This paper investigates hedging performance of misspecified models. Rather than constructing super-replicating strategies this paper suggests quantifying the hedging error by introducing another contingent claim that covers any losses from using the misspecified model. This suggestion is motivated by the high costs of super-replicating strategies and the market practice of calibrating models to observed market prices. Numerical tests of using the Black-Scholes model for hedging in an economy driven by the Constant-Elasticity-of-Variance model show that the value of this claim is very low even for substantially different parameters.
By Nicki Søndergaard Rasmussen