中文摘要：

为了准确描述基础资产波动率的＂聚集性＂、＂杠杆效应＂及时变特征,将有偏GARCH模型引入最小二乘蒙特卡罗美式期权定价（LSM）方法中,同时使用Lévy过程修正GARCH模型中的随机数分布,进而表现金融数据的非正态特征和＂跳跃＂特征,建立了Lévy-GARCH-LSM模型。基于我国13只美式权证、百慕大权证,和S＆P 100指数的美式期权的实证研究,发现：（1）Lévy-GARCH模型不仅能精确地描述基础资产的统计特征,而且还提供了较为准确的美式权证定价结果;（2）引入Lévy过程和GARCH模型都能提高权证定价精度,而各类非对称GARCH模型之间没有显著差异;（3）由于我国权证市场交易制度不完善、投机成分严重,与S＆P100美式期权的误差相比,基于无套利假设下的Lévy-GARCH模型用以定价我国权证仍存在较大误差。

英文摘要：

This paper introduces a sort of non-Gaussian and asymmetric GARCH model into the Least Squares Monte Carlo American option pricing approach （LSM）, to accurately describe ＂volatility clustering＂, ＂leverage effect＂ and ＂time varying states＂ styles for underlying assets. We use Levy processes to modify the innovation＇s distribution in GARCH dynamics to capture the non-normality and jump behaviors in financial data, and then develop a Levy-GARCH-LSM model for American option pricing. Based on our empirical research, including 13 American warrants in Chinese market and American options of S＆P 100 index, we find： （1） L6vy-GARCH models not only perfectly describe the statistical properties of underlying assets, but also provide more accurate pricing results; （2） Levy processes and GARCH dynamics can improve the ability of warrant pricing, but there is no significant difference between these asymmetric GARCH dynamics; （3） due to the imperfect transaction mechanism and over-speculation in Chinese warrants market, the pricing errors based on the non-arbitrage L6vy-GARCH models for Chinese market are much larger than that for S＆P100 American options.