中国管理科学与工程学会金融计量与风险管理研究会第一届学术年会

大会报告摘要（按报告时间顺序排序）

报告人简介：袁先智博士是上海“千人计划”和四川“千人计划”金融工程特聘专家， 国际金融工程杂志（International Journal of Financial Engineering）主编， 和多家学术专业杂志的编委。他目前是中国系统工程学会金融系统专业委员会副主任委员，中国金融数学金融工程精算和保险专业委员会执行副主任等行业位置，也是国内外多所高校/科研机构的访问、兼职、讲座教授。

袁先智博士目前专注于公司在大数据框架下的支持金融行业的解决方案，特别是在资本市场中债券定价系统的建立，和支持小，中和大型贷款融资的解决方案。他领导公司金融和相关业务部门成功地完成了包含“长江”，“新华08”等一批具有里程碑影响的金融服务解决方案项目。他带队完成的小微企业信用评级系统，颠覆性地开启了小微企业贷款“大数据时代”，有效地解决了广大小微企业客户的融资难题。 袁先智认为，建立互联网金融诊断图谱，“解决企业信 用评估，改善信息不对称，关键数据要呈现。” 为推动中国在金融风险管理方面的理论与实践相结合作出了巨大的贡献。

袁先智先后在四川大学，加拿大多伦多大学， 戴尔豪斯大学获得数学学士，数学硕士，金融工程硕士，统计学硕士， 数学博士等学位。此后在澳大利亚，加拿大，美国和中国的著名高校，国际领先的金融机构，交易公司和财务咨询公司工作， 也为共和国总理提供过解读金融危机等方面的专题报告。

袁博士有在国内外（美国，加拿大，和澳大利亚）超过25年工作和学习的经历。目前在金融工程/金融数学的理论和业界实践应用和非线性泛函分析和相关应用开展研究工作。在国内外SCI 和SSCI学术刊物发表了140多篇专业论文，出版2本专著。在非线性分析和相关的KKM 理论以及在数理（金融）经济学，博弈论，优化理论，非线性集值分析，和非线性集值分析不动点理论方面的应用方面研究方面，取得一系列处于国际领先水平的系统性结果。面前袁博士正带领团队建立具有中国特色的金融科技行业的国际标杆。

在金融业界实践方面, 自1999年以来, 先后在全球领先的KPMG, Deloitte财务和咨询公司, 美国的TXU能源交易公司, 加拿大Montreal 银行等国际公司工作。 他积累了一系列理论和业绩相结合的专业经验和技能, 特别是自2008年以来, 袁博士为德勤中国 （Deloitte China）组建了“定价和计量风险部门”，并为中国银行、工商银行、建设银行、农业银行、招商银行，浦东发展银行等的国内大型金融机构提供基于新巴塞尔协议的专业风险计量与管理服务。

中国父母热衷于送子女出国留学，本文关注了这一现象背后的影响因素。基于中国家庭追踪调查（CFPS）2014年调研数据，我们发现除家庭收入、子女构成以及教育观念等因素之外，父母双亲对社会公平感的认知明显影响到其送子女出国留学的意愿。研究发现，有送子女出国留学倾向的家庭比没有该倾向的家庭社会公平感低34.65%，社会公平感的缺失会导致父母送子女出国留学的倾向提高10.57%。进一步地，母亲缺乏社会公平感的影响更大，亲身经历过不公正对待的父母更加倾向于送子女出国留学。此外，社会公平感缺失的影响不能被教育观念存在差异的因素解释。这些结果表明居民对社会公平感的认知会影响其子女受教育的选择，这不仅加剧了境外消费服务贸易的失衡，而且可能会降低国家对本国青年人才的吸引力。本文对“出国留学热”的动机提供了新的解释，并揭示了居民的缺乏社会公平感可能存在的长期负面影响。

报告人简介：吴卫星，对外经济贸易大学教授、博士生导师，金融学院院长，对外经济贸易大学重点研究基地应用金融研究中心主任，教育部长江学者青年项目获得者。中国管理科学与工程学会金融计量与风险管理研究会副理事长，中国金融工程学会理事，中国国际金融学会常务理事、副秘书长，《经济研究》和《管理世界》编委会委员，《财贸经济》执行编委。

