在封閉解GARCH選擇權模型下，報酬與變異數序列是否有相同的驅動因子?

Abstract

本研究檢驗在封閉解GARCH選擇權模型中的假設下，報酬與變異數序列的關係。論文中設定多方假設檢定來探求兩數列間的因果關係與檢定Heston & Nandi (2000)GARCH選擇權模型中的假設在實證上之結果。各國間指數的實證結果顯示報酬與變異數並未存在同期的完全正相關的特性，亦即同期關係。實證顯示存在單一方向關係，即報酬領先變異數，或者回饋關係，兩數列受到過去資訊交互影響。不同的GARCH模型有相同的結果。此結論可解釋為何HN模型在某些選擇權交易的應用上，如避險行為上，有較差的表現。

This paper examines the dynamic relations between return and volatility series under the assumption of closed form GARCH model. A multiple hypotheses testing method is employed to identify causal relations between the two series and to test the empirical implication of the assumption of Heston and Nandi (2000) on GARCH option pricing model. The international empirical results show that returns and volatility series do not perfectly correlated instantaneously, that is contemporaneous relation. There exists unidirectional, return lead volatility, or feedback relation; two series are cross-correlated by past information. Different GARCH models also have the same result. It is found that return leads volatility. This result help explain why HN model has inferior performance in some option application, such as hedging.