山西大学2014年统计会议

大会邀请报告

发布时间：2014-08-01 05:09:00 发布者：阅读次数：

大会邀请报告一： Topic: Liquidity risk and asset pricing: Evidence from daily data, 1926–2009 Reporter：Prof. Weimin Liu，University of Nottingham ，UK 刘卫民英国诺丁汉大学
Abstract：There exist many proxies for liquidity such as transaction costs based (such as Amihud and Mendelson, 1986; Lesmond, Ogden, and Trzcinka, 1999; Hasbrouck, 2009; and Corwin and Schultz, 2012), price impact based (such as Amihud, 2002; Pastor and Stambaugh, 2003; and Sadka, 2006), trading quantity based (such as Datar, Naik, and Radcliffe, 1998; and Brennan, Chordia, and Subrahmanyam, 1998) and trading speed based (such as Liu, 2006). It is commonly recognised that illiquid securities expose to high liquidity risk because of the difficulty to convert them to cash, especially in bad times when investors’ demand for cash is high. Given the wisdom of the risk—return relation, a good liquidity measure should predict return. Accordingly, for asset pricing, which tries to reveal the relation between risk and expected return, it is a natural task to ascertain which measure(s) of liquidity better predict return. The first part of the paper performs this task by examining eight liquidity measures and identifies two best return predictors over the 1926-2009 period: turnover and the trading discontinuation proxy. Compared with the size and value premia, the liquidity premium is more significant and robust. 　There are a number of asset pricing models developed in the literature: the capital asset pricing model (CAPM) of Sharpe (1964), Lintner (1965), and Mossin (1966); the intertemporal CAPM (ICAPM) of Merton (1973); the arbitrage pricing theory (APT) of Ross (1976); the consumption based CAPM (CCAPM) of Lucas (1978) and Breeden (1979); the three-factor model of Fama and French (1993) (FF3FM); the liquidity augmented CAPM (LCAPM) of Liu (2006); etc. Among them, the CAPM is the earliest and most commonly used in practice such far, whereas academic research has heavily relied on the FF3FM nowadays. However, a large number of pricing anomalies challenge the CAPM to explain cross-sectional returns. The FF3FM also shows limited power to describe average returns and subsequently some studies attempt to incorporate more factors such as the momentum factor of Carhart (1997) and the profitability factor of Fama and French (2014). In addition, both the ICAPM and the APT are multi-factor models, but their derivations do not specify what these factors are. In practice, different researchers have proposed and empirically investigated different factors. This paper shows that the state nature of liquidity is robust to the pre-1963 period over when there is little research on liquidity risk based pricing. Large fluctuations in market liquidity correspond to economic conditions. Conditional on economic and liquidity states, liquid stocks apparently more highly priced than illiquid stocks in bad states, implying flight to liquidity. The implication naturally arises from which investors are unwilling to invest during economic downturns, when consumption is low and investment opportunities are poor, unless they can expect a higher return. Consistently, we find that investors indeed require higher expected returns for holding assets that are more sensitive to fluctuations in market liquidity, and the LCAPM provides a good description of expected returns. 　This paper also assesses the marginal effect of liquidity risk versus liquidity level. The assessment has direct implications for performance evaluation, portfolio analysis, and empirical research. If liquidity level rather than sensitivity to liquidity variations predicts cross-sectional returns, then including a liquidity factor in an asset pricing model is gratuitous. In contrast, liquidity risk is important for asset pricing if liquidity risk subsumes liquidity level in explaining the cross-section of security returns. The results suggest that liquidity level lacks significant power to predict stock returns beyond liquidity risk. Weimin Liu has been undertaking a secondment to the University of Nottingham Ningbo China (UNNC) campus since September 2012. His research interests include asset pricing, efficient market hypothesis tests, corporate finance, and mutual fund industry. He has published in leading international finance journals such as the Journal of Financial Economics (JFE), the Review of Financial Studies (RFS), the Journal of Banking and Finance (JBF), the Journal of Business Finance and Accounting (JBFA), and the European Financial Management (EFM).