     吴卫星教授博士毕业于中国科学院数学与系统科学研究院，先后在韩国高丽大学，美国哥伦比亚大学做访问学者和高级访问学者。主要研究方向为资产定价、金融工程、风险管理、家庭金融。先后主持了国家自然科学基金面上项目、国家社会科学基金重点项目，国家社会科学基金重大项目。在国内外知名学术期刊如Management Science，Journal of Empirical Finance，Journal of Quantitative Economics、《管理世界》、《金融研究》、《系统工程理论与实践》等发表论文70余篇。其论文曾经获得孙冶方金融创新奖、中国金融教育发展基金会“2008年金融教育优秀科研成果”研究论文类一等奖；中国金融学会优秀论文奖二等奖等多项奖项。

金融系统中有两类基本的风险传染渠道：1）债务违约可能通过金融机构间互相借贷形成的债务网络引发信用风险传染；2）资产甩卖可能通过机构共同持有资产形成的资产重叠网络引发市场风险（价格下跌）传染。本研究探讨这两种风险传染的叠加机制和可能由此引发的系统性风险的测算问题。

报告人简介：朱书尚，湖南湘西人，本科（1997）和硕士（2000）毕业于湘潭大学， 2003年毕业于中国科学院系统科学研究所，获博士学位。2003年7月到2012年1月于复旦大学管理学院任教（讲师、副教授）。现任中山大学管理学院教授、博士生导师。现任中国运筹学会金融工程与金融风险管理分会常务理事、秘书长；中国系统工程学会金融系统工程与风险管理专业委员会理事；中国优选法统筹法与经济数学研究会经济数学与管理数学分会常务理事。

研究领域：包括金融工程与风险管理（研究兴趣：投资组合优化、金融风险管理、金融风险传染等）。在Operations Research, Mathematical Finance, IEEE Transaction on Automatic Control, INFORMS Journal on Computing, Journal of Banking and Finance, Journal of Economic Dynamics and Control, Journal of Computational Finance, Journal of Risk, Quantitative Finance等国内外期刊上发表论文四十余篇。

主要研究方向包括金融衍生产品的设计与定价、公司债券、金融风险管理、信用风险、投资组合分析等；先后主持了4项国家自然科学基金项目、参与了国家973项目和国家自然科学基金重点项目等；在国内外知名学术期刊如European Financial Management、Quantitative Finance、Finance Research Letters、Annals of Economics and Finance、Optimization、Theoretical Computer Science、《管理科学学报》、《系统工程理论与实践》等发表论文50余篇。协助成思危先生出版《虚拟经济概览》、《人民币国际化》、《The Chinese Stock Market》等3部专著；曾任新华社、中债登记结算公司、北京市金融局、北京市住建委等单位的项目评审专家。

随着医疗水平的提高及人们健康意识的增强,世界各国的人口死亡率呈明显下降趋势,人口寿命的增长给养老金体系带来较大的支付压力和偿付风险.针对我国人口死亡率数据少、部分年份数据缺失等问题,首先采用经济学的协整理论,用台湾人口死亡率对中国人口死亡率时间因子做回归,并利用Lee-Carter模型预测了中国男性人口死亡率.然后借鉴十七世纪欧洲的“唐提 (Tontine)”年金理念,依据自助养老年金思想(GSA),提出两种可以增加收益比的方案,构建适合我国国情的自助养老年金体系.最后将预测得到的男性人口死亡率代入自助养老年金体系进行实证分析.研究结果表明: GSA模型具有精算平衡、收益稳健、终身领取等特点.并且优化后的自助养老年金体系,可以让更多的个体参与到年金池中,从而使个体风险得到有效的抑制,提高个人养老保险的人群覆盖率

报告人简介：肖鸿民，女，1967年11月生。兰州大学理学博士，西北师范大学数学与统计学院教授，统计学硕士生导师，美国《数学评论》评论员。主要研究方向为保险精算、金融统计及风险管理。在 Computers & Mathematics with Applications, Communications in Statistics - Theory and Methods、应用数学学报、工程数学学报等SCI及国内学术刊物上公开发表论文30余篇，参编著作3部。主持完成国家自然科学项目“基于保单进入过程的风险系统的建构与应用研究”（71261023），甘肃省高等学校科研项目“相依重尾随机变量的性质及其应用研究”(1001-10)，参与完成多项国家自然基金项目和国家教委博士点专项基金项目。曾获得省教育厅教学成果奖，甘肃省高校科技进步一等奖，指导学生获得大学生数学建模比赛全国一等奖。2014年10月-2015年10月，国家留学委派出以访问学者身份访问美国佐治亚州立大学（Georgia State University）罗宾逊商学院（J. Mack Robinson College of Business）风险管理与保险系（Department of Risk Management and Insurance）。

专题报告 （Session 1 -高频金融数据分析）

Specification testing for volatility function of a diffusion process using high-frequency data will be challenging to the existing testing methods because these testing methods are based on the true equilibrium prices which are unobservable and perturbed by the microstructure noise in practice. In this paper, we develop an estimated price based Chen-type testing approach to solve this high-frequency volatility function testing problem when the microstructure noise can be assumed to be a parametric function of the trading information. We can prove that our testing method has the same asymptotic properties as that of Chen-type based on the true equilibrium prices and we also provide simulation studies for displaying the clear advantage of our testing method. Empirical application demonstrates that the parametric function of volatility does not match the real data, which may be due to an additional noise in practice.

