A machine learning approach to construct quarterly data on intangible investment for Eurozone

Abstract

We develop a novel approach to construct quarterly time series data for annually measured intangible investment variables. We accomplish this by using machine learning methods to explore the relationship between these variables and key macroeconomic time series available on a quarterly frequency. The proposed approach offers some advantages over other econometric techniques. Specifically, it does not require any ex-ante assumptions for the link between the quarterly time series and their annual counterpart, while minimizing the need for computationally expensive algorithms and necessitating almost no data pre-processing. To demonstrate the usefulness of the constructed data, we present some business cycles facts for the intangible economies of Eurozone and estimate a dynamic factor model.