad, Gazet van Antwerpen, Het Belang van Limburg, and Het Volk) and the index based on the French language includes five newspapers (L'Avenir, La Dernière Heure, La Libre Belgique, Le Soir, and L'Echo). A final Belgium index is then defined as the average of the Belgium-Dutch and Belgium-French indices. The indices span from January 2001 to the present, at a monthly interval.
The construction follows the procedure in "Measuring Economic Policy Uncertainty" by Scott Baker, Nicholas Bloom, and Steven Davis, reflecting normalized frequency counts of news articles related to economic policy uncertainty. The newspaper-level Belgium-Dutch and Belgium-French indices (frequency counts scaled by the total number of articles per month) are normalized to unit standard deviation, and the respective averaged series is normalized to mean 100, both for a reference period ranging from the start of the time series up to December 2010. The index for Belgium is labeled in following annotated chart.
The keywords for article selection were defined starting from the Dutch words by Kroese et al. (2015). Next, candidate words to add to the keywords were generated through an existing word embedding space. The authors then manually sifted through the propositions and decided on a word list of around 150 words combined for the economic (E), policy (P) and uncertainty (U) categories. Instead of doing the same for the French keywords, the authors opted to translate the Flemish words to French to maximize keywords equivalence.
The constructed EPU indices clearly reflect national as well as worldwide evolutions. This duality is no surprise as Belgium lies at the heart of Europe, with Brussels housing for instance the headquarters of the European Commission. The validation exercise shows sufficient correlation with EPU indices for the Netherlands, Germany, and Europe, as well as stock market volatility, and the official Belgian consumer confidence indicator. The research note also presents an automated news content analysis, using noun and proper noun extraction together with biterm topic modeling. Such analysis helps to understand the underlying sources of uncertainty explicitly mentioned in the news articles. Lastly, the authors show the evolution of daily EPU news coverage during early 2020, and unveil interesting dynamics about the rise and possible fall of uncertainty in light of the COVID-19 pandemic.
Note: The index received a major update in 2021 following the refinement of the keywords in the Dutch and French language. There are now around 100 words combined for the economic (E), policy (P) and uncertainty (U) categories. It also includes a new filter that restricts the article selection to news discussing Belgium. This filter is applied to both the EPU articles and the total number of articles per month as part of the EPU computation. In addition, the update closed an existing data gap in the coverage of the EPU Belgium index in the French language for the year 2006. The older version of the index can be obtained from the authors. Olivier Delmarcelle, Arno De Block, David Ardia, Keven Bluteau, and Kris Boudt contributed to the development of this update.