The program of the RDTSE

 

Wang Yanan Institute for Studies in Economics (WISE) has been set up as an international-oriented institute. It signifies Xiamen University’s aim to maintain its leading position in economic research in the nation, and aspires to become an international center for research and teaching excellence. Thanks to the contribution of the late president, Professor Wang Yanan, who was first to translate Adam Smith’s The Wealth of Nations and Karl Marx’s On Capital into Chinese, economics in Xiamen University has always been among the top programs in China.

WISE focuses intensively on high-quality postgraduate training, frontier research, and international academic interactions and cooperation. WISE is particularly inspired to achieve the following:

       Publishing research results in top domestic and international academic journals;

       Enhancing the curriculum design in economic studies at XMU, including Econometrics, Finance, Macroeconomics, Labor Economics and Political Economy;

       Striving to excel WISE in research and nurture a team of young economists and econometricians with academic leadership;

       Taking up “think tank” role to provide first-class policy analysis and consultancy national and regional governments;

       Establishing itself as an influential regional center in Asia-Pacific for international academic exchange in economics and finance.

WISE has since set up the following academic research centers, including the Center for Econometrics Research, the Center for Financial Research, the Center for Macroeconomics Research, the Center for political Economy Research, the Center for Statistics Research, China Center for Research in Labor Economics and Social Security, SAS Center for Excellence in Econometrics, Cornell-XMU Exchange Center, and XMU-SMU Research Center for China Capital Market. At the current stage, WISE has chosen econometrics, finance, and labor economics as priority fields in its development strategy.

WISE currently has twenty-six faculty members, including eleven full-time faculty recruited from prestigious universities overseas and four part-time faculty from the U.S. who visit WISE regularly each year. With a full capacity of fifty, WISE is seeking high-quality researchers in all areas in economics and finance at all ranks, both in international and domestic job hunting markets.

The Centre for Forecasting Science (CEFS) was founded in January 2006, which is subordinate to academy of mathematics and system science of Chinese academy of Science. CEFS consists of four research departments: Farm Products Forecasting Department, Strategy Resources Forecasting Department, Macroeconomic Forecasting Department and International Market Forecasting Department.

CEFS has a profound tradition in basic disciplines and predominant advantages in interdisciplinary research. Most of great forecast issues in national economy are researched. All of these researches provide important suggestions to the government in decision making.

The missions of CEFS are: (1) to research the important forecasting problems in China’s economic and social development and provide better evidences for the government; (2) to develop forecasting theories and technologies and achieve original contributions in forecasting science.

The center’s target is to become the main research centre of forecasting science in China, to become one of the influential international research institution in forecasting science and also become one of the main base to incubate first-rate talents of forecasting science.

Please refer to the Conference Agenda for a more detailed timetable of presentations. This guide primarily concerns with Reception, Accommodation, Registration, Meals, Transportation and Ticket-booking arrangements during the conference.

The Conference Committee has arranged transportations for all the presenters at the airport, which requires the presenters to provide accurate arrival/departure information to/from Xiamen. At the arrival, please pay attention to the “RDTSE 2008” poster to get the proper transportation arrangement.

1)      For all participants, accommodation is provided by Yi Fu Building (campus hotel, +86-592-2087988). The registration time is on 9th May and the Venue is at the lobby of Yi Fu Building, Xiamen.

2)      The Conference only accounts for the Speakers’ daily rent in the hotel (3 nights) while our guests would pay for any incidental cost incurred otherwise.

1) 9th May	Supper	6:00PM	Yi Fu Building
2) 10th May	Breakfast		Yi Fu Building
	Lunch	12:20PM	Yi Fu Building
	Supper	6:00PM	Seafood Restaurant
3) 11th May	Breakfast		Yi Fu Building
	Lunch	12:10PM	South Pu Tuo Temple(Vegetarian Food)
	Supper	6:00PM	To be determined
4) 12th May	Breakfast		Yi Fu Building

(Locations are subject to change; please refer to the announcement made by the Conference Reception.)

