中国是世界上最大的棉花生产和消费国，棉花产量和消费量分别占全球的近1/3。1998年以来，中国棉花市场和价格逐渐开始开放，随着2004年6月棉花期货在郑州商品交易所上市，棉花价格日益成为全社会关注的热点之一，中国的棉花价格不仅影响到国内棉花生产，棉农收入及涉棉企业的发展，对国际棉花价格也产生着举足轻重的影响。对高频金融时间序列波动特征进行系统深入的理论分析与应用研究是数量经济和系统科学学科的一个重要研究领域。中国棉花期货市场建立时间短，棉花价格制定、预测发展还不是很成熟，而纽约棉花期货市场建立于1870年，其规避风险和发现价格的功能充分发挥出来，已成为美国政府制定有关棉花政策主要参考依据，也是除中国以外其他主要产棉国棉农和涉棉企业套期保值的主要场所。因此本文重点研究美国纽约棉花期货市场的价格，探索其价格形成的内在机制，在此基础上，研究中美两国棉花期货价格的相关性，希望能对中国棉花价格的预测提供参考。本论文以1973年至2006年的纽约棉花期货价格序列为研究对象，应用ARFIMA,GARCH模型，协整理论等数量经济理论和方法，并应用EViews5.1，Matlab等软件的模拟计算功能，对纽约棉花期货价格的长记忆性，波动特征以及与中国棉花期货价格的关系进行了实证研究。本文首先检验了纽约棉花期货价格的统计特征，发现其分布均与正态分布有显著差异，有明显的高峰厚尾现象；在此基础上对其记忆性进行了检验，发现其存在长期记忆性；接着，利用ARFIMA模型对长记忆性特征进行了拟合，拟合后对残差的ARCH检验表明具有异方差性，再应用GARCH模型对残差进行了分析建模，捕捉到了波动的时变性。然后，以中国棉花进出口量为依据，将纽约棉花期货价格分段，利用ARIMA，GARCH模型对不同时间段棉花期货价格的考察，发现：1）虽然整体来看纽约棉花期货市场存在长记忆性，但分段进行单位根检验均认为棉花期货价格存在单位根，为I(1)过程；2）不同时段的棉花期货价格显示存在着异方差现象。最后，对郑州棉花期货收盘价格和纽约棉花期货收盘价格进行了协整检验，在说明郑州棉花期货价格的变动是纽约棉花期货价格变动的Granger原因的基础上（这一结果表明中国棉花市场对世界棉花市场有重要的影响，并与现在的普遍认识是一致的），得到郑州棉花期货价格的预测模型。

As the biggest cotton producing and consuming country, China takes almost 1/3 of the world cotton market. Since 1998 China cotton market starting opening gradually, along with the cotton future went public in Zhengzhou commercial trade market in 2006, cotton price is becoming one of the social concerns.One of the important research area of econometric and system science is the further academic and empirical research. Since China cotton future market is still developing while New York cotton future market was built up in 1870 and is well developed, this paper is focusing on New York cotton price close price to get some useful conclusion and suggestion to model China cotton future price.Data source is NYBOT, from 1973 to 2006, fist study the statistic characteristics of the daily cotton future price, find the price distribution is not normal, it has fat tails and a higher kurtosis, and also there is long memory in the cotton future market, so choose ARFIMA model to estimated the long memory and GARCH model to analysis the heteroskedasticity of residual. Then according to China cotton amount of import and export, the series is divided to 6 sections to analysis the character of each period. Then we find that: 1) though the whole cotton price has a long memory but for each period, they all have a unit root and is I(1) process; 2) Different period may have heteroskedasticity characteristics individually. At last, do the co-integration analysis of Zhengzhou and New York cotton future price and find that the change of Zhengzhou cotton price is the Granger cause of New York cotton price, base on the model give the forecast price.