Title: Components of Uncertainty
Authors: Vegard Larsen - Norges Bank (Norway) [presenting]
Abstract: Newspaper content is used to create category specific measures of uncertainty. The news are classified using a topic model called Latent Dirichlet Allocation. This model allow us to label a large number of news articles by their underlying content. Uncertainty measures are calculated based on the count of uncertainty terms over time within specific types of news articles in Norways largest business newspaper. The uncertainty measures are available at a daily frequency and they capture well known events linked to both uncertainty at an aggregate level, but also category specific events. We investigate the response from changes in different types of uncertainty to aggregate macroeconomic variables in a structural VAR framework. Several of the category specific uncertainty measures foreshadow weaker macroeconomic outcomes. But, different uncertainty categories can have different effects. Grouping related categories together, we find that an increase in uncertainty related to the categories Household, Financial, and Oil lead to a sizable negative effect on the Norwegian economy. Policy uncertainty seem not to be important for the Norwegian business cycle when it is grouped together, but an increase in uncertainty related to Fiscal policy is followed by weaker economic conditions.