Empirical Analysis Of The Volatility Of Copper Futures Price With Markov Switching GARCH Models

With the global economic integration and the rapid development of Chinese futures market, the impact of risks from international financial market is becoming significant. Analysis of the volatility characteristics of futures price accurately is the fundament of recognizing market risk and so enhancing the market efficiency. In recent years, diverse researches, focusing on the financial time series' volatility, has been carried out with different GARCH models. Nevertheless, as the coefficients remain constant, conventional GARCH models are unable to capture the non-linear feature of a nonstationary series. For this reason, the results may not be accurate or rigorous. Markov switching model, proposed by Hamilton (1989), provides a new approach to describe financial time series which contains regime changes.This paper introduces state variables into traditional GARCH models, and establishes Markov-Swithing GARCH models for analysing the volatility of Shanghai copper futures. The data set analyzed in this paper is the three-month consecutive price data with the sample period from April 1995 to April 2011. After some basic statistical tests, non-linear tests and regime-swithcing tests, this paper presents three types of GARCH models, for standard GARCH model, Exponential GARCH model, and Threshold GARCH model, respectively. Based on this, Markov-Switching GARCH models are putted forward as a contrast with those linear models. Besides, some fat-tailed distuributions, to be specific, the Students't and the Generalized Error Distribution (GED), are applied in this paper as complements for classic gaussian assumption.The greatest innovation of this thesis lies in the applying Markov-Switching GARCH models into futures volatility reseach originally, and by comaring the fitting results of different models, it is able to lay the foundation for the prediction of rate of return of futures and also quantitative analysis for risk identification in Chinese futures market.According to the result of parameter estimation and goodness-of fit statistics, this paper shows that there is significant GARCH effect in the rate of return of Shanghai copper futures, while leverage effect is statistically nonsignificant. In Markov-Switching GARCH models, parameters are allowed to switch between a high and a low volatility state and the result of parameter estimation varies evidently between the two. Compared with low volatility state, the duration of high volatility state is shorter, and the volatility persistence is much lower. The empirical analysis illustrates that MS-GARCH models outperforms most conventional GARCH models in capturing characteristics of futures volatility.