The problem of allocating responsibility for risk among members of a portfolio arises in a variety of financial and risk-management contexts. Examples are particularly prominent in the insurance sector, where actuaries have long struggled with issues of distributing profit loadings and/or capital (net worth) across a number of distinct exposure units or accounts according to their relative contributions to the total risk of an insurer’s portfolio. Although substantial work has been done on this problem, a general solution for inhomogeneous loss distributions has not been presented. The purpose of this article is to show that the value-assignment method of non-atomic cooperative games proposed by Aumann and Shapley (1974) provides an effective solution to risk-allocation problems of the most general form.

　　2. Optimal Hedge Ratio of Commodity Futures Using Bivariate DCC-CARR and DCCGARCH Models

　　In this paper, we consider various methods in minimum-variance hedge strategy. We compute the Optimal Hedge Ratios (OHRs) between spot and futures for some commodity prices using different econometric methods. We use the Dynamic Conditional Correlation - Conditional Autoregressive Range (DCC-CARR) model proposed by Chou et. al. (2005) to compute the OHRs. Other alternative methods used for comparison include the ordinary least squares (OLS) estimator, Constant Conditional Correlation models, and DCC-GARCH model of Engle (2002). Different methods are compared with each other in their performance of variance-reduction. For withinsample hedge, the results show that the DCC-CARR model performs better than other hedge models for the selected commodities with the exception of gold. For the out-sample hedge, the DCC-CARR model is the best model for all commodities. In conclusion, we suggest that the DCC-CARR model is the better model for investors to find the minimum-variance of a portfolio.

　　3. Linear Time Series Processes with Mixed Data Sampling and MIDAS Regression Models

　　Economic data are not available at the same sampling frequency. We start by studying linear time series processes with mixed data sampling (MIDAS). Much has been written about temporal aggregation in linear models, yet the literature assumed that all processes are sampled at the same (high) frequency and studies the consequences of temporally aggregating the process. We start with a general setting of linear time series processes and study three cases (1) all processes are sampled at high frequency, (2) all processes are sampled at some low common frequency and last but not least (3) some processes are sampled at a higher frequency than others. We then ask: if some series are available at a higher frequency, how much do we gain by considering the mixture of aggregated and disaggregated data. We also compare the MIDAS setting with Kalman filtering. We turn next to the case where one series is sampled at low frequency and a number of series are sampled at a high frequency. This setting leads to the definition of MIDAS regression models. We draw comparisons with distributed lag models and MIDAS regressions. We also introduce so called reverse MIDAS regressions, the latter allow us to measure the impact of low frequency data onto high frequency processes, whereas MIDAS regressions achieve the opposite. Most importantly, we derive conditions, verifiable via an empirical test, that allow us to empirically verify the circumstances under which mixed data sampling achieves the same forecasting MSE efficiency as the hypothetical situation where all the series are available at high frequency. The paper concludes with a detailed Monte Carlo study that examines (1) forecast performance comparisons between the three cases (1) all processes are sampled at high frequency, (2) all processes are sampled at some low common frequency and last but not least (3) some processes are sampled at a higher frequency than others. The simulation study also examines how well our empirical test performs in practice. We also examine discretization bias issues in a simulation setting and last but not least we compare MIDAS regressions with the Kalman filter approach.

　　4. Cash Sub-Additive Risk Measures under Interest Rate Ambiguity

　　A new class of cash sub-additive measures R to assess the risk of financial positions under interest rates is introduced. R can be used to assess other financial and insurance risks, such as default risk. A link between R and cash invariant convex risk measures is established and the dual representation us derived using cash invariant risk measure property. Finally, R can be used to disentangle the intrinsic risk of financial positions and the risk due to the numbers.

　　5. A Central Limit Theorem for Computation of Option Prices for Stochastic Volatility Models

　　We consider European options on a price process that follows the log-linear stochastic volatility model. There are three stochastic integrals in the option pricing formula that are costly to compute. We derive a central limit theory that provides approximations to these integrals. For thirty days at the money options with parameter settings appropriate to foreign exchange data, our formula improve computation speed by a factor of 1000 over brute force Monte Carlo at comparable accuracy. This improvement in speed gets computational efficiency to the point where Markov Chain Monte Carlo (MCMC) statistical methods are feasible. We verify that MCMC statistical methods perform well on simulated joint price and options data that follow the log-linear stochastic volatility model. We provide estimates of model parameters from daily data on the Swiss Franc to Euro and Japanese Yen to Euro over the period of 1999 to 2002.

　　6. A Functional Approach to Interest Rate Modelling

　　In this paper, we model interest rates as functionals of a basic Markov state variable, which couples Ornstein-Uhlenbeck process and Bessel process. We obtain a whole class of extremely parsimonious models. This simple approach unifies many previously proposed diffusion models of interest rates, and provides an easy scheme for establishing other interest rate models. This is a joint work with Shunlong Luo of AMSS of Academy of Sciences and Qiang Zhang of City University of Hong Kong.

　　7. The Comparative Study Between The CARR model And The GARCH Race Models On The Volatility Of The Financial Market

　　In this paper we provide the CARR race models based on the CARR model, and we utilize them with the financial market daily time series to carry on the empirical analysis. We find that the CARR race models are more effective in forecast in these financial markets than the GARCH race models that have the same format with them.

