Abstract:
A Bayesian approach to option pricing is presented, in which posterior inference about the underlying returns process is conducted implicitly, via observed option prices. A range of models which allow for conditional leptokurtosis, skewness and time-varying volatility in returns, are considered, with posterior parameter distributions and model probabilities backed out from the option prices. Fit, predictive and hedging densities associated with the different models are produced. Models are ranked according to several criteria, including their ability to fit observed option prices, predict future option prices and minimize hedging errors. In addition to model-specific results, averaged predictive and hedging densities are produced, the weights used in the averaging process being the posterior model probabilities. The method is applied to option price data on the S&P500 stock index. Whilst the results provide some support for the Black-Scholes model, no one model dominates according to all criteria considered.