After VAR Models Do's and Don'ts, a second such presentation from 2012 covers the design of a special pass band filter. Filters are used to decompose macroeconomic and financial market time-series data into for example trend + cycle + seasonal + random components. The underlying idea of such decompositions is that different “forces” drive for example long-term economic growth compared to intraday trading effects on a stock exchange.
This pass band filter is central to the one-of-a-kind frequency domain methodology that we use for generating scenarios. We use it to separate trends from business cycle and monthly intra-year fluctuations. Together with Dynamic Factor Models (DFM) for each of these components it constitutes the “bi-orthogonal decomposition approach” which drives the time-varying risk and return across economies, asset classes and investment horizons in our scenarios.
Based on 10 years of practical experience we have improved it along the way but essentially we still work with this very same filtering approach. We are happy to share its workings with you by means of this presentation. Click here to find out more about our Economic Scenario Generator (ESG) models.