You may contact the following agencies for booking affairs.

MF：2226666	CZ：5117777；6022936-7558
FM：2210600	MU：5732401
CA：5084383；5084375	SC：5555555；5562377
KA：5117702	CX：00852-27471888
MH：2106088；2106188	PR：2394729；2394730
MI：2053275；2037127	KA：2680140；2680141
SQ：2053275	NH：2052317；5732888

a) Academic affairs:	Zongwu CAI	2186328
	CY SIN	2185729
	Ying FANG	2181763
b) Publicity:	Sophie SONG	2180855
c) Reception:	Iris ZHANG	2187878
	Daisy XU	2181003
	Jeoly BAO	2181269
	Vicky LI	2186170
d) Conference-related affairs:	Roki LUO	2186137
	Jacky CHEN	2182991
	Will ZHANG	2182185
e) Treasury:	Iris ZHANG	2187878

6.       A half-day trip around Xiamen will be specially offered on 12th May by the Committee which will guide our guests around Gu Langyu Island (the Island of Architectures and Piano) and the Ring Road. For those guests who are interested in this trip, please register during your registration. For those who sighed up then, please gather at 8:00am, 12th May at Lobby at Yi Fu Building; the trip will last from 8:00AM to 1:00PM.

7.       Please present your conference ID card when you have lunch and supper at the appointed restaurants.

8.       Please turn your cell phone into silence or vibration mode when the conference is in progress.

9.       Photography can be downloaded in the homepage of the Wang Yanan Institute for Studies in Economics (WISE) at www.wise.xmu.edu.cn .

10.   Computers with access to the internet are provided for the convenience of Presenters.

11.   For clubbing and pubs, please go to Bin Lang Xiao Qu (Bin Lang Community) at Hu Bin Dong Lu (Eastern Hu Bin Road).

12.   For safety cautions: Although Xiamen is one of the most comfortable and safest cities in China, we suggest you take good care of yourself and your belongings. A company is better than alone when you are going out at night. Additionally, we strongly suggest you not to swim in the sea.

 

 

 

                                          Wang Yanan Institute for Studies in Economics

                                                          May, 2008

Theme: Nonlinear Time Series Econometrics: Theory and Applications

May 10-11, 2008

Wang Yanan Institute for Studies in Economics (WISE), Xiamen University, China

Center for Forecasting Science (CEFS), Chinese Academy of Sciences, China

Journal of Econometrics

 

Venue: Room 201, Qun-Xian Building II

Saturday, May 10
Session A: 8:30am-8:45am: Opening Ceremony
Chair: Chor-Yiu Sin, Xiamen University, China, cysinhkbu@gmail.com

 

Yongmiao Hong, Cornell University, USA

Cheng Hsiao, University of Southern California, USA and Xiamen University, China

Xiaoguang Yang, Chinese Academy of Sciences, China

 

Chair: Cheng Hsiao, University of Southern California, USA and Xiamen University, China, chsiao@usc.edu

 

Speaker: Peter M. Robinson, London School of Economics, UK, p.m.robinson@lse.ac.uk
"Issues in Semiparametric Modeling of Multivariate Long Memory Time Series"

9:30am-9:40am

Open Discussions

 

Session C: 9:40am-10:20am: Keynote Session I (Cross-Sectional Independence)

Chair: Xiaotong Zhang, Nankai University, China, xttfyt@public.tpt.tj.cn

Speaker: Cheng Hsiao, University of Southern California, USA and Xiamen University, China, chsiao@usc.edu

M. Hashem Pesaran, University of Cambridge, UK and University of Southern California, USA

Andreas Pick, University of Cambridge, UK

"Diagnostic Tests of Cross Section Independence for Nonlinear Panel Data Models"

 