　　8. Endogenous VaR Confidence Level and Empirical Studies with CAViaR Models

　　In testing the robustness of VaR models, it will be affected seriously by setting the confidence levels exogenously, since when it is too small, we can not carry on a valid test, but if it is too large, we will go away from the risk management practice (e.g. required level by Basel Committee’s regulation for risk monitoring and management). Because of the non-normality of financial variables, the estimators of coefficients?goodness of fit and efficiency of prediction are rather sensitive to the choice of confidence level for many VaR models. Thus, as many researchers have showed, the out-of-sample prediction ability of the models varies dramatically from that of in-sample. One of the typical examples for the argument is in the Conditional Autoregressive Value-at-Risk (The CAViaR) model introduced by Engle (2002). In this paper, based on the idea of backtesting, we investigate the selection of confidence level according to the endogenous factors of the models and consider to set the confidence level endogenously. We expand the application of DQ (Dynamic quantile) test and design a likelihood statistic of Weighted Sum of Absolute Errors (WSAE) to test the prediction efficiency of model, with which we can systematically survey the movement of the prediction efficiency in an interesting domain of the confidence level so as to get an optimal level for the risk management practice. We do it empirically with CAViaR models.

　　9. On Minimum-Variance Arbitrage Portfolio

　　In this short note, based on the convex analysis theory, the existence of the minimumvariance arbitrage portfolio (MVAP) is obtained. Then an analytical formula and some basic properties of MVAP are presented after a comparison with the Korkie and Turtles results (2002).

　　10. American Currency Option Pricing in an Environment of Mean-Reverting Interest Rates and Stochastic Volatilities

　　This paper describes an investigation into American currency-option pricing models. By considering and extending the Amin and Bodurtha (1995) model, the paper builds an improved model with stochastic mean-reverting interest rates based on the CoxIngersoll-Ross (1985) model and thereafter stochastic mean-reverting volatilities (Heston, 1993). The considerably more complex model is approached via an enhanced Monte Carlo simulation approach (Duck et al., 2005) developed from the method of Longstaff and Schwartz (2001). To implement the model practically, the parameters from the empirical estimation papers of Dupoyet (2006), and Treepongkaruna and Gray (2003) are used. Using the example of a liquid cash currency option, the pricing of the option is sought as a more general case of other exchange-rate related options.

　　11. Extension and Optimization of Kernel Graduation

　　In this work, we improve kernel graduation and the precision of boundary graduation is increased remarkably. A new Gamma kernel estimator for graduation is studied and an optimal Gamma kernel is derived. This results are compared and the methods are applied to the fifth census of China.

　　12. The Roles of Margin Adjustment on Controlling Futures Market Risk and the Reform of Margin System Empirical Analysis from Dalian Commodity Exchange

　　This paper makes theoretical and empirical analysis of the roles of margin adjustment on adjusting futures market risk based on the data from Dalian Commodity Exchange and method of VaR(Value-at-Risk). We draw conclusions that market risk decreases when margin lever is increased and does not change when the margin lever is decreased. Based on the conclusions and the realities in China market, we make suggestions on the reform of the margin system in futures market, complying with the basic principles of prudentiality and opportunity cost.

　　13. Does “Quan Qian” Motivation Exist? — Evidence from Rights Offering and Public Offering in China

　　This paper documents a series of empirical studies focusing on the “Quan Qian” motivation in rights offerings and public offerings of China. Our result shows that firms have strong propensity towards rights issue if they are motivated by “Quan Qian” compared with other prevailing theories. It is also justified that “Quan Qian” incentive dominates in the rights offerings between 1995 and 2000 while this incentive is not significant in the public offerings. Furthermore, the empirical results indicate that “Quan Qian” motivation can explain particular phenomena of seasoned equity offerings, including ROE manipulation, changing the use of proceeds and blockholder’s waiver of subscription rights.

　　14. Efficiencies of Life Insurers in China – An Application of Data Envelopment Analysis

　　Using Data Envelopment Analysis (DEA), we analyzed the relative efficiency of Chinese life insurers between 2000 and 2003, identify the drives of technical efficiency, and discuss the characteristic of the life insurance industry in China. This paper also compares the technical efficiencies, pure technical efficiencies, and scale efficiencies between different groups of life insurers. Through these comparisons, the styles of development and the characteristics of operation are identified. In addition, with the advantage of DEA, this paper also discusses the topics of scale economies, shadow price, improvement space, and Malmquist index. We believe our work is beneficial for researchers and practitioners to better understand the Chinese life insurance industry. Some of our suggestions are presented at the end of the paper.

　　15. Do Board Characteristics Have an Impact on a Listed Companys Behavior of Committing Fraud?

　　This paper selected 195 A-share listed companies which received enforcement actions from the regulators in China due to committing fraud spanning from 2001 to 2005 as sample. Applying Binary Probit regression, General Linear Model and Ordered Probit regression analysis, it investigated the relation between the Board characteristics and the properties of corporate fraud, surrogated by the probability of occurrence, the frequency and the degree of fraud. The study finds that: First, contrary to the Agency theory, the results show that the general manager duality significantly helps to reduce fraud and alleviate the degree of fraud, which supports the Stewardship theory. Second, the high frequency of Board meetings does not mean that the directors are more likely to perform their duties diligently, but to react to the poor performance of a company.

