余额宝收益率的影响因素分析及预测研究

学校代码： 10286 分类号： F830 密 级： 公开 U D C ： 336 学 号： 131841 余额宝收益率的影响因素分析与预测研究研究生姓名： 白洁 导 师 姓 名： 何建敏申请学位类别 管理学硕士 学位授予单位 东 南 大 学 一级学科名称 管理科学与工程 论文答辩日期 2016 年 1 月 16 日 二级学科名称 金融工程 学位授予日期 20 年 月 日 答辩委员会主席 教授 评 阅 人 教授 教授 2016 年 1 月 16 日硕 士 学 位 论 文余额宝收益率的影响因素分析与预测研究专 业 名 称： 管理科学与工程 研 究 生 姓 名 ： 白洁 导 师 姓 名 ： 何建敏 教授 摘要IResearches upon the Factors of Yu-Ebao Returns and Its PredictionA Thesis Submitted toSoutheast UniversityFor the Academic Degree of Master of ManagementBYBai JieSupervised byProfessor HE Jian-minSchool of Economics and Management Southeast UniversityJanuary 2016摘要II东 南 大 学 学 位 论 文 独 创 性 声 明本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成果。尽我所知，除了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得东南大学或其它教育机构的学位或证书而使用过的材料。与我一同工作的同志对本研究所做的任何贡献均已在论文中作了明确的说明并表示了谢意。研究生签名： 日期： 东 南 大 学 学 位 论 文 使 用 授 权 声 明东南大学、中国科学技术信息研究所、国家图书馆有权保留本人所送交学位论文的复印件和电子文档，可以采用影印、缩印或其他复制手段保存论文。本人电子文档的内容和纸质论文的内容相一致。除在保密期内的保密论文外，允许论文被查阅和借阅，可以公布（包括刊登）论文的全部或部分内容。论文的公布（包括刊登）授权东南大学研究生院办理。研究生签名： 导师签名： 日期： 摘要III摘要当前互联网金融的快速发展引起了社会各界的高度关注，余额宝作为它的一个创新产品，以高收益率、强灵活性和低门槛等优势开创了一个全民理财的新环境，有力地加快了利率市场化的步伐。以余额宝为代表的互联网金融产品正在革命性地改变着传统金融的面貌，尤其是在移动支付、资源平台、大数据和搜索引擎等方面，极大地改变了人们的商业习惯，给用户带来了无限惊喜，给商业银行造成了巨大影响。截至 2014 年 12 月底，余额宝规模达到5789 亿元，不仅成为我国规模最大的货币基金和公募基金，也是全球第 4 大货币基金。因此，对余额宝收益率的影响因素和预测分析进行研究具有重大的时代意义。本文提出了基于 EEMD 分解的余额宝收益率分析方法，分别是基于 EEMD-VAR 的余额宝收益率影响因素研究方法和基于 EEMD-GARCH 的余额宝收益率预测研究方法。首先，运用 EEMD 技术将余额宝收益率原始序列分解成若干个不同频率的分量（包括若干个本征模函数和一个剩余分量） ，并根据 t-检验重组成高频分量、低频分量和趋势分量，分别对应的是市场波动项、重大事件影响项和趋势项；然后,分别对引起这三大项变动的可能影响因素进行定性分析，并与导致余额宝收益率发生明显变化的实际事件相结合，对影响因素进行量化。随后,通过 VAR 模型对余额宝收益率与其各项影响因素之间的长期均衡关系和短期波动模式进行实证研究。结果表明：1)余额宝收益率与其影响因素间所构成的关系系统是稳定的；2) 银行间同业拆借利率和广义货币供应量对余额宝收益率的影响程度和方差贡献度最大，表明国内市场资金面和货币政策的松紧程度对余额宝收益率的变动起着重要作用；3)汇率水平和银行存贷比均与余额宝收益率呈负相关，但影响程度并不显著。由于余额宝收益率是非线性和非平稳的时间序列，所以精确预测余额宝收益率是一项有挑战性的工作。传统的统计学模型建立在数据是线性的假设之上，很难捕捉到隐藏在余额宝收益率序列中的非线性模式，通常不能得到精确的预测结果。GARCH 模型是专门针对金融数据量体定做的回归模型，特别适用于收益率波动性的分析和预测。但由于模型复杂程度约束和信息冗余度的影响，尤其是对于非平稳、非线性的余额宝收益率来说，单一预测方法往往仅能兼顾一些主要影响因素，使得预测精度有所下降。因此，针对余额宝收益率预测问题，本文提出一种基于 EEMD（集成经验模式分解）和 GARCH(广义自回归条件异方差)的非线性组合预测方法。该方法运用 EEMD 技术将余额宝收益率原始序列分解成若干个不同频率的分量，并重组成高频、低频和趋势分量，分别代表市场波动价格、重大事件价格、趋势价格；针对此三个序列，构建不同的 GARCH 模型分别进行预测，得到各序列预测值后进行加摘要IV总得到 EEMD-GARCH 的预测值，再与通过单一 GARCH 建模的余额宝收益率原始预测值进行比较分析；结果表明 EEMD-GARCH 预测效果比 GARCH 模型好。总体来说，本文另辟蹊径，提出 EEMD-VAR 结构来研究余额宝收益率的影响因素，结合定性定量分析，从数据实证的角度全面深入地剖析余额宝收益率的关键影响因素。这不仅可以很好地解释它高低波动的内在原因，可以为市场各参与者提供良好的投资决策参考，对其未来走势的判断和风险控制也具有重要的理论价值和实践意义。关键词：余额宝收益率；影响因素；EEMD-VAR；EEMD-GARCH 摘要VAbstractThe current rapid development of Internet financial has caused high attention from all sectors of community. Yu Ebao, as one of its innovative products, has created a new environment of universal financial with its high yield, strong flexibility, low threshold and other advantages, accelerating the pace of market-oriented interest rate greatly. Represented of Yu Ebao for Internet financial products is changing the face of traditional financial revolutionary, especially in mobile payment, resource platform, big data and search engines, etc., which has greatly changed peoples business practices, bringing infinite surprise to users and making great influence on commercial banks. Up to the end of December 2014, the scale of Yu Ebao has reached 578.9 billion yuan, not only becoming the largest monetary funds and public funds, but also the worlds fourth-largest monetary fund. Therefore, to study the influence factors and forecast analysis of Yu Ebao Returns is of great significance.This thesis proposed a different Yu Ebao correlation analysis method based on EEMD decomposition, which were the study of the influence factors of Yu Ebao Returns based on EEMD-VAR and the forecast study of Yu Ebao Returns based on EEMD-GARCH. First, the Yu Ebao Returns series were decomposed into several components (including a set of intrinsic