Session D: 10:50am-12:20pm: Nonlinear Modeling for Nonstationary Time Series
Chair: Zongwu Cai, University of North Carolina at Charlotte, USA , zcai@uncc.edu

 

10:50am-11:20am
Speaker: Yoosoon Chang, Texas A&M University, USA, yoosoon@tamu.edu

"Endogeneity in Nonlinear Regressions with Integrated Time Series"

 

11:20am-11:50am

Speaker: Chien Ho Wang, National Taipei University, Taiwan,

Robert M. de Jong, Ohio State University, USA

"Asymptotics for Scaled Periodic Transformations of Integrated Time Series"

 

11:50am-12:20pm

Speaker: Zongwu Cai, University of North Carolina at Charlotte, USA , zcai@uncc.edu

Qi Li and Joon Y. Park, Texas A&M University, USA

"Functional-Coefficient Models for Nonstationary Time Series Data"

 

Chair: Xiaoguang Yang, Chinese Academy of Sciences, China, xgyang@iss.ac.cn

 

2:00pm-2:40pm
Speaker: Anil Bera, University of Illinois at Urbana-Champaign, USA,

"Birth of a Nonlinear Model"

2:40pm-3:20pm
Speaker: Viktor Todorov, and George Tauchen, Duke University, USA, get@econ.duke.edu

"Volatility Jumps"

 

Chair: Ying Fang, Xiamen University, China, yifst1@gmail.com

 

3:50pm-4:20pm

Speaker: Bin Chen, University of Pittsburgh, USA, binchen@pitt.edu

Yongmiao Hong, Cornell University, USA

"A Unified Approach to Validating Univariate and Multivariate Conditional Distribution Models in Time Series"

 

4:20pm-4:50pm
Speaker: Yoon-Jin Lee, Indiana University, USA, lee243@indiana.edu

Yongmiao Hong, Cornell University, USA

"A Loss Function Approach to Specification Testing and Its Relative Efficiency to the GLR Test"

 

4:50pm-5:20pm
Speaker: Liangjun Su, Peking University, China, lsu@gsm.pku.edu.cn

Halbert White, University of California, San Diego, USA

"Testing Structural Change in Partially Linear Models"

 

5:20pm-5:50pm
Speaker: Ying Fang, Xiamen University, China, yifst1@gmail.com

"Instability Test and Nonparametric STAR Model with an Application to Chinese Macroeconomic Time Series"

 

 

 

Chair: Chor-Yiu Sin, Xiamen University, China, cysinhkbu@gmail.com

 

8:30am-9:10am

Speaker: Yiguo Sun, University of Guelph, Canada

Zongwu Cai, University of North Carolina at Charlotte, USA

Qi Li, Texas A&M University, USA, qi@econmail.tamu.edu

"Consistent Nonparametric Test on Parametric Smooth Coefficient Model with Nonstationary Data"

 

9:10am-9:50am

Speaker: Valentina Corradi1, University of Warwick, UK

Andres Fernandez and Norman R Swanson, Rutgers University, USA, nswanson@econ.rutgers.edu

"Information in the Revision Process of Real-Time Datasets"

 

 

Session H: 10:20am-11:50am: Nonlinear Testing for Nonstationary Time Series
Chair: Jiti Gao, The University of Adelaide, Australia, jiti.gao@adelaide.edu.au

10:20am-10:50am
Speaker: Mehmet Caner, North Carolina State University, USA, mcaner@ncsu.edu

Keith Knight, University of Toronto, Canada

"No Country for Old Unit Root Tests: Bridge Estimators Differentiate between Nonstationary versus Stationary Models and Select Optimal Lag"

 

10:50am-11:20am

Speaker: Jiti Gao, The University of Adelaide, Australia, jiti.gao@adelaide.edu.au

11:20am-11:50am

 