　　16. Nonparametric Estimation of Copula Functions For Dependence Modeling

　　Copulas are full measures of dependence among components of random vectors. Unlike the marginal and the joint distributions which are directly observable, a copula is a hidden dependence structure that couples a joint distribution with its marginals. This makes the task of proposing a parametric copula model non-trivial and is where a nonparametric estimator can play a significant role. In this paper, we propose a kernel estimator which is the mean square consistent everywhere in the support of the copula function. The bias and variance of the copula estimator are derived which reveal the effects of kernel smoothing on the copula estimation. A smoothing bandwidth selection rule based on the derived bias and variance is proposed. The kernel estimator is then used to formulate a goodness-of-fit test for parametric copula models.

　　17. Estimating Covariation: Epps Effect, Microstructure Noise

　　This paper is about how to estimate the integrated covariance < X, Y > of two assets over a fixed time horizon [0, T], when the observations about X and Y are “contaminated” and when such noisy observations are at discrete, but not synchronized, times. We show that the usual covariance estimator is biased, and the size of the bias is more pronounced for less liquid assets. We also provide optimal sampling frequency which balances the tradeoff between the bias and various sources of stochastic error terms, including non-synchronous trading, microstructure noise, and time discretization.

　　18. Multiscale Jump and Volatility Analysis for High-Frequency Financial Data

　　Volatilities of asset returns are pivotal for many issues in financial economics. The availability of high frequency intraday data should allow us to estimate volatility more accurately. Asset prices often contain jumps, and high-frequency financial data are inevitably contaminated with market microstructure noise. Existing methods can deal with noisy data for the continuous diffusion price model or handle the jump-diffusion price model without noise. This talk will present estimation of integrated volatility and jump variation for noisy high-frequency financial data with jumps. The proposed wavelet based multi-scale methodology can cope with both jumps in the price and market microstructure noise in the data, and estimate both integrated volatility and jump variation from the noisy data. We establish convergence rates for the proposed estimators of integrated volatility and jump variation. In particular, we show that the integrated volatility can be estimated asymptotically under the jump-diffusion price model as well as under the continuous diffusion price model. Simulations are conducted to assess the performance of the proposed estimators and to compare them with existing ones. Theoretical and numerical analysis show that the proposed estimators outperform existing methods for noisy high-frequency data under the jumpdiffusion model, and have comparable performance for the continuous diffusion model and noiseless jump-diffusion model. The methods are illustrated by applications to two high-frequency exchange rate data sets.

　　19. Realized Volatility: A Review

　　This paper reviews the exciting and rapidly expanding realized volatility literature. A simple discrete time model is presented in order to motivate the main results. The continuous time specification is considered, with and without the presence of microstructure noise, which causes severe problems in terms of the consistency of the daily realized volatility estimator. Independent and dependent noise processes are examined. The most important solutions currently available for the consistency problem are presented, and a critical exposition of different techniques is given. Various modelling issues are tackled, and the main empirical findings are summarized. Several applications are also discussed.

　　20. Benchmarked Executives’ Compensation and Optimal Option Repricing

　　This paper argues that executive compensation containing relative performance sensitivity to insulate the managers from systematic risk offers a reference for how sensitive Executive officer’s compensation should be, and provide low cost and credible information to perform ex post adjustment in minimizing the monitoring cost. A simple smoothed relative performance model, constructed on the overarching argument that the mean of the inducible effort for different types of managers is decided by their marginal substitution rate regarding current consumption and human capital investment, is constructed first. It is proved that such compensation scheme maximize the incentive of a typical and more than typical manager while effectively screen out less than typical manager. The model is extended to option repricng scenario. Via examining option repricing as a relatively benchmarked compensation, the extended model generates quantitatively critical conditions on when and why the option should be repriced. An optimal repricing range follows in correspondence. The conditions show the negative relationship between option repricing and market performance may be spurious due to a biased sample obtained on a comparatively short time interval. Examinations using empirical data from previous research suggest current option repricing practices are well confined by the upper bound of the optimal range hence avoid the incentive dampening effect. In contrast, the lower bound is, very likely, breached and consequently makes current practices inefficient in screening out less than typical managers. Other relevant Important empirical finding so far are also discussed in the rubric of the model, of which include the size effect of option repricing, management entrenchment and power problem, to name a few.

　　21. Option Bounds and Second Order Arbitrage Opportunities

　　Analogous to Mertons no arbitrage option bounds, second stochastic dominance option bounds give the range of possible option values baring second order arbitrage opportunities. In this paper we first derive second order stochastic dominance option bounds given the prices of any number of other options. We show that the optimal option bounds are given by piecewise constant pricing kernels. When these bounds are violated there are second order arbitrage opportunities in the market. We then establish the optimal arbitrage strategies to make profits from these opportunities. The results make it possible to test second stochastic dominance using data from options markets.

　　22. Corporate Law-breaching Behavior and Managerial Shareholding under Different State Ownership Structure — The Empirical Study on China Listed Firms

　　Taking into account the different degree of political control in all state-owned firms, we analyze the law-breaching behaviors of firms under different state ownership structure, using the data of all A-share firms in China from 2000 to 2004. We find that the probability of a firm’s misbehavior has a positive relationship with the legal-owned concentration and has a negative relationship with the state-owned concentration. Moreover, under the state-owned structure (not the legal-owned structure), managerial shareholding can efficiently decrease the probability of corporate misbehaviors as a whole, but the marginal effect decreases.

