An empirical study on model risk

This dissertation focuses on some important issues regarding model risk. The first part examines the relative performance of the Black-Scholes model and some of its alternatives in hedging exotic equity options. We test these models in a procedure consistent with the way practitioners use them. To overcome the difficulty of lack of data for exotic options, we look at the difference between model price and the value of replicating portfolio for the hedged options rather than pricing errors. The results show that the alternative option pricing models outperform the Black-Scholes model in hedging exotic options that are close to vanilla ones, such as short-term barrier options and compound options. However, they do not necessarily outperform the Black-Scholes model in hedging options with severe exotic features, such as long-term barrier options. Our results also indicate that model performance depends on the degree of path dependence of the option under consideration.; The second part of the study examines the practical compatibility of the LIBOR and the swap market models when they are used for hedging purposes. These models are calibrated to the cross sectional Black implied volatilities for caps and swaptions respectively, and the test is based on their effectiveness in hedging floors and swaptions that are not used in the calibration. We find that the LIBOR market models outperform the swap market models in hedging floors but perform slightly poorer than the swap market models in hedging swaptions. Our results also show that incorporating a humped volatility structure into these models does not significantly improve their hedging performance.