mode functions and a Residual component) with different frequencies by EEMD, then recombining them into high frequency, low frequency and residual components according to the t-test, respectively corresponding to market fluctuation, major events and trend components. Next, qualitative analysis were made about the influence factors of the big three changes, combining Yu Ebao Returns with actual events causing significant changes and quantifying the influence factors. Then empirical research was made through VAR model on the long-term equilibrium relationships and short-term fluctuation situations between Yu Ebao Returns and its various affecting factors. The results showed that: 1) The system consisting of Yu Ebao returns and its influential factors is stable;2) The Interbank lending rates and the broad money supply have the largest influencing degree and contribution on the Yu Ebao Returns, showing that the tightness degree of domestic market financing level and monetary policy are the most important affecting factors of Yu Ebao Returns; 3) Exchange rates and the bank loan ratio are negatively correlated with Yu Ebao Returns, but the impact is not significant. It is challenging to predict Yu Ebao Returns because Yu Ebao Returns series are non-linear and non-stationary. The traditional statistical models are built on the linear assumption, so it is difficult to capture the non-linear model hidden in the Yu Ebao Returns sequence, thus it is difficult 摘要VIto get the accurate prediction of Yu Ebao Returns. GARCH model was the regression model created specifically for the financial data quantity body custom, particularly applicable to the volatility analysis and forecasting of different kinds of returns. But because of the constraint of model complexity and the impact of information redundancy, especially for non-stationary and non-linear Yu Ebao Returns, a single forecasting method often can only focus on some main influence factors, thus making the prediction accuracy drop down. Therefore, in view of the prediction problem of Yu Ebao Returns, this paper proposes a nonlinear combination prediction method to combine EEMD (Ensemble Empirical Mode Decomposition) and GARCH (generalized autoregressive conditional heteroscedasticity).This method applies EEMD to decompose the Yu Ebao Returns series into several components with different frequencies, which are then composed into three subseries according to their frequency level, namely high frequency, low frequency and the trend frequency. The three new sequences represent the market fluctuation price, major event price and trend price respectively. Different GARCH models are constructed for the three new sequences to predict the final value of each sequence respectively, and the final predictive value of EEMD-GARCH can be obtained via the aggregation of the three predict values above. At last, by comparing and analyzing the EEMD-GARCH value with the single GARCH model value, the result shows that EEMD-GARCH method has higher prediction accuracy than single GARCH model.In general, this paper put forward a different approach-the EEMD-VAR structure to study the influence factors of Yu Ebao Returns, combining with the qualitative analysis and quantitative comprehensive. Then decompose the key factors affecting Yu Ebao Returns in-depth from the perspective of empirical data analyses. This can not only explain the reasons causing the high and low fluctuation of Yu Ebao Returns and provide good investment decision-making reference for the market participants, but also has important theoretical value and practical significance for the judgment of its future trend and risk control. Key words: Yu Ebao Returns; Influencing Factors; EEMD-VAR; EEMD-GARCH 目 录VII目 录摘要 .III目 录 .VII第一章 绪论 .91.1