外溢效果、動態波動和預測：指數交易型 債卷(ETN)實證分析

The Spillover Effects, Volatility Dynamics and Forecasting: Evidence from Exchange-traded Notes

狄強(Francis Diaz)

指導教授 ： 陳若暉

中原大學/商學院/商學博士學位學程/博士(2012年)

https://doi.org/10.6840/cycu201200133

摘要

本篇論文探討外溢效果、動態波動率和非線性效應，預測一個較新穎的投資工具，Exchange-traded Notes (ETNs)。本文含四篇實證論文，利用始於2006年且交易頻繁的24種ETNs資料為主。第一篇論文探討七種權益型ETNs之外溢效果和槓桿效應，應用ARMA-GARCH和ARMA-EGARCH之報酬與波動性和追蹤股票指數。本研究發現兩種財務工具的單邊與雙邊關係，各自有正面與負面的影響。第二篇利用MGARCH模型和強韌性分析(robustness check)來分析七種商品型的ETN和對應的ETN期貨之報酬率。實證結果顯示ETN和對應的ETN期貨具有長期的持續性和動態的相關性。實證結果也發現落後期的共變異數和落後期的市場衝擊(cross-product shocks)是ETN和對應ETN的函數。第三篇利用ARFIMA-FIGARCH模型且發現貨幣型的ETN報酬具有不穩定和非可逆性(non-invertibility)的特性。實證結果中也發現貨幣型的ETN報酬具有明確的混沌現象。其中的DBS檢定，RS檢定和相關維度分析(CD)是利用ARMA和GARCH模型的殘差來分析。第四篇利用灰關聯分析，並且發現貨幣型ETN以波動率指數在報酬率上提供最好的效果，另美元指數亦提供指數價值最好的效果。本研究也利用四種類神經網路模型分析，發現週期性神經網路模型(RNN)相較於其它三種模型(倒傳遞類神經，時間遞延神經網路和幅狀基底函數網路)具有最佳的預測準確性。

關鍵字

Exchange-traded Notes (ETNs) ； 外溢與槓桿效果 ； 動態波動率 ； 長期記憶特

並列摘要

This dissertation examines the spillover effects, volatility dynamics, nonlinearities and predictability of a relatively new investment instrument, exchange-traded notes (ETNs). The empirical database utilized a total of 24 actively traded ETNs for the four essays starting from their inception on 2006. The first essay uses the Autoregressive Moving Average-General Autoregressive Conditional Heteroskedasticity (ARMA-GARCH) and Autoregressive Moving Average-Exponential General Autoregressive Conditional Heteroskedasticity (ARMA-EGARCH) in studying the spillover and leverage effects of returns and volatilities of seven equity ETNs and their tracked stock indices. This study finds unilateral and bilateral relationships between the two financial instruments, with each also having positive and negative influences. The second paper utilizes three Multivariate General Autoregressive Conditional Heteroskedasticity (MGARCH) models, and a robustness check to analyze seven commodity ETNs with their corresponding futures contract returns. The research showed the presence of long-run persistence and dynamic correlations between the two financial instruments. The study also found that ETNs and futures contracts returns are functions of their lagged covariances and lagged cross-product shocks. The third study uses autoregressive fractionally integrated moving average-fractionally integrated generalized autoregressive conditional heteroskedasticity (ARFIMA-FIGARCH) and found that currency ETN returns have non-stationarity and non-invertibility properties. The paper also discovered deterministic chaos tendencies of currency ETN returns, ARMA and GARCH residuals through the Brock, Dechert, and Scheinkman (BDS) test, Reschaled Range (R/S) test and Correlation Dimension (CD) analysis. The fourth essay utilizes the grey relational analysis (GRA) and found that the volatility index and the US dollar index provide the greatest effect on returns and index values of currency ETNs, respectively. The research also applies four types of artificial neural networks (ANN) model and showed that the recurrent neural network (RNN) model produces the highest forecasting accuracy compared to back-propagation perceptron (BPN), time-delay recurrent neural network (TDRNN) and radial basis function neural network (RBFNN) models.

並列關鍵字

Exchange-traded Notes (ETNs) ； Spillover and Leverage Effects ； Volatility

參考文獻

Tsai, R.Y. (2008) Investigating the relationships for CRB index with stock price, exchange

Background References:

Chakravarty, R. and Praveen, D. (2010) Exchange-traded currency derivatives markets in

Chen, J. and Huang, C. (2010) An analysis of the spillover effects of exchange traded funds’, Applied Economics, vol. 42, no. 9, pp. 1155-68.

Deville, L. (2008) Exchange traded funds: History, trading and research, Handbook of