大会邀请报告二： Topic: Timings, Terms and Agency Problems in Mergers and Acquisitions Reporter：Prof. Peijun Guo，Yokohama National University ，JPN 郭沛俊日本横滨国立大学
Abstract：There are a great number of researches about the motives for M&A. It has been summarized that the theories of merger motives could be classified into seven groups: efficiency theory, monopoly theory, raider theory, valuation theory, empire-building theory, process theory and disturbance theory. We focus on the operational synergies in the framework of efficiency theory. In this research, the gains of M&A not only result from economies of scale, but also products consolidation. On the one hand, because production technology creates economies of scale, the post-merger firm can produce more products than the total products that they can produce before merger. On the other hand, through sharing technologies, patents and brands, the post-merger firm is willing to select only one type of initial product with a relatively high profit to produce. We continue a line of research using real option models to analyze mergers and acquisitions. In this paper, we focus on the product market instead of the stock markets which have been taken into account in the existing literature. Lambrecht (2004) also considered the product market. However, in his model, the bidder, the target and the post-merger firm produce the same product; the synergy gain is only generated from economies of scale. In our model, we suppose the two firms produce two types of products. Such assumption can more appropriately characterize a real world situation. After merger, the two firms consolidate their products and produce one of their two initial products so that the synergy gains result not only from economies of scale, but also products consolidation. 　Also, we study agency problems between managers and outside shareholders. We follow the ideas of Myers (2000) and Lambrecht (2007) by assuming that the total value of the firm is shared by the managers and the outside shareholders, and they are both self-interested. However, we have different interests of research from theirs. Assuming that the managerial rents are a constant proportion of the firm values, we propose real option models to analyze several merger strategies of managers corresponding to the different situations where outside shareholders know different information on operational synergies. 　In this research, we analyze the timings, terms and agency problems in mergers and acquisitions where each firm optimally exercises its own exchange option and the timing and terms are determined endogenously. The theoretical analysis shows that the competition amongst the bidders will speed up merger and decrease the winning bidder's share in the post-merger firm; the agency problem will speed up or delay takeover. The managerial insights have been gained by the theoretical analysis. Dr. Peijun Guo is Professor of Decision Sciences, Faculty of Business Administration, Yokohama National University, Japan. Prof. Guo is Associate Editors of Information Sciences and the Editorial Board Members of several Journals.His research interests involve operation research and management science, mainly in decision analysis under uncertainty and uncertainty modeling. His papers also appear in European Journal of Operational Research, IEEE Transactions on SMC, Part A: Systems and Humans, Computational Statistics and Data Analysis, Information Sciences, Fuzzy Sets and Systems etc.
大会邀请报告三： Topic: 基于t-copula信用资产组合风险度量 Reporter：Prof. Rongda Chen，Zhejiang University of Finance and Economics 陈荣达浙江财经大学
