Business cycle measurement with semantic filtering: A micro data approach

Abstract

In this paper we develop a business cycle measure that can be shown to have excellent ex-ante forecasting properties for GDP growth. For identifying business cycle movements, we use a semantic approach. We infer nine different states of the economy directly from firms' responses in business tendency surveys. Hence, we can identify the current state of the economy. We therewith measure business cycle fluctuations. One of the main advantages of our methodology is that it is a structural concept based on shock identification and therefore does not need any - often rather arbitrary - statistical filtering. Futhermore, it is not subject to revisions, it is available in real-time and has a publication lead to official GDP data of at least one quarter. It can therefore be used for one quarter ahead forecasting real GDP growth.