Not all VIXs are (Informationally) equal: Evidence from affine GARCH option pricing models

This paper examines which VIX maturity to use in affine GARCH model estimation, when the objective is to do option pricing. Utilizing the Model Confidence Set approach repeatedly, we rank the best VIXs across different dynamic models. Our results highlight the importance of estimating with VIXs and show that with the appropriate VIX a reduction of up to 38% in option pricing errors can be obtained. Our results also show that the 1-year VIX is the worst to use, that the 1-month VIX is an overall favourite, and that the choice of VIX maturity matters mostly for more flexible models. • We examine which VIX maturity to use in affine GARCH model estimation. • Our results highlight the importance of estimating with VIXs. • With the appropriate VIX a reduction of up to 38% in option pricing errors can be obtained. • The 1-year VIX is the worst to use and the 1-month VIX is an overall favourite. • The choice of VIX maturity matters mostly for more flexible models.