Estimation des changements des cours du café et du cacao: Filtre de Kalman, filtre de Hodrick-Prescott et modélisation à partir de processus markovien

This aim of this study is to estimate the price of coffee and cocoa using a methodology based on Hodrick-Prescott filter, Kalman filter and a Markov Switching Model which, unlike linear models, allows the parameters to vary depending on the economic situation, the transitions between regimes are governed by a Markov chain. Our results show that the Hodrick-Prescott filter gives only the general trend of the series while the Kalman filter approach the best real trend of the series. The Markov Switching Model breaks the series in the regimes. Thus, this latter model gives the probability of transition from one regime to another and then gives the average length of stay in a particular regime.