　　23. Co-Integration and Error-Correction Model for China’s Money Supply Function

　　Conventionally, when setting up a linear regression model the time series are usually presumed to be steady to ensure that the variables estimated by common least square method are consistent and in asymptotically normal distribution. However, most economic time series are unsteady and they are easily resulting in so-called “false regression” when linearly regressed. So, an error-correction model is set up on the basis of the co-integration theory and by virtue of statistics, and handled data, especially the empirical analysis of money supply, which is done by the econometric Eviews statistic software. The autocorrelation and heteroskedasticity of residual of the error-correction model were tested, and the results showed that the model can get rid of the “false regression” effectively and explain well the economic phenomena.

　　24. Option Pricing with Aggregation of Physical Models and Empirical Learning

　　Financial mathematical models are useful tools for option pricing. These physical models provide a good first order approximation to the underlying dynamics in the financial market. Their pricing performance can be significantly enhanced when they are combined with statistical learning approaches, which empirically learn and correct pricing errors through estimating state price densities. In this paper, we propose a new semiparametric technique for estimating state price densities and pricing financial derivatives. This method is based on a semiparametric approach to estimating the survivor function of a normalized state variable and is easy to implement. Our method can be combined with any model-based pricing formula to correct the systematic biases of pricing errors and enhance the predictive power. Empirical studies based on S&P 500 index options show that our method outperforms several competing pricing models in terms of predictive and hedging ability.

　　25. Conditional Quantile Estimation for GARCH Models

　　Conditional quantiles is an essential ingredient in various risk measures. In nowadays, estimation of (conditional) quantiles is a common practice in risk management operations. The purpose of this paper is to propose a robust, flexible approach to estimating conditional quantiles using a quantile regression approach. The GARCH process has proven to be highly successful in modelling financial return data, and is arguably the most frequently used class of models in financial applications. In this paper, we investigate a new, robust approach of estimating conditional quantiles based on GARCH type models. Quantile regression estimation of GARCH models is highly nonlinear. We discuss the problem of estimating this type model using traditional recursive methods for nonlinear quantile regression, and propose two new methods of estimating quantiles of GARCH models. Asymptotic properties of the proposed methods are investigated.

　　26. Detection of Changes in Return byWavelet Smoother with Conditional Heteroscedastic Volatility

　　In this paper, we propose two empirical wavelet coefficient estimators for change points or gamma-sharp cusps, namely, the integral estimator and discretized estimator for the wavelet coefficient of the return function in the nonparametric drift-plus-diffusion model. These estimators can be used as statistics to test the jumps or gamma-sharp cusps of the return function. The model allows for lagged-dependent variables and other mixing regressors. The asymptotic distributions of the statistics are established, hence, the asymptotic critical values can be obtained analytically. Meanwhile, the test can be used to determine some consistent estimators for the locations of change points or gamma-sharp cusps. The jump sizes and locations of change points as well as gamma-sharp cusps can be consistently determined via the test procedures. Some Monte Carlo simulations have been performed in order to check the powers and sizes of the test statistics. Some practical examples in finance and economics have been considered to detect changes in the return by the empirical wavelet coefficients. This is a joint work with Gongmeng Chen, Department of Accountancy, The Hong Kong Polytechnic University and Yoon K. Choi, Department of Finance, College of Business Administration, University of Central Florida.

　　27. Dynamic Conditional Correlation Analysis Between Stock and Interest Rate

　　This paper investigates the dynamic conditional correlation between stock and interest rate by employing a dynamic multivariate GARCH model and ACC model. The statistics show that the stock return is negatively correlated with the daily interest rate. The evidence reveals that the correlation coefficient between stock and interest is time varying. Analyzing the dynamic path of the correlation coefficients suggests that the increase in negative correlation from 1999 is related to the mature of Chinese financial market.

　　28. The Optimal Hedging Ratio of SHFE Cooper Futures

　　This paper uses a long memory model to analyze the SHFE Cooper Futures. We study the lead-lag relationship between futures price and spots price, and the characteristics of futures market. Utilized in this study are the SHFE futures data and the ChangJing spots data, we examines the performance of various hedge ratios estimated from different econometric models: The VAR model, the EC model, and the FIEC model. Our analysis identifies that both futures and spots have long memory and the futures market creates the hedge opportunity. The FIEC model is introduced as a new model for estimating the hedge ratio and it providing better post-sample hedging performance in the return-risk context.

　　29. A Numerical Solution of Ruin Probability of Erlang(2) Risk Processes with Constant Interest Force

　　We consider a Sparre-Andersen risk process with a constant interest force for which the claim inter-arrival distribution is Erlang(2). To study the corresponding ruin probability, the integro-differential equation is derived. However, not only its analytical solution is difficult to find but also the initial condition is not obvious. In this paper, we propose using the technique of importance sampling to estimate the initial condition and solving the corresponding integro-differential equation by block-by-block method similar to Paulsen, Kasozi and Steigen Paulsen (2005). We also present an effective way to bound the error caused by the estimated initial condition. This is a joint work with Yijun Hu and Yan Liu in Wuhan University and Remus K.W. Ho in the Chinese University of Hong Kong.

　　30. Multivariate Time Series Modelling: Common Factors and Nonstationarity.

　　We propose a new method for estimating factor multivariate time series models. One distinctive feature of the new approach is that it is applicable to nonstationary time series. The unobservable (nonstationary) factors are identified via expanding the innovation space step by step; therefore solving a high-dimensional optimisation problem by many low-dimensional sub-problems. Numerical illustration will also be presented.