摘要：对信用资产组合的风险度量研究一直以来都是学者关注的重点，而对于异质信用资产组合风险度量研究不多。本文研究一种能够刻画异质性信用资产组合尾部相关性的有效度量风险的方法，并对上市股票相关数据做实证分析。通过Black-shores Merton模型的求解得到了结构模型所需要的每支股票的时刻公司资产价值以及标准收益率，通过核密度估计得到每支股票的边际分布。利用t-copula系函数以及非线性估计我们分别得到了组合的相关结构以及异质资产组合的各因子系数，进而得到了组合的整体违约概率。再基于t-copula下的重要抽样技术，得到了组合的风险VaR值和ES值，结果表明重要抽样技术在保证精确度的前提下的确减少来了抽样的方差，同时引入ES值可以更好的进行风险监管，为极端事件的研究以及投资者配置经济资本和监管者进行风险监管提供了依据。陈荣达, 浙江财经大学金融学教授。主要从事金融工程和金融风险度量方面的学习与研究，主持国家自然基金项目2项，获浙江省哲学社会科学优秀成果三等奖1项、浙江省高校科研成果2项。在《Economic Modelling》、《Physica A》、《Annals of Operations Research》、《Mathematical Problems in Engineering》、《管理科学学报》、《系统工程理论与实践》、《数量经济技术经济研究》等期刊发表论文40余篇，其中SSCI、SCI收录7篇，EI收录12篇。出版专著2部。
大会邀请报告四： Topic: Time inconsistency, self-control and internal harmony: A planner-doer game framework Reporter：Prof. Xiangyu Cui，Shanghai University of Finance Economics 崔翔宇上海财经大学
Abstract：For time inconsistent multi-period mean-variance portfolio decision, we develop a two-tier planner-doer game model with self-control, in which planner and doers represent different interests of the same investor at different time instants. The planner (the willpower to resist short term temptations) can impact preferences of doers through commitment by punishment, while the applied total penalty in turn affects the planner’s preference. Dealing with time inconsistency is to achieve a degree of internal harmony (measured quantitatively by expected cost of self-control) through aligning interests of planner and doers. We further extend this game framework to general time inconsistent stochastic decision problems. 　In particular, we assume one planner and T doers in our game problem formulation to represent different facets of the same investor under different circumstances at different time instants. The doer at time t (called self t) represents the investor at time t who originally bows fully to short term temptation, and decides a carried out strategy for time period t. On the other hand, the planner in the model represents the willpower of the investor to resist such temptations and attempts to manipulate the behavior pattern of the doers through a self-control mechanism, “commitment by punishment”. More specifically, the planner decides a planned investment strategy before the investment, announces it to the T doers in advance and commits that any deviation by a doer from the planned investment strategy would cost the doer a type of punishment. 　In this planner-doer game model, due to the existence of commitment by punishment, the doer at time t modifies his attitude of full temptation and behaves to balance the investment performance of the truncated time horizon from time t to terminal time T against possible punishment applied to him by the planner. At the same time, the planner also softens his position in insisting the pre-committed policy and intends to balance the long term investment performance determined by the carried out strategies of T doers over the entire time horizon against the total punishments imposed on these T doers. Applying penalty to both doers and the planner aligns the interests of doers and the planner, which may further result in internal harmony between doers and the planner. 　On one hand, our formulation is a leader-follower game between the planner and T doers (termed as the upper tier game). On the other hand, due to the non-smoothness of the variance term, the objective function of self t is dependent on the carried out strategies of future selves in a non-separable manner. Thus, in addition to the upper tier game between planner and doers, our formulation also involves a lower tier sequential game among T doers. The coupling of these two games in the upper and lower tiers makes our planner-doer game model a two-tier game model. We derive the explicit equilibrium strategy of this two-tier planner-doer game model, which is termed as Doer-Perfect Nash equilibrium strategy. 崔翔宇，博士，上海财经大学统计与管理学院助教授。从事动态投资组合优化和行为金融学研究。主要研究方向为动态投资组合选择，风险管理，数量金融和行为金融。
