Over the years we have done a lot of R&D around this topic, and shared the results through various conference presentations. Although some cases are more than a decade old, the content of these presentations is still very relevant today.
One such presentation from 2006 covers the do’s and don’ts of working with Vector Auto Regressive (VAR) models. VAR models are a class of time-series models which remain widely used for many applications. The early generations of our ESG models were built on VAR models. And although these models have since then evolved into a much more advanced frequency domain methodology, its unique Dynamic Factor Models (DFM) are still at heart driven by VAR models for Principal Component Factors (PCA) per frequency band.
Over the years we have learned what works well and what does not work so well when dealing with VAR models. We are happy to share these learnings with you in what is the first of a series of “scenario modeling evergreens” and hope that you will find the content useful. Click here to find out more about our Economic Scenario Generator (ESG) models.