　　31. Incorporating Economic Objectives into Bayesian Priors: Portfolio Choice Under Parameter Uncertainty

　　Parameter estimation risk is pervasive in risk management. In the context of portfolio selection, this paper proposes a way to allow priors to reflect the objective of maximizing an expected utility. Using monthly returns of the Fama-French 25 assets and their three factors from January 1965 to December 2004, we find that the objective-based priors out-perform alternative priors substantially, with annual certainty-equivalent gains of over 10% in many cases. The better performance is present even in repeated sampling experiments, suggesting that objective-based Bayesian optimal portfolios are superior decision rules even judged by the classical statistical criterion.

　　32. Is Systematic Risk Priced in Options?

　　In this empirical study, we demonstrate the importance of systematic risk in option prices. We do so by examining two testable hypotheses relating both the level and slope of implied volatility curves to the systematic risk of the underlying asset. Using daily option quotes on the S&P 100 index and its 30 largest component stocks, we show that after controlling for the underlying asset’s total risk, a higher amount of systematic risk leads to a higher level of implied volatility and a steeper slope of the implied volatility curve. The findings are robust to various alternative specifications and estimations.

　　33. Estimation Bias in Testing Asset Pricing Models By Mimicking Portfolios

　　This paper address the estimation bias problem in testing time series asset pricing models. Since we have no idea about the true risk factors but only use some linear portfolios as their mimicking, these representatives themselves include noise term that will affect the estimation and result in the estimation biases. Due to the fact that traditional statistical test shave no power to check whether the significant correlation comes from useful risk information or useless noise information, they are not robust as the benchmarks for evaluating different asset pricing models. Moreover, it could be misleading for studying risk-return tradeoff if we just depend on such traditional tests. Based on residual analysis, Hong and Lee (2005a, 2005b) General Spectral Derivatives Test could effectively differentiate the correlation by risk information from that by noise information and give the correct direction for asset pricing model specification. These results are shown in simulation and the approach is applied to Fama-French portfolios in the end. SMB and HML really reduce the model specification error. They both include some useful risk information and can be regarded as risk factors. However, three factor model is not enough to fully explain the changes of stock returns, suggesting that other risk factors or nonlinearity should be introduced for future asset pricing research.

　　34 The Seeking SolitaryWave in Nonlinear Finance Market – The Trade Price Fluctuation Speculated Modeling and Practicing

　　This paper gains the geometry configuration of the differential coefficient manifold by the space reconstructing for the trade data in finance market, discovers the YangMills functional in finance market, obtains a meaningful conserved quantity through corresponding non-Abel localization gauge symmetry transformation, and educes the finance solution, which explains that there is strict symmetry between manifold fiber bundle and gauge field in finance market. This paper straightly simulates and validates a fact of solution existing by synchronization trade experiment of real-time simulating and firm-offer testing some stocks and futures. The solution discovered in finance trade market indicates that there is a kind of new substance and energy existing form in finance trade market (future, stock). The model can provide a quantitative decisionmaking basis for speculation, investment and risk manage in practice.

　　35. Warrant Pricing under Creating Mechanism in China

　　An efficient algorithm is developed to price European warrants in China, in the presence of transaction costs, using utility indifference pricing approach, introduced by Hodges and Neuberger (1989). In China, warrants are different from both normal warrants because the former are not issued by the company but its large shareholders and securities companies, and normal options, because investors can’t write them. Securities companies must deposit the same amount of shares as guarantee when they want to write call warrants and deposit the same amount of money as guarantee when they want to write put warrants. Short sales on shares are prohibited in China. There are transaction costs with both warrants and shares. These characteristics make the warrant pricing significantly different from that in other countries.

　　36. Scalar Measures of Volatility and Dependence for the Multivariate Models of Financial Markets

　　The variance-covariance matrix i s a multi-dimensional array of numbers, gathering all the information about the individual variabilities and the pairwise linear dependence of a set of variables. However, the matrix itself is difficult to interpret in a concise way. Following Frisch (1929), we suggest a scalar measure of overall variabilities (and dependence) by collapsing all the elements in a variance-covariance matrix into a single quantity. The determinant of the covariance matrix ∑, called the generalized variance, can be used as a measure of overall spread of the multivariate distribution. Similarly, the positive square root of the determinant |R| of the correlation matrix, called the scatter coefficient, will be a measure of linear independence among the random variables while collective correlation +(1-|R|)^1/2 is a measure of linear dependence. In an empirical application to the five Asian market returns, these statistics perform the intended roles successfully. In addition, these are shown to be able to reveal the empirical facts which can not be uncovered by the traditional methods. Particularly, we show that both the contagion and interdependence among the national equity markets could be confirmed in contrast to the previous studies .

　　37. Liquidity and Short Rates in the Inter-Bank Market

　　This paper studies the repos rates and CHIBOR rates in the inter-bank market of China. It is found the spreads of CHIBOR rates and repo rates are not related to default risk, but related to liquidity risk obviously. It is also found that term risk premiums for CHIBOR rates and repo rates are both significant. The longer maturities of the interest rates, the higher risk premiums; and the higher expected or unexpected liquidity of the interest rates, the lower risk premiums of the interest rates.

　　38. A Gaussian Calculus for Inference from High Frequency Data

　　In the econometric literature of high frequency data, it is often assumed that one can carry out inference conditionally on the underlying volatility processes. In other words, conditionally Gaussian systems are considered. This is often referred to as the assumption of “no leverage effect”. This is often a reasonable thing to do, as general estimators and results can often be conjectured from considering the conditionally Gaussian case. The purpose of this paper is to try to give some more structure to the things one can do with the Gaussian assumption. We shall argue in the following that there is a whole treasure chest of tools that can be brought to bear on high frequency data problems in this case. We shall in particular consider approximations involving locally constant volatility processes, and develop a general theory for this approximation. As applications of the theory, we propose an improved estimator of quarticity, an ANOVA for processes with multiple regressors, and an estimator for error bars on the Hayashi-Yoshida estimator of quadratic covariation.