Low-frequency historical data, high-frequency historical data and option data are three major sources, which can be used to forecast the underlying security's volatility. In this paper, we propose a unified GARCH-It/^{o}-IV model, which is the first explicit model integrating three information sources in the literature. Instead of using options' price data directly in the model, we extract the option-implied volatility from the option data and construct its dynamics. We provide the quasi-maximum likelihood estimators for the parameters and establish their asymptotic properties. In empirical analysis, we show that the proposed GARCH-It/^{o}-IV model has better out-of-sample volatility forecasting performance than the popular models, such as GARCH+IV, Realized GARCH(IV), Realized GARCH(RV), HAR-RV and GARCH-It/^{o}.

In this paper, a new model, called Mix-GRACH model, for measuring volatility under mixture frequency data have been established. Due to drawing information from low and high frequency data simultaneously with data-driven weight, the Mix-GARCH model could own more better performance comparing with the existed competitor. The estimation of parameter and its asymptotic theory are investigated in detail and the nice practical performance of the proposal model illustrated by applying to the chosen data.

Who gains and who loses in derivatives, and why? We consider these issues using proprietary brokerage data on a warrants market in China. Individual investors underperform via misunderstanding option payoffs: they treat expiring deep OTM warrants as stock, overpay for skewness, and pay for warrants that are assuredly worthless owing to price limits in the underlying asset market. Trading by institutions and large individual investors is more sophisticated. Regulators set warrants’ price limits that preclude convergence to fundamentals, which materially affects investors’ profits. Overall, our results highlight the importance of ensuring financial sophistication amongst investors and regulators in derivatives markets.

Recent developments document a strong and robust negative relationship between skewness and average return in the financial market. In this work, we consider a one-period-two-date economy with rank-dependent utility agents. After deriving the equilibrium asset pricing formula under probability weighting, we show that in the derived three-moment-CAPM model, the existence of probability weighting would increase the risk premium of coskewness. The reason is that through probability weighting, the rank-dependent utility agents could overweight the tail events, and thus develop a preference for coskewness. The empirical analysis further confirms our findings. We find that the stock’s expected  return is negatively related to its coskewness in high-probability weighting periods but unrelated to its coskewness in low-probability weighting periods.

We investigate the impacts of reference point formation on wealth growth and inequality under a framework of Cumulative Prospect Theory for a network of multiple investors with socially determined reference points. Under our framework, wealth inequality is sufficient for sustained wealth growth for a general class of reference point formation rules. For a specific reference point formation rule, the long run behavior of the wealth gap, wealth growth and inequality measured by Gini coefficient depends on how ambitious investors update their reference points. An application of our model to a large network consisting of several subnetworks reveals that subnetworks demonstrate different wealth growth rates and the inequality of the large network is growing, which is supported by empirical observations.

There is compelling evidence that volatility is negatively related to future returns in most markets studied.  Overconfidence is often proposed as an important explanation for this phenomenon.  We structure a new test to detect the overpricing of higher volatile stocks under self-contribution and limited attention. Our empirical strategy is to develop a market price bubble index to capture the bubble-like feature of a stock market perceived by investors and use this bubble index to distinguish between high- and low-bubble periods. A regime switching strategy is then developed to test the hypothesis that the negative relation between volatility and reward, as implied by the low-volatility anomaly, becomes very weak or even reversed in direction in the bubble-like periods characterized by extremely high returns.

文章使用主动占比和追踪误差分别从特有风险和系统性风险的角度测度基金经理的主动管理能力。研究发现，基金经理对业绩具有显著的影响。在控制基金特征等因素之后，基金经理的主动管理能力与业绩等因素相关，结论表明，业绩好的基金经理通常具有较高的主动管理能力。持续性分析表明，主动管理能力高的基金经理具有显著的持续性，表明他们具有真实的主动管理能力。此外，基金经理变更的研究表明，经理变更对基金业绩具有负向的影响。

