Search

This study tests the ability of three one-factor, and three two-factor models in the HJM class, to explain the pricing of the Eurodollar futures options on the Chicago Mercantile Exchange from 1 January 96 to 31 December 98. The work extends the paper of Amin and Morton (1994) by introducing two-factor models, which have the advantages over the one-factor models for the yields of different maturities being imperfectly correlated. On each day, about 10.7 calls and 9.9 puts with different categories of moneyness and maturity are collected to minimise the difference of market and model prices for determining volatility parameters for each function. Three criteria are imposed for evaluating the models: (1) the ability of each model to fit option transaction prices; (2) the predictability of option prices for each model; (3) the profitability of hedging strategies for each model. The results show that the models with a combination of different sensitive volatility parameters provide smaller fitting and prediction errors than the models with a simple volatility parameter. However, volatility parameters in these models are less stable and highly correlated. We also documented that ITM options are overpriced relative to at- and OTM options. in addition, the degree of overpriced options is an increasing function of maturity. In general, two-factor models fit better for long-lived options than one-factor models, whereas there is no significant difference for both types of models for fitting shortlived options. Based on out-of-the-sample approach, those models, which have higher explanatory power in in-sample fit usually perform better, indicating volatility parameters are not unstable over time. Finally, two hedging strategies, buy-and-hold and rebalanced, are implemented to reap the findings of mispriced options. On average, the models with an exponential factor produce negative profits, as a result of overfitting the long-term options. Although this group of models fits significantly better than others, those models with flexible parameters have resulted in overfitting the biases. Although lower prediction errors are found for these models, absolute and square root models are still preferable since they can detect mispriced options by making more profits. Of these two trading strategies, the buy-and-hold strategy generates more profits for all models in most circumstances.