High-mixed-frequency forecasting models for GDP and inflation

This paper analyzes the technical and practical issues involved in the use of data at mixed frequencies (quarterly and monthly and, possibly, weekly and daily) to forecast monthly and quarterly economic activity in a country. In particular, it considers alternative high-frequency forecasting models for GDP growth and inflation in the Philippines, utilizing indicators that are observable at different frequencies and with particular focus on dynamic time-series models that involve latent factors. The study compares the forecasting performance of this approach with more commonly used data-intensive methods that have been developed in applications in the U.S. and Europe. These alternative approaches include Mixed Data Sampling (MIDAS) Regression and Current Quarter Modeling (CQM) with Bridge Equations. While these alternatives are mostly data-intensive, the dynamic latent factor modeling with mixed frequencies presents a parsimonious approach which depends on a much smaller data set that needs to be updated regularly. But it also faces additional complications in methodology and calculations as mixed-frequency data are included in the analysis…