A0860
Title: Mixing data of different sampling frequencies in the frequency domain
Authors: Marc Wildi - Zurich University (Switzerland) [presenting]
Abstract: Synthesizing information from data sampled at different frequencies - say daily, weekly, monthly or quarterly data - has a strong appeal in the context of real-time economic monitoring, since indicators can be up-dated continuously, as new information drops in. We propose a new mixed-frequency approach entirely designed in the frequency-domain. Specifically, we tackle the problem of integrating the various (unequally sampled) spectra within a common multivariate framework (MDFA: Multivariate Direct Filter Approach) and we derive closed-form solutions for various practically relevant optimization principles, including minimal (mean-square) revision errors. We discuss some of the advantages of the frequency-domain over the time-domain in the context of mixed-frequency data. Finally, the methodology is applied to a high-frequency indicator of US-GDP, combining quarterly and monthly macro-, weekly employment- and daily market-data.