专题报告 （Session 3- 金融计量）

In modern financial markets, volatility measures the degree of dispersion for a trading asset and plays an important role in portfolio allocation, performance evaluation, and risk management. Traditional approach adopted to model the dynamic evolution of volatility process uses parametric models such as generalized autoregressive conditional heteroscedasticity (GARCH) model and considers low-frequency data, data at daily or longer time horizons. Modern approach investigates volatility process with high-frequency data that refer to intra-daily observations. Various methodologies have been developed in literature for volatility estimation with high-frequency data, well-known estimators include multi-scale realized volatility (MSRV) estimator, pre-averaging realized volatility (PRV) estimator and kernel realized volatility (KRV) estimator. However, the interconnection between data gathered at two different time scales cannot be ignored. In this talk, we introduce unified models for modeling high-frequency financial data that can accommodate both the continuous-time Ito diffusion and the discrete-time realized GARCH model by embedding the discrete realized GARCH model in the continuous instantaneous volatility process. The key feature of the proposed model is that its conditional daily integrated volatility has an autoregressive structure that has been seen within a realized GARCH model. Given the autoregressive structure in the conditional daily integrated volatility, we propose a quasi-likelihood function for parameter estimation and establish its asymptotic properties. To improve the parameter estimation, we propose a joint quasi-likelihood function that is built on the marriage of daily integrated volatility estimated by high-frequency data and daily implied volatility obtained from option data. We conduct a simulation study to check finite sample performance of proposed methodology and an empirical study with the Bank of America stock and option data.

In this paper we develop a new Ito-Taylor expansion approach to deriving analytical approximations to the transition densities of multivariate diffusions. The obtained approximations can thereby be used to carry out the maximum likelihood estimation for the diffusions with discretely sampled data. This new approach is universally applicable for a wide spectrum of models, particularly time-inhomogeneous non-affine irreducible multivariate diffusions. We manage to show that the expansion converges to the true probability density under some regularity conditions. The method enjoys a significant computational efficiency over the other alternatives in the literature because we provide explicit recursive formulas for the expansion coefficients. Extensive numerical experiments on a variety of models demonstrate the accuracy and efficiency of the estimators based on the approximate densities yielded from our approach.

A crucial step in developing personalized treatment strategies is to identify the latent subgroups of patients of a heterogeneous population. In this article, we consider a general class of heterogeneous transformation models for subgroup identification, under which an unknown monotonic transformation of the response is linearly related to the covariates via subject-specific regression coefficients with unknown error distribution. This class of models is broad enough to cover many popular models, including the heterogeneous linear model, the heterogeneous Cox’s proportional hazard model and the heterogeneous proportional odds model. Without any priori grouping information, we propose a robust method based on the maximum rank correlation and a concave fusion to automatically identify the latent subgroup structure and estimate the subgroup-specific treatment effects simultaneously. We establish the theoretical properties of our proposed estimate under regularity conditions. A random weighting resampling scheme is used for variance estimation. The proposed procedure can be easily extended to handle censored data. Numerical studies including simulations and a real data analysis demonstrate that the proposed method performs reasonably well in practical situations.

专题报告 （Session 4 -风险管理与保险）

This paper examines the evolution of systemic risk in the Chinese banking system using the risk measure under skewed-t distribution. Our empirical results reveal that dynamic Chinese banks’ systemic risk contribution reach the highest during stock market crash in 2015. In addition, we construct a forward-looking systemic risk measure by projecting   on lagged firm characteristics such as Value-at-Risk (VaR), size, leverage, maturity mismatch and ROA. The panel regression results imply that banks’ size, leverage, maturity mismatch positively drive their systemic risk contribution changes, and ROA show negative relationship with their systemic risk contribution. Moreover, banks’ own VaR positively affects their systemic risk contribution, which means that riskier banks contribute more to Chinese financial system. This forward-looking measure can potentially be used in macro prudential policy applications.

This paper studies the pricing of liquidity risk in Chinese corporate bonds for the period from January 2008 to December 2017. We construct a liquidity measure which captures liquidity risk from four dimensions mentioned in Liu (2006). Incorporating liquidity pricing factor into term-and-default model, we propose a new bond pricing model.  The results suggest that liquidity risk is a significant determinant of corporate bond returns in China.

This paper studies the optimal contract design problem in risk-sharing between an insurer and a reinsurer. The insurer purchases reinsurance for risk-control and decides her retention level with an objective to minimize her ruin probability. The reinsurer has control over the reinsurance premium and aims to maximize her expected discounted profits up to the time when the insurer goes bankrupt. In a stochastic differential game-theoretic framework, the optimal reinsurance strategy and the optimal reinsurance premium strategy are characterized as a Nash equilibrium and are obtained by solving a system of coupled Hamilton-Jacobi-Bellman equations. We obtain explicit solutions for the game problem when both the insurance and the reinsurance premiums are calculated according to the standard-deviation principle or the expected value principle, respectively. We also perform numerical examples and provide economic interpretations for the results.