大会邀请报告五： Topic: Gradient-Based Simulated Maximum Likelihood Estimation on Stochastic Volatility Models Reporter：Prof. Jianqiang Hu，Fudan University 胡建强复旦大学
Abstract：Parameter estimation and statistical inference pose challenges for stochastic volatility models, especially those driven by pure jump Levy processes. Maximum likelihood estimation (MLE) is preferred when a parametric statistical model is correctly specified, but traditional MLE implementation for stochastic volatility models is computationally impractical. 　We first study the parameter estimation problem for a special Levy-driven Ornstein-Uhlenbeck stochastic volatility model. Estimation is regarded as the principal challenge in applying these models since they were proposed by Barndorff-Nielsen and Shephard (2001). Most previous work has used a Bayesian paradigm, whereas we treat the problem in the framework of maximum likelihood estimation, applying gradient-based simulation optimization. For the Gamma Ornstein-Uhlenbeck stochastic volatility model, the classic perturbation analysis can be used to estimate the gradient of the log-likelihood. A hidden Markov model is introduced to formulate the likelihood of observations; sequential Monte Carlo is applied to sample the hidden states from the posterior distribution; smooth perturbation analysis is used to deal with the discontinuities introduced by jumps in estimating the gradient. 　To extend the application of the gradient-based simulated maximum likelihood estimation on more general stochastic volatility models, we study the maximum likelihood estimation for a generic stochastic volatility model which covers the pure jump Levy-driven stochastic models, using simulation and sensitivity analysis based on characteristic functions. Hilbert transform is applied to invert the characteristic functions to estimate the values of cumulative distribution functions at discrete points and use linear interpolation to approximate the cumulative distribution functions; for models where the random variable driving the transition function is continuous, we derive an infinitesimal perturbation analysis estimator for the log-likelihood using characteristic functions; for the case where the random variable driving the transition function has positive probability masses, smooth perturbation anlysis and the support independent unified likelihood ratio and infinitesimal perturbation analysis are applied to derive the gradient estimator for the log-likelihood using characteristic functions. 　Numerical experiments are presented to show the efficiency and wide applicability of the proposed method. We apply the gradient-based simulated maximum likelihood estimation to calibrate the Gamma Ornstein-Uhlenbeck stochastic volatility model using daily market data of S&P 500 and H&S 300 from 1st June, 2007 to 1st June, 2011, and compare different models based on BIC which uses MLE as input. 胡建强，复旦大学管理学院管理科学系教授，博士生导师。1996年任美国波士顿大学(Boston University)终身教授。他的主要科研方向是运筹学中各种随机系统的分析、优化和控制，研究内容包括仿真模拟、排队论、供应链和物流管理、电讯通信系统、金融市场及衍生产品等，十几年来共发表了80多篇专业论文。现任《运筹学学报》编辑，《Automatica》副主编，《Journal of the Operations Research Society of China》编辑，中国运筹学会金融工程及金融风险管理分会会长，中国运筹学会理事。
大会邀请报告六： Topic: Dynamic α–stable Copula for CDO pricing Reporter：Prof. Hua Li，Zhengzhou University 李华郑州大学
