| Summary: | 碩士 === 國立臺灣科技大學 === 財務金融研究所 === 95 === In this paper, the first expounded on gas emissions trading origin and theoretical basis, and explore abroad toxic gas emission control system has evolved, and major global gas emissions related financial products market, Finally, the right to collect gas emissions related to the evaluation of financial products related literature, and the Kyoto Protocol commitment to the first phase (pilot stage) European GHG emissions allowance EUAs(European Union Allowances, EUAs) three major spot markets Powernext, EEX and Nordpool to empirical study to discuss which model that in-the-sample fitted ability, and the out-of-sample predictive ability performs better. The assumption returns of EUAs obeyed the ARMA mean equation models, and combined with GARCH, GJR-GARCH, EGARCH three GARCH conditional variance model, with the assumption of GARCH residual in Gaussian distribution or Student's t distribution. Frist, respectively analysis the EUAs three major spot markets return sequences, how many lag stages of the ARMA model is the optimal formula, then, respectively mix of three different GARCH models and different distribution of GARCH models’s residual to assess the best lag stages of models, and to verify in the EUAs three major spot markets return sequences, what kind of model fitted ability of in-the-sample data is more suitable than others, finally using the “rolling window” method comparison under different forecasting horizons, discusses EUAs out-of-sample returns data, what kind of model predictive ability of out-of-sample is more appropriate than others. The empirical results show that the EUAs three main spot market price return sequence, GJR-GARCH and EGARCH condition variance model no matter in-the-sample fitted ability, or the out-of-sample predictive ability are relatively better than GARCH model, this may be attributed that the EUAs three main spot market price return sequence, the existence of a conditional asymmetric volatility assumption. In addition, this study also found that the EUAs three main spot market price return sequence, in-the-sample fitted ability, the assumption of GARCH residual in student's t distribution is more suitable than Gaussian distribution, but the out-of-sample predictive ability, the assumption of GARCH residual in Gaussian distribution is more appropriate than student's t distribution , therefore, in this study shows that the model fitted ability with the predictive ability is not necessarily consistent results.
|
|---|
