Uncertainty is acknowledged to be a source of economic fluctuations. But, does the type of uncertainty matter for the economy’s response to an uncertainty shock? This paper offers a novel identification strategy to disentangle different types of uncertainty. It uses machine learning techniques to classify different types of news instead of specifying a set of keywords. The paper finds that, depending on its source, the effects of uncertainty on a macroeconomic variable may differ. I find that both good (expansionary effect) and bad (contractionary effect) types of uncertainty exist.
Larsen, Vegard Høghaug; Thorsrud, Leif Anders & Zhulanova, Julia (2020)
News-driven inflation expectations and information rigidities
Using a large news corpus and machine learning algorithms we investigate the role played by the media in the expectations formation process of households, and conclude that the news topics media report on are good predictors of both inflation and inflation expectations. In turn, in a noisy information model, augmented with a simple media channel, we document that the time series features of relevant topics help explain time-varying information rigidity among households. As such, we provide a novel estimate of state-dependent information rigidities and present new evidence highlighting the role of the media in understanding inflation expectations and information rigidities.
Bergholt, Drago; Larsen, Vegard Høghaug & Seneca, Martin (2019)