Session I: 2:00pm-3:20pm: Keynote Session IV (Continuous Time Series and Long Memory Nonlinear Time Series Model)
Chair: Shengchu Pan, The Central University of Finance and Economics, China, pan_shengchu@263.net

 

2:00pm-2:40pm

Speaker: Michael McAleer, University of Western Australia, Australia,

Marcelo C. Medeiros, Pontifical Catholic University of Rio de Janeiro, Brazil

"A Multiple Regime Smooth Transition Heterogeneous Autoregressive Model for Long Memory and Asymmetries"

 

2:40pm-3:20pm

Speaker: Joon Y. Park, Texas A&M University, USA, jpark@tamu.edu

"Martingale Regression and Time Change"

 

Session J: 3:50pm-5:50pm: Financial Time Series

Chair: Sungyong Park, Xiamen University, China, park.sungyong@gmail.com

3:50pm-4:20pm
Speaker: Hai Lin, Xiamen University, China, cfc@xmu.edu.cn, Junbo Wang,

Chunchi Wu, University of Missouri, Columbia, USA

"Price Discovery and Trading After Hours in the U.S. Treasury Market"

 

4:20pm-4:50pm

Speaker: Chor-Yiu Sin, Xiamen University, China, cysinhkbu@gmail.com

"Stationarity, Ergodicity and Finite Moments of the Log-ACD, and Comparisons with the Power ACD"

4:50pm-5:20pm
Speaker: Wei Sun, University of Karlsruhe, Germany, sun@statistik.uni-karlsruhe.de

Stoyan V. Stoyanov, FinAnalytica, Inc., USA,

Frank J. Fabozzi, Yale School of Management, USA

"Analyzing Nonlinear and Asymmetric Dependence in German Equity Market: Approach with Multivariate Skewed Student’s t Copula"

 

5:20pm-5:50pm
Speaker: Sungyong Park, Xiamen University, China, park.sungyong@gmail.com

"Conditional Value at Risk and alpha-Risk under Regression Quantiles"

 

Session K: 5:50pm-6:00pm: Closing Remarks

Chair: Zongwu Cai, University of North Carolina at Charlotte, USA

 

Anil K. Bera, University of Illinois at Urbana Champaign, abera@ad.uiuc.edu

Abstract: Appearance of nonlinearity is the rule rather than the exception in economic models since the effects of some economic variables on others are simply not additive. Econometricians, however, concentrated their efforts primarily in developing and analyzing linear models, at least during the first four decades since the inception of the field of econometrics in 1920s. Possibly, the first formal nonlinear model in econometrics literature is the constant elasticity of substitution (CES) production function that appeared in 1961. As we all know, this nonlinear function has found its use far beyond the production function literature. For example, in general equilibrium analysis, preferences are usually presented by the CES utility function. This is also adopted in the international trade theory where “countries” take the place of “agents,” and “endowments” are replaced by “production levels,” Chipman 2008. Beyond economics, the much used Box-Cox (1964) transformation in the statistics literature has essentially the CES form. Advent of CES function led applied econometricians to use linear approximations; examples of this practice are the use of translog, generalized Leontief  and Rotterdam linear system in place of the CES function. Effects of linear approximation on estimation and testing when a non-linear model would be more appropriate are discussed in White 1980, Byron & Bera 1983. Recently suggested nonlinear autoregressive conditional heteroskedasticity (NARCH) models (Higgins and Bera, 1992, and many articles following that) have its origin in the Box-Cox transformation. More recently, the literature on model selection, maximum entropy (ME) and empirical likelihood (EL) estimation start with the Cressie-Read (1998) power divergence (CRPD) family which turned out to be a Box-Cox transformation of Karl Pearson (1900) chi-squared distance measure. However, “nothing” under the sun is completely new; every beginning has its own beginning. We can go back as far as Hardy, Littlewood and Polya (1934) to find an early expression of the CES function, what they called “mean value function of certain order.” In this historical note, I will cover many things and seemingly unrelated connections, but will concentrate on the origin of Arrow, Chenery, Minhas and Solow (1961), The Review of Economics and Statistics, that derived the CES form from empirical economic evidence and the “first principle” of economic theory.

