Predicting Stock Market Drawdowns using Polymodels

We propose a Systematic Risk Indicator derived from a polymodel estimation. Polymodels allow us to measure the strength of the links that a stock market maintains with its economic environment. We show that these links tend to be more extreme before a market crisis, confirming the well-known increase of correlations while proposing a more subtle perspective. A fully automated and successful trading strategy is implemented to assess the interest of the signal, which is shown to be strongly significant, both from an economic and statistical point of view. Results are robust across different time-periods, for various sets of explanatory variables, and among 12 different stock markets.

Download presentation