Abstract：Static factor Copula methods have been the most popular methods for CDO pricing in both the academia and the financial industry. However, it not only exibits the correlation smile phenomenon which is not in consistent with the model assumption, but also can not be used to price CDOs with the same underlying portfolio but different maturities. This paper studies the dynamic factor copula model for CDO pricing based on the-stable distributions, the two above mentioned problems can be overcome in the dynamic factor copula framework. 李华，郑州大学副教授，硕士生导师，郑州大学金融工程研究所所长。主要研究方向:衍生产品的设计定价和风险防范，信用风险，期货套利和套保策略，证券智能交易策略和算法实现，金融数学模型的数值方法等。参加完成国家自然科学基金2项，主持完成国盛证券企业孵化上市策略研究项目，目前主持交通银行资产管理项目和2项河南省科技厅基础前沿项目。
大会邀请报告七： Topic:基于社会管理的农户信用评级﹑信用贷款决策模型与系统研发 Reporter：Prof. Sulin Pang， Jinan University 庞素琳暨南大学
摘要：广东省云浮市郁南县于2009年在广东省率先试行“信用村、信用镇”，开辟农村信用贷款和金融扶贫的先河，以符合农民最基本特征的社会管理方法，利用金融扶贫、金融扶持做为政府、金融机构和农民之间沟通的桥梁和手段，通过农村诚信文化建设，大胆探索农村创新金融发展，解决了“三农”贷款难的问题，提升了农村社会管理水平。　　本报告创新性地将农户信用评级指标体系的基础架构建成一个具有多层级的单向网络结构。先根据三级信用评级指标体系的逻辑结构，构建了四层单向网络，给出了各级指标计算公式及信用评分计算公式；并由此进一步给出具有一级和二级农户信用评级指标体系信用评分公式的特例；然后在此基础上推广到具有四级以上信用评级指标体系，给出了具有四级以上信用评级指标体系的信用评分公式，由此最终得到从一级到多级的农户信用评级指标体系信用评分公式的通式。为了解决农户信用评级问题，还设计了一个线性分段分类器来将多层单向网络输出的结果进行信用等级分类，建立了农户信用评级的规则以及农户信用评级单向网络线性分段评价模型，讨论了基于农户信用评级的银行贷款授信所满足的性质。　　在农户信用评级基础上，研究基于银行贷款风险损失比的农户信用贷款决策模型以及银行相应的信用贷款利率机制. 通过以农户个体合理性和银行最大可接受的贷款风险损失比作为约束条件，分别在农户项目成功概率对银行期望收益的影响以及农户项目成功概率同时对银行期望收益和农户期望收益都产生影响两种不同情形的假设下，建立了相应的农户信用贷款决策模型，给出了两种不同的银行最优贷款利率机制。还举出实例，针对5级分类中农户不同的信用等级以及相应所获得的银行贷款授信，设计了4组不同的组合数据，分别在贷款申请金额、农户自有财富、银行最大可接受的贷款损失比以及农户项目期望收益率发生不同的变化时，农户最优项目成功概率以及银行最优贷款利率发生的变化。讨论了在银行不同的贷款利率下农户项目净期望收益率的变化以及农户项目期望收益率的合理区间. 　　最后将理论建立的模型应用到广东省云浮市郁南县农户信用评级中，当对272个农户进行信用评级时，得到了符合郁南县农户信用评级实际的评定结果，准确率达到100%。　　本研究还研发了《云浮市农户信用评级与信用查询系统》，实现了农户信息采集、农户信用查询、农户信用评级、关联部门查询等功能。　　庞素琳，暨南大学应急管理学院教授，博士生导师。研究领域包括：金融工程与风险管理、应急管理与应急技术、移动网络故障与定位、网络评价与预警技术、计算机软件开发等。在国内外重要学术期刊发表论文近100篇。出版学术专著3部。获得授权发明专利2项。研究成果获得广东省科学技术奖二等奖1项、广东省哲学社会科学优秀科研成果奖三等奖1项。主持国家、省（部）市级项目、政府和企业委托或招标项目等20多项。2008年入选教育部新世纪优秀人才支持计划。
大会邀请报告八： Topic: The Default Probability in the Credit Risk and its Computation under the Structural Model Reporter：Prof. Yongjin Wang, Nankai University 王永进南开大学
Abstract：In this talk, we study credit risks, following the structural framework. For this purpose, we first propose a stochastic model fitting for some specific asset price dynamics in the financial market. In our consideration, we introduce the so-called "regulated" asset price dynamics (with floor and ceiling), typically in the currency markets. We quantitatively model them by the "reflected" stochastic processes. Then we are going to discuss the default issues, which caused by the asset price fluctuations. On the other hand, the advanced probabilistic analysis indeed allowed us to explicitly compute the Probability of Default in the model. This theoretically and practically is fundamental for pricing the Credit Risks and the Credit Derivatives. This talk is a survey of recent joint works with the coauthors. 王永进，南开大学商学院金融和财务管理学教授、博士生导师，南开大学商学院副院长。同时双聘南开大学数学学院概率统计学教授、博士生导师。其主要研究领域包括：风险定价和金融期权，金融信用风险与信用衍生品；以及现代概率论，随机过程的理论与计算。迄今为止已在现代概率论、数量金融和金融工程等领域国际学术期刊上，发表学术研究论文50余篇。并先后主持教育部重大项目“金融信用风险的量化研究”以及系列国家自然科学基金项目等。
大会邀请报告九： Topic: 石油价格变化对金融市场的影响 Reporter：Prof.Chongfeng Wu, Shanghai Jiao Tong University 吴冲锋上海交通大学
摘要：本文主要分析油价冲击对金融市场的影响。首先分析油价冲击对世界各国股票市场的影响，然后运用原油市场信息预测短期和长期汇率变化，从样本外的角度分析油价冲击对汇率市场的影响。吴冲锋，上海交通大学安泰经济与管理学院金融学教授、博士生导师、教授委员会主任、金融工程研究中心主任、经济学院副院长等，是国内最早从事金融工程研究与教学的学者之一。 20多年来一直从事金融工程（资产定价、金融产品创新和金融风险管理）及其相关问题研究，曾先后主持过国家杰出青年科学基金项目、国家自然科学基金重点项目等以及上海期货交易所等科研项目30多项；曾获国家教委科技进步奖3项、中国高校人文社会科学研究优秀成果奖2项、上海市决策咨询建议奖2项、上海市哲学社会科学优秀成果奖3项；曾获得国家教委和国务院学位办授予“做出突出贡献的中国博士学位获得者”。先后在国内外重要期刊发表许多论文，其中被SSCI收录期刊论文超过30篇。此外，还先后担任中国金融学会常务理事兼金融工程专业委员会副主任，上海市金融工程研究会理事长，上海市金融学会常务理事，教育部高等学校金融学类教学指导委员会委员等。
大会邀请报告十： Topic: Efficient Monte Carlo valuation for portfolio risk Reporter：Prof. Chenglong Xu, Tongji University 徐承龙同济大学