This paper investigates an insurance design problem, in which a bonus will be given to the insured if no claim has been made during the whole lifetime of the contract, for an expected utility insured. In this problem, the insured has to consider the so-called optimal action rather than the contracted compensation (or indemnity) due to the existence of the bonus. For any pre-agreed bonus, the optimal insurance contract is given explicitly and shown to be either the full coverage contract when the insured pays high enough premium, or a deductible one otherwise. The optimal contract and bonus are also derived explicitly if the insured is allowed to choose both of them. The contract turns out to be of either zero reward or zero deductible. In all cases, the optimal contracts are universal, that is, they do not depend on specific form of the utility of the insured. A numerical example is also provided to illustrate the main theoretical results of the paper.

基于谢尔宾斯基垫片上的边渗流构建了一个非线性的随机交互价格动态模型, 试图重构和研究实际证券市场收益率序列的复杂性行为. 引入分数阶数组熵和分数阶数组复杂性来分析金融信号的复杂性行为. 为了更深入地理解随机价格进化的波动性质, 分析了随机对数收益率和波动的复杂性行为, 主要有幂率分析, 分数阶样本熵和分数阶数组复杂性. 为了更进一步证实模型的合理性和有效性,对真实证券市场数据集做了对比统计研究. 实证结果表明了该价格模型能在某种程度上重构真实证券市场的复杂性行为.

专题报告 （Session 5 - 投资组合优化）期权投资组合的风险度量和优化

本文第1部分通过研究正态随机向量二次多项式的特征函数、矩和累积量，并应用分布函数近似和线性规划的方法，给出了基于Greeks的期权投资组合方差/VaR/CVaR的估计；本文第2部分讨论了期权投资组合的单期和（离散时间）多期优化问题，考察了模型的不适定性并分析了其原因，同时指出通过求解一个线性方程组即可得到所有“Greeks-有效”期权投资组合；本文第3部分基于鞅方法并利用随机贴现因子建立了连续时间下动态“均值-方差”最优的期权投资策略选择模型，并进行求解；本文第4部分使用多维空间中的波动率曲线建立凸分析优化问题，并在完备市场模型下分别讨论了使用市场Greeks和理论Greeks对冲组合风险的期权波动率套利策略的动态优化。

This work studies the constrained optimal execution problem with a random market depth in the limit order market. Motivated from the real trading activities, our execution model considers the execution bounds and allows the random market depth to be statistically correlated in different periods. Usually, it is difficult to achieve the analytical solution for this class of constrained dynamic decision problem. Thanks to the special structure of this model, by applying the proposed state separation theorem and dynamic programming, we successfully obtain the analytical execution policy. The revealed policy is of feedback nature. Examples are provided to illustrate our solution methods. Simulation results demonstrate the advantages of our model comparing with the classical execution policy.

For long investment horizon, dynamic control of value at risk (VaR) is indispensable to achieve high performance of risk management. Unfortunately, the resulted dynamic mean-VaR portfolio selection formulation is time inconsistent, which also leads to a non-tractability in deriving optimal investment policy. We tackle this long-standing challenge from a new angle. We first prove an equivalence between the mean-VaR and the mean-safety-first formulations when the risk parameters in the two formulations satisfy certain relationship. Solving the time-consistent dynamic mean-safety-first formulation using either dynamic programming or martingale approach enables us to identify corresponding optimal investment policy for the time-inconsistent mean-VaR problem formulation. Furthermore, investigating the property of the induced confidence level in truncated mean-VaR problems gives rise to the stagewise nested VaR structure, which guides us adjusting the VaR level dynamically and adaptively during the investment process such that the global mean-VaR goal is attained.

We propose a new smooth nonparametric kernel (NPK) portfolio optimization method for managing downside risk. Lower partial moments (LPM) are used as our measure of downside risk and we include transaction costs into our optimization model. We demonstrate that the mean-LPM model is a convex optimization model when the order of LPM is equal to or greater than one. We numerically and empirically compare our method with the traditional moment (MOM) method and find that our method produces portfolios that dominate those found using MOM. Numerical results show that our NPK method outperforms the MOM method in terms of estimation accuracy. Empirical tests show that NPK outperforms MOM in terms of portfolio performance using a wide variety of performance metrics. Furthermore, NPK has lower transaction cost at higher orders of LPM. In addition, NPK outperforms a naive equally-weighted strategy and the market index.