　　39. A Simulation Method for Testing the Time Homogeneity of Credit Rating Transitions

　　The measurement of credit quality is at the heart of the models designed to assess the reserves and capital needed to support the risks of both individual credits and portfolios of credit instruments. A popular speci.cation for credit-rating transitions is the simple, time-homogeneous Markov model. While the Markov specification cannot really describe processes in the long run, it may be useful for adequately describing short-run changes in portfolio risk. In this specification, the entire stochastic process can be characterized in terms of estimated transition probabilities. However, the simple homogeneous Markovian transition framework is restrictive. We propose a test of the null hypotheses of time-homogeneity that can be performed on the sorts of data often reported. We apply the tests to 4 data sets, on commercial paper, sovereign debt, municipal bonds and S&P Corporates. The results indicate that commercial paper looks Markovian on a 30-day time scale for up to 6 months; sovereign debt also looks Markovian (perhaps due to a small sample size); municipals are well-modeled by the Markov specification for up to 5 years, but could probably bene.t from frequent updating of the estimated transition matrix or from more sophisticated modeling, and S&P Corporate ratings are approximately Markov over 3transitions but not 4.

　　40. A SFIR Approach to Financial Derivative Valuation

　　An innovative SFIR (stepwise filtration and regression) approach is proposed for financial derivative valuation. The SFIR method is a recursive semi-parametric approach which is applicable to computing the derivative prices when the transition probability function of the underlying process is known. Valuation of various financial derivatives including European options, American options, and convertible bonds can be solved by the SFIR method. The initial derivative values are obtained by an iterative procedure consists of two steps, the first involves approximating the payoff by a multi-piece regression function and the second involves computing the one-step backward payoff function by filtration. The proposed schemes are performed to compute the values of convertible bonds, European, and American options of the Black-Scholes, jumpdiffusion, and GARCH models. Approximation errors of the SFIR method and the backward filtration of the regression functions are derived for these models. Both the theoretical findings and the simulation results show the SFIR approach to be very tractable for numerical implementation and provides a unified and accurate method for pricing financial derivative.

　　41. The Numerical Stability of Runge-Kutta Methods for DDEs with Many Delays

　　This paper deals with the numerical stability analysis of implicit Runge-Kutta methods for the numerical solution of delay differential equations (DDEs) with many delays. The stability behavior of such methods is studied when they are applied to a test equation with many constant delays and complex coefficients. The concept of stability is introduced. It is proven that an implicit Runge-Kutta method is stable if and only if it is stable for ordinary differential equations (ODEs).

　　42. Robust Local Linear Regression Estimator for Dependent Spatial Processes

　　A robust version of local linear regression smoothers augmented with variable bandwidth is proposed and investigated for dependent spatial processes. The weak and strong consistency as well as asymptotic normality for the local linear M-estimator of the spatial regression function g(x) are established under some mild regularity conditions. Furthermore, an additive model is considered to avoid the curse of dimensionality for spatial processes and an estimation procedure based on combining the marginal integration technique with local linear M-estimator is developed.

　　43. AGGREGATION OF NONPARAMETRIC ESTIMATORS FOR VOLATILITY MATRIX

　　An aggregated method of nonparametric estimators based on time-domain and statedomain estimators is proposed and studied. To attenuate the curse of dimensionality, we propose a factor modeling strategy. We first investigate the asymptotic behaviors of nonparametric estimators of the volatility matrix in the time domain and in the state domain. The asymptotic normality is separately established for nonparametric estimators in the time domain and state domain. These two estimators are asymptotically independent. Hence, they can be combined, through a dynamic weighting scheme, to improve the efficiency of the estimated volatility matrix. The optimal dynamic weights are derived and it is shown that the aggregated estimator uniformly dominates the volatility matrix estimators using time-domain or state-domain smoothing alone. A simulation study, based on an essentially affine model for the term structure, is conducted and it demonstrates convincingly that the newly proposed procedure outperforms both time- and state-domain estimators. Empirical studies endorse further the advantages of our aggregated method.

　　44. Testing Asset Pricing Models in the Presence of Measurement Error: A Nonparametric Approach

　　This paper provides a uni.ed framework for the estimation and testing of the prominent asset pricing models studied in the literature. Most asset pricing models imply an Euler Equation that may be solved to obtain an intertemporal relation between the conditional expected excess return of stock market investment and its conditional variance. The main difficulty in testing the above relations is that the conditional variance of the market is not observable. Most of the existing literature is based on the estimation of parametric ARCH or stochastic volatility models for the underlying returns. We focus on an nonparametric measure of monthly return variability called realized volatility, which is easily computed from high-frequency intra-period returns. Also, since the conditional variance of the market is unobservable, using an estimate in the time-series regression suffers from a standard error-in-variables problem leading to inconsistency in the parameter estimates. Most of the existing literature ignores the measurement error problem and hence inference can be misleading. Our approach explicitly incorporates the measurement error in the volatility proxy while deriving the limiting distribution of the estimated parameters.