Abstract：Monte Carlo simulation is widely used to measure the credit risk and systemic risk in portfolios of loans, corporate bonds, and other instruments subject to possible default. The accurate measurement of credit risk is often a rare-event simulation problem, because default probabilities are low. This makes importance sampling and other acceleration methods attractive. However the traditional acceleration methods, such as the control variate and importance sampling methods are often model dependent. This paper provides some new acceleration procedures for the measure of portfolio risk by the combination of algorithm(soft ware) and GPU(hard ware). 徐承龙，同济大学数学系风险管理研究所教授，博士生导师，兼任上海市计算科学E-研究院特聘研究员，Math Review 评论员。从事计算金融的研究，承担过国家自然科学基金和国家973重点项目子课题《信用风险分析和信用衍生产品定价》（主要研究人员）以及多项上海市科研基金项目。在国内外发表论文50多篇，教材7本。曾获得上海市优秀教学成果一等奖（排名第二），宝钢优秀教师奖。
大会邀请报告十一： Topic: 增长期权的消耗、创造与股票收益 Reporter：Prof. Yong Zeng，University of Electronic Science and Technology 曾勇电子科技大学
Abstract：Assets-in-place and growth options codetermine asset components, systematic risk and expected stock returns. In a pioneering study, Berk, Green and Naik (1999) develop a model in which the exercise of growth options is related to dynamic changes of relative importance of different asset components and the risk characteristics of asset. They provide a new explanation for Fama-French three factor model as well as important implications for the relation between investment and stock returns. Subsequent empirical literature provides evidences for the role of capital investment (the exercise of investment options) in explaining the cross section of future stock returns. However, the literature has paid less attention to the fact that the exercise of investment options (such as ordinary capital expenditure or growth in assets) and creation of growth options (such as R&D expenditure or strategic M&A) may have opposite effects on asset structure and risk characteristics. Therefore, aiming to differentiate the roles between the exercise and creation of growth options, this paper examines simultaneously how past ordinary capital expenditure and R&D expenditure affect future stock returns, and provides a consistent explanation from the perspective of rational investment and risk-based pricing. 　Using a sample of non-financial listed firms on Shanghai and Shenzhen Exchanges during 2007 to 2013, we define ordinary capital expenditures as the ratio of cash paid for the purchase and construction of fixed assets, intangible assets and other long term assets to total assets, R&D intensity as the ratio of R&D expenditures to total assets, and expected stock returns as monthly returns in subsequent fiscal year. Based on Fama-Macbeth cross-sectional regression, we find that ordinary capital expenditure and R&D expenditure have negative and positive effects on expected stock returns, respectively. The results are consistent with existing evidences in literature and the intuitive that R&D investment can help to create growth options while ordinary capital investment usually consumes growth options, and thus the systematic risk will decrease with ordinary capital investment and increase after R&D investment. Our results are robust when we use sub-samples in manufacturing or information technology industries and other measures of R&D intensity and ordinary capital expenditures standardized by sales, market capitalization and investment growth. 　With the fact that financial constraints have a great influence on firms’ investment decision, we extend to investigate how the relation between investment and stock returns is affected by financial constraints. Results show that financial constraints will strengthen significantly the positive R&D-returns relation, while we find no significant influence of financial constraints on the negative ordinary capital expenditures-returns relation. 曾勇，电子科技大学经济与管理学院教授、博士生导师、资本市场研究中心主任，国务院特殊津贴专家。多年从事金融工程、资本市场与公司金融等领域的研究，作为主持人承担了国家自然科学基金、教育部“新世纪优秀人才支持计划”、教育部高等学校博士学科点专项基金、教育部人文社会科学规划基金等资助的多项重要科研项目。研究成果曾获得国家科技进步奖三等奖、四川省科技进步奖一等奖和二等奖、电子工业部科技进步奖一等奖、中国高校人文社会科学优秀成果二等奖、教育部高等学校科学研究优秀成果奖二等奖、四川省哲学社会科学优秀成果奖二等奖等多项奖励，在国内外学术刊物发表学术论文100余篇，出版学术著作5部。《金融学季刊》副主编，《系统工程学报》、《系统工程理论与实践》、《管理工程学报》、《管理学报》、《管理评论》和《投资研究》等编委。