Zongwu Cai, University of North Carolina at Charlotte, USA, zcai@uncc.edu

Qi Li and Joon Y. Park, Texas A&M University, USA

Abstract: This paper studies functional coefficient regression models with nonstationary time series data, allowing also for stationary covariates. A local linear fitting scheme is developed to estimate the coefficient functions. Both the consistency and asymptotic distributions of the estimators are obtained, showing different convergence rates for the stationary and nonstationary covariates. A two-stage approach is proposed to achieve estimation optimality. When the coefficient function is a function of a nonstationary variable, the new findings are that asymptotic bias is the same as stationarity covariate case but convergence rate differs, and further, the asymptotic distribution is mixed normal, associated with the local time of a standard Brownian motion. The asymptotic behavior at boundaries is also investigated.

Mehmet Caner, North Carolina State University, USA, mcaner@ncsu.edu

Keith Knight, University of Toronto, Canada

Abstract: This paper develops Bridge estimators in a mixed model of  possibly nonstationary and stationary variables. We show that these estimators differentiate between stationary and nonstationary variables. This both adds to unit root literature as well as model selection. Instead of testing, estimation is used to choose between nonstationary versus stationary cases. It can also simultaneously select the optimal lag. This prevents pretesting/lag selection before unit root testing. In terms of model selection literature, we extend the results on all stationary models to mixed models. We show that existing model selection ideas do not solve this problem, and we introduce an extension that achieves the oracle property.

Yoosoon Chang, Texas A&M University, USA, yoosoon@tamu.edu

Abstract: This paper considers the nonlinear regression with integrated regressors that are contemporaneously correlated with the regression error. We, in particular, establish the consistency and derive the limiting distribution of the nonlinear least squares estimator under such endogeneity for the regressions with the integrable or asymptotically homogeneous regression function. For the regressions with both the integrable and asymptotically homogeneous regression functions, it is shown that the estimator is consistent and has the same rate of convergence as for the case of the regressions with no endogeneity. Whether or not the limiting distribution is affected by the presence of endogeneity, however, depends upon the type of the regression function. If the regression function is asymptotically homogeneous, the limiting distribution of the least squares estimator has an additional term reflecting the presence of endogeneity. On the other hand, the endogeneity does not have any effect on the least squares limit theory, if the regression function is integrable. Regardless of the presence of endogeneity, the least squares estimator has the same limiting distribution in this case. To illustrate our theory, we consider the nonlinear regressions with logistic and power regression functions with integrated regressors that have contemporaneous correlations with the regression error.

Bin Chen, University of Pittsburgh, USA, binchen@pitt.edu

Yongmiao Hong, Cornell University, USA

Abstract: Modeling conditional distributions in time series has attracted increasing attention in economics and finance. We develop a new class of omnibus specification tests for time series conditional distribution models using a novel approach, which embeds the distribution function in a spectral framework. Our tests check a large number of lags and are therefore expected to be powerful against neglected dynamics at higher order lags, which is particularly useful for non-Markovian processes. Despite using a large number of lags, our tests do not suffer much from loss of a large number of degrees of freedom, because our approach naturally downweights higher order lags, which is consistent with the stylized fact that economic or financial markets are more affected by recent past events than by remote past events. Unlike the existing methods in the literature, the proposed omnibus tests cover both univariate and multivariate conditional distribution models in a unified framework. They fully exploit the information in the joint conditional distribution of underlying economic processes. Moreover, a class of easy-to-interpret diagnostic procedure is supplemented to gauge possible sources of model misspecifications. Distinct from Cramer-von Mises and Kolmogorov-Smirnov tests, which are also based on the distribution function, our test statistics all follow a convenient asymptotic N (0; 1) distribution and enjoy the appealing “nuisance parameter free” property that parameter estimation uncertainty has no impact on the asymptotic distribution of the test statistics. Simulation studies show that the tests provide reliable inference for sample sizes often encountered in economics and finance.

