Estudio de la volatilidad estocástica con parámetros variables en el tiempo: una aplicación al precio del crudo Oriente del Ecuador

Abstract

The present work exposes the stochastic volatility in mean model with time-varying parameters (SVM-TVP), to model Oriente crude oil price from Ecuador. An approach based on Bayesian inference is presented to estimate the model and Markov chains Monte Carlo (MCMC) methods are used, such as Metropolis-Hasting (MH) algorithm, the Gibbs sampling and the acceptance-rejection Metropolis-Hasting (ARMH) algorithm to simulate the a posteriori distribution of the parameters of the model. An algorithm is proposed, which is based on the Gibbs sampling, to estimate the marginal likelihood function of the SVM-TVP model. In addition, the performance of the SVM-TVP model is compared with the other conditional volatility models such as: GARCH, GARCH-M, GARCH-J, GARCH-2 and their counterparts with stochastic volatility, using: marginal likelihood, the bayesian factor, Akaike information criterion (AIC) and Schwarz information criterion (BIC).