Outlier propagation from volatility forecasts to the realized–implied volatility spread (VRP)

Test the conjecture that extreme values in realized variance forecasts do not translate into extreme values in the realized-minus-implied volatility spread used to sort option portfolios, due to co-movement in implied volatility; characterize the distributional relationship between volatility forecast outliers and VRP outliers.

Background

The authors observe heavy-tailed realized variance with extreme outliers that complicate forecast evaluation. They hypothesize that when realized volatility is large, implied volatility is often large as well, potentially muting outliers in the realized–implied spread used for option portfolio sorting, which could explain differences between volatility forecast performance and option-return outcomes.