大会邀请报告十二： Topic: 隐含因子的信息含量 Reporter：Prof. Zhenlong Zheng, Xiamen University 郑振龙厦门大学
Abstract：It is widely accepted in classic fixed-income papers that the current term structure contains all the information of future changes in interest rates, and information outside of the interest rate term structure has little effect on the prediction of future interest rates. In other words, the term structure of interest rates follow a Markov process. However, this view is inconsistent with the empirical results. How can we explain and complete the gap between the empirical results and the classic theory theoretically? This paper uses Duffee’s(2011) framework of hidden factor to specify the form of Gauss affine model, and finds that we will obtain some different results when the hypothesis that state variables obey Markov process is relaxed. 　In addition, papers using macro variables to explain term premium have some common characteristics. They usually priori assumed some macroeconomic variables could predict term premium and then conducted empirical studies to validate it. Although most researches verified the influence of macro variables on the term permium, the final conclusions are sometimes quite different due to the differences in the sample asset, sample period and the market. Can we change the way of thinking and extract a comprehensive factor firstly that can predict the term premium, and then look back for its economic implications? Along Duffee’s (2011) framework of hidden factor, this paper finds a new way. 　With the sample of China's interbank bond market during January 2005 - May 2012, this paper draws the following conclusions in the new framework of hidden factor: ①By extending the affine model from three factors to five factors,variance ratio used to measure information content significantly reduces from 0.99 to 0.41,which confirms the existence of hidden factors and shows that most of the term premium forecasting information is hidden outside of the interest rate curve.②The existence of hidden factors reminds us that when we study the term premium with the three factor affine model, we should be more cautious since that a large number of information may be omitted. There are several ways to overcome the above shortcomings. One is to increase the number of factors just as this paper. The other is to establish a joint model including macro variables and latent factors simultaneously as Joslin, Priebsch ,Singleton(2009) and Barillas(2010).③This paper confirms the hidden factor does contains abundant macro information with the linear regression and VAR analysis .The correlation of the hidden factors and the price index like CPI and M2 is very strong while the correlation between the hidden factor and the economic growth index is quite week. It is noteworthy that the information of exchange rate is particularly important ,even more than CPI and M2.This may be caused by China’s special monetary policy and exchange rate system. 郑振龙，厦门大学金融工程教授，博士生导师。研究方向主要集中在金融工程领域，主要研究内容为衍生品设计、定价、风险管理和交易策略。先后作为主持人和主要参与者承担了18项国家和省部科研课题的研究，还主持了15项横向课题。出版了30多部著作和教材，发表了近200篇学术论文，并获得20多项省部级优秀社科成果奖。2002年获得霍英东教育基金会第八届青年教师奖，2004年入选教育部新世纪优秀人才支持计划，2009年入选国家百千万人才工程，2011年享受国务院政府特殊津贴。