Three factor stochastic volatility model estimation and inference : forecasting fish pool conditional moments ; determinants of the conditional volatility

The paper applies the General Scientific Model methodology of Gallant and McCulloch implementing MCMC simulation methodologies to build a multifactor stochastic volatility model for the latent (and unobservable) volatility of Fish Pool financial front month contracts. The main objective is to find appropriate descriptions (stochastic processes) emphasizing schemes for derivative pricing purposes without the assumption of constant volatility. Based on these estimates and inference, prediction under the MCMC framework can be done for both the mean and volatility. The scientific model building approach used here can therefore produce useful and superior volatility and correlation updating schemes. The paper reports Fish Pool conditional one-step-ahead (density) moments, particle filtering for one-step-ahead conditional standard deviations, conditional variance functions for evaluation of shocks, analysis of multi-step-ahead dynamics, and conditional moment’s persistence. The analysis adds insight and enables forecasts to be made, thus building up the methodology for developing valid scientific models for commodity markets.