　　45. Nonparametric Methods for Estimating Conditional VaR and Expected Shortfall

　　In this article we propose a new nonparametric estimation method to estimate the conditional value-at-risk and expected shortfall functions based on the weighted double kernel local linear estimator of the conditional distribution function. The conditional value-at-risk is estimated by inverting the estimated conditional distribution function. The nonparametric estimator of the conditional expected shortfall is constructed by a plugging-in method. First, we establish the asymptotic normality and weak consistency of the weighted double kernel local linear estimator of the conditional distribution for time series data at both boundary and interior points. Also, we also show that the weighted double kernel local linear conditional distribution estimator not only preserves the bias, variance, and more importantly, automatic good boundary behavior properties of the double kernel local linear estimator and the weighted Nadaraya-Watson estimator, but also has the additional advantages of being always a distribution itself, continuity, and differentiability. Secondly, we show that the proposed weighted double kernel local linear estimators for both the conditional value-at-risk and expected shortfall are weakly consistent and normally distributed under the time series context at both boundary and interior points. Moreover, an automatic bandwidth selection method is proposed based on the nonparametric version of the Akaike information criterion. Finally, an empirical study is carried out to illustrate the performance of the proposed estimators.

　　46. Statistical Inference in CIR Stochastic Volatility Model

　　In this paper, the problem of estimating coefficients in Cox-Ingersoll-Ross stochastic volatility model are studied. The moment estimates of the stationary mean m and the stationary variance v of the CIR process is given. By assuming the parameters m and v “known”, the relation between the scale parameter α and the volatility β is obtained. The conditional moment estimate and the approximate maximum likelihood estimate are discussed. The simulation evidence indicates that the conditional moment estimate is more accurate then the approximate maximum likelihood estimate

　　47. Land of Addicts? An Empirical Investigation of Habit-Based Asset Pricing Models

　　This paper studies the ability of a general class of habit-based asset pricing models to match the conditional moment restriction simplied by asset pricing theory. We treat the functional form of the habit as unknown, and to estimate it along with the rest of the model’s finite dimensional parameters. Using quarterly data on consumption growth, assets returns and instruments, our empirical results indicate that the estimated habit function is nonlinear, that habit formation is better described as internal rather than external, and the estimated time-preference parameter and the power utility parameter are sensible. In addition, the estimated habit function generates a positive stochastic discount factor (SDF) proxy and performs well in explaining cross-sectional stock return data. We find that an internal habit SDF can explain a cross-section of size and book-market sorted portfolio equity returns better than (i) the Fama and French (1993) three-factor model, (ii) the Lettau and Ludvigson (2001b) scaled consumption CAPM model, (iii) an external habit SDF proxy, (iv) the classic CAPM, and (v) the classic consumption CAPM.

　　48. The Study on the Relationship between the trading volume and the price volatility based on the MDH Theory and the day-of-the-week effect

　　In this paper, we take the stock index of Shanghai and Shenzhen markets as the research object and introduce the trading volume?the trading volume considering the autocorrelation and the Day-of-the-week Effect into the GARCH model. The study finds that the trading volume has already had the explanation effect to the volatility of the stock index to a certain extent. But the trading volume considering the autocorrelation can’t explain the GARCH effect of the stock price effectively. The Day-of-the-week Effect has the function on the explanation which adds fuel to the flames regarding the trading volume to the stock price volatility.

　　49. Characteristic Function-Based Testing for Multifactor Continuous-Time Markov Models via Nonparametric Regression

　　We develop a nonparametric regression-based goodness-of-fit test for multifactor continuoustime Markov models using the conditional characteristic function, which often has a convenient closed form or can be approximated accurately for many popular continuoustime Markov models in economics and finance. By exploiting the Markov property, our omnibus test fully utilizes the information in the joint conditional distribution of underlying processes and hence is expected to have good power against a vast class of continuous-time alternatives in the multifactor framework. A class of easy-to-interpret diagnostic procedures is also proposed to gauge possible sources of model misspecifi- cation. All our tests have a convenient asymptotic N(0, 1) distribution under correct model specification. Simulations show that our tests provide reliable inference for sample sizes often encountered in empirical finance, and there is no need to use bootstrap procedures.

　　50. Feature and Source of Excess Return Volatility of Closed-end Fund

　　This paper examines if there is an excess return volatility of closed-end fund in a bear market. Excess volatility tests are used to compare the difference between the fund variance and the portfolio variance. Our main findings can be summarized as follows: There is an evidence of excess return volatility and the average closed-end fund’s weekly return is 59 percent more volatile than its assets. The excess volatility in fund return is not consistent across different asset size funds in different periods of observation. Closed-end fund returns sometimes under-react to returns on the underlying portfolio and sometimes overreact to returns on the underlying portfolio. There is clear evidence of a reduction in excess volatility in the late bear market. 43% . 52% of the average fund’s excess risk is explained by market risk? mall-firm risk? and risk that affects all closed-end funds.

　　51. An Empirical Study of Pricing and Hedging Collateralized Debt Obligation

　　This paper studies pricing collateral debt obligation (CDO) using Monte Carlo and the analytic method. Both methodologies are established within the framework of the reduced form model. One-factor Gaussian Copula is used for treating default correlation among collateral portfolio. Based on the two methods, the portfolio loss, the expected loss in each tranche, tranche spread and default delta sensitivity are analyzed with respect to different parameters such as maturity, default correlation, default hazard rate and recovery. The experiment shows that Monte Carlo method is slow and not robust in the calculation of default delta sensitivity. The analytic approach has comparative advantages for pricing CDO. Furthermore, the implied default correlation and base correlation are empirically studied.

