Processos estocásticos dos preços das commodities: Uma abordagem através do filtro de partículas

Abstract

The spot prices of a commodity can be decomposed into two stochastic factors, the long term equilibrium price levels and the short term deviation in prices, which represents temporary changes. These two factors are not observable in the market. We consider that the long term equilibrium price follows a geometric Brownian motion and the short term deviations in prices follow an Ornstein-Uhlenbeck process with Poisson jumps. It is our objective to test this hypothesis. The addition of a Poisson process makes the model non-Gaussian. The Kalman filter methodology can not be used here. But there exists another methodology known as Particle filter method, which is appropriate in our case. We have used it and found that the model with jump process explains better than the model without it.