中文摘要：

目前我国外汇储备配置的研究大多集中在外汇储备规模和基于固定收益资产的币种配置上．本文认为，应首先区分不同资产状态，以此为前提再对外汇储备进行币种配置．采用均值-方差-偏度．峰度（MVSK）模型，基于外币现金、外国国债和外国股票三种资产类型进行最优币种配置研究．最优结果显示，三类美元资产存在一定差异，但美国国债和美国股票的比重基本超过50％．为排除人民币钉住美元的特殊汇率政策和可能的政治因素的影响，用同样的方法分别对日本和瑞士的外汇储备进行优化，也得到类似结果．分析可知，虽然当前美元货币不断贬值，但是相对于其它国家的金融资产，美国国债和美国股票资产安全性好，因此它在组合中的比重仍然很大．

英文摘要：

Most of the existing literatures focus on the scale of foreign exchange reserves and on the currency diversification of fixed income portfolio. We think it is necessary to distinguish the different types of the international assets before the process of currency diversification. This paper introduces a mean- variance-Skewness-Kurtosis （MVSK） optimization framework to manage the portfolios of foreign currency, foreign government bound and foreign stock respectively. We find that although difference does exist, the weights of both U.S. government bound and U.S. stock excess 50%. In order to expel the impact caused by the policy of pegging the Renminbi to the U.S. dollar and the potential political factors, this paper does robust tests by examining the portfolios of Japan and Switzerland with the same method, and arrives at similar results. This paper provides an explanation to the apparent paradox between the U.S. dollar depreciation and the significant weights of U.S. government bound and U.S. stock： the big weight is consistent with U.S. dollar assets＇ lower volatility compared with other counties＇ assets, which guarantees security of foreign reserves.