　　52. Multifactor Non-Affine Term Structure Models

　　Finding conditional moments and asset prices in continuous-time diffusion models requires numeric methods if the diffusion does not fall within a relatively limited class of processes. For example, most term structure models in the literature fall within the affine class (see Duffie and Kan, 1996) or the linear-quadratic class (see Ahn, Dittmar, and Gallant 2002). Kimmel (2005) develops a technique for series approximation of conditional moments or asset prices in a large class of non-linear diffusion models, and shows that, for many such models, the approximations converge uniformly in time horizon (for conditional moments) or maturity (for bond prices). However, this method is restricted to scalar diffusions, or multiple diffusions with independent state variables. We extend this technique to a larger class of multiple diffusions, such that the series approximations converge to the conditional moment or bond pricing function for all time horizons or maturities. However, uniform convergence in time horizon or maturity typically requires introduction of multiple (artificial) time dimensions, and construction of power series approximations in these multiple time variables. The path of true time is then a curve through a higher-dimensional space in multiple artificial time dimensions. The restriction of the power series approximations (expressed as a function of the multiple time variables) to the embedded curve representing true time gives the true conditional moment or asset price function; furthermore, the power series approximations often converge uniformly for arbitrarily large time horizon or maturity. Bond prices for many multifactor term structure models currently in the literature may be approximated (uniformly in maturity) in this way, but the technique applies to many other multifactor models that have not previously appeared in the literature.

　　53. Estimation of Serially Correlated Microstructure Noise

　　Theory on quadratic variation which self can be as a good measure of volatility of returns shows that realized variance based on the effective (frictionless) stock prices can be used as an unbiased and consistent estimate of quadratic variation under certain mild conditions. However, the effective price is unobserved in the real market because it is contaminated by the so called microstructure noise and thus realized variance based on observed prices loses its good statistical properties, and even worse with increasing of sampling frequency. The existing disturbing of microstructure noise calls for noise-reduction technique. Based on the unbiased estimate of realized variance under serially correlated noise proposed by Hansen & Lunde (2004), we went a further step by deducing a measure of variance of noise. Such measure utilizes the differences between the traditional estimate of RV under noise-i.i.d. assumption and Hansen & Lunde’s unbiased estimate of RV under serially correlated noise assumption, all of which are estimated at different frequencies. We also proposed the methods of how to determine those sampling frequencies. Compared to what exist in most current literature that assume noise series are i.i.d., our measure based on a more loose assumption (serially correlated noise) and therefore is more realistic.

　　54. Nonparametric Estimation for the Conditional Mean and Variance Functions of Taiwan Stock Returns

　　In this paper, the CHARN (conditional heteroskedastic autoregressive nonlinear) model is applied to study the conditional mean and volatility functions for Taiwan’s stock returns from Jan. 3, 1991 to Nov. 30, 2005. The CHARN model is estimated with the nonparametric regression approach, particularly local linear estimation. Typically, a version of the plug-in idea for local linear estimator proposed by Ruppert, Sheather and Wand (1995) is used in this paper. For the whole sample period, the fitted conditional mean exhibits positive slope and is significantly from zero when yt.1 are in [.2 %,.0.5 %] and [0.5 %, 2 %]. This fact may be the result of some degree of market inefficiency. The estimated conditional variance function exhibits a U-shaped structure, also called a ”smiling face”, which means that risks of return are much higher for extreme values taken in the past days. For the segment of yt.1 in [.4 %, 4 %], the estimated conditional variance is larger when the lagged returns are negative and smaller when the lagged returns are positive. This supports the asymmetry of volatility caused by leverage effect, i.e., negative lagged return has larger impact on the subsequent volatility than positive lagged return does. To investigate the effects of two manifest events, asian financial crisis in 1997 and 911 air craft attack in 2001, on Taiwan’s stock market. The whole sample is divided into three sub-samples. One is before asian financial crisis (from Jan. 3, 1991 to July 1, 1997), another is between the asian financial crisis and 911 (from July 2, 1997 to Sep. 11, 2001), and the other is after the 911 air craft attack (from Sept. 13, 2001 to Nov. 30, 2005.) The CHARN model is estimated for these three designed sub-samples. Based on the estimated conditional mean functions, the efficient market hypothesis is hold for the first and third sub-samples. The estimated conditional variance function from the second sub-sample period is relatively different from the estimated ones from the first and third sub-sample periods and the whole sample period. These finding conclude that the impact of asian financial crisis on Taiwan’s stock market is larger than the 911 air craft attack does.

　　55. Tail-Adaptive Procedures of the Generalized Secant Hyperbolic Distribution: A Financial Application

　　The generalized secant hyperbolic distribution (GSHD) has been studied recently as a modeling tool in financial data analysis. The GSHD is completely specified by location, scale and shape parameters. It has also been shown elsewhere that the rank procedures of location are regular, robust, and asymptotically efficient on a certain range of the shape parameter. In this paper we demonstrate that the shape parameter may be understood as a tail weight parameter of the distribution, and introduce a three-test adaptive rank procedure of location based on various estimators of the tail weight of the GSHD. We then compare the results using adaptive non-parametric estimators with that of MLE typically used for GARCH modelling. This distributional assumption when applied to real data illustrates the different leptokurtosis that exist in the market between asset classes. Its closed form inverse cumulative distribution function also makes GSHD useful in predicting financial risk quantiles.