Ying Fang, Xiamen University, China, yifst1@gmail.com

Abstract: This research proposes an instability test based on the comparison between a parametric fixed coefficients fitting and a nonparametric time-varying coefficient fitting. We formally test a sample of 85 monthly Chinese macroeconomic time series and their 7140 bivariate relations. Our results suggest time-varying coefficients models. We then propose a nonparametric index-nonlinear STAR model.

Bin Chen, University of Pittsburgh, USA

Yongmiao Hong, Cornell University, USA, yh20@cornell.edu

Abstract: We develop a nonparametric regression-based goodness-of-fit test for multifactor continuous-time Markov models using the conditional characteristic function, which often has a convenient closed-form or can be approximated accurately for many popular continuous-time Markov models in economics and finance. Our omnibus test fully utilizes the information in the joint conditional distribution of underlying processes and hence is consistent 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 misspecifications. All our tests have a convenient asymptotic N(0,1) distribution under correct model specification. Simulations show that our tests have reasonable sizes in finite samples, and good power against several alternatives, including misspecifications in the joint dynamics even if the dynamics of individual components is correctly specified. This feature is not attainable by some existing tests.

Cheng Hsiao, University of Southern California, USA and Xiamen University, China, chsiao@usc.edu

M. Hashem Pesaran, University of Cambridge, UK and University of Southern California, USA

Andreas Pick, University of Cambridge, UK

Abstract: We show that the Lagrangian multiplier test of cross-sectional independence test for nonlinear linear models is equivalent to testing cross-sectional correlation coefficients of the generalized residuals (e.g. Gourieoux, Monfort, Renault and Trognon (1987)) are equal to zero. In nonlinear models, the definition of the residual sometimes could be ambiguous and we consider two approaches: deviations of the observed dependent variables from their expected values and generalized residuals. We consider two tests based on these definitions of residuals. One is based on the average pair-wise residual correlation coefficients as Pesaran’s (2004) CD test. The other is based on the average of the squared correlation coefficients as the Lagrangian multiplier test in the linear case. In Monte Carlo experiments it emerges that the CD test has the correct size for any combination of N and T whereas the test based on the square of the correlation coefficients relies on T large relative to N. We illustrate the use of cross-sectional dependence test by considering the roll-call votes of the 104th U.S. Congress and find considerable dependence between the votes of the members of Congress.

Jiti Gao, The University of Adelaide, Australia, jiti.gao@adelaide.edu.au

Abstract: This paper considers a class of nonlinear time series econometric models with nonstationarity. We propose a nonparametric kernel test for the conditional mean and then establish an asymptotic distribution of the proposed test. Both the setting and the results differ from earlier work on existing results with stationarity. In addition, we develop a new bootstrap simulation scheme for the selection of a suitable bandwidth parameter involved in the kernel test as well as the choice of a simulated critical value. The finite-- sample performance of the proposed test is assessed using one simulated example and one real data example.

Yoon-Jin Lee, Indiana University, USA, yoonjin_lee@hotmail.com

Yongmiao Hong, Cornell University, USA

Abstract: The generalized likelihood ratio (GLR) test has been proposed by Fan, Zhang and Zhang (2001) as a generally applicable method to test parametric, semiparametric or nonparametric models against nonparametric alternative models. It is a natural extension of the maximum likelihood ratio test for parametric models and fully inherits the advantages of the classical likelihood ratio test. Both true likelihood and pesudo likelihood functions can be used. Like the classical likelihood ratio test, the GLR test enjoys the appealing Wilks phenomena in the sense that its asymptotic distribution is independent of nuisance parameters and nuisance functions, and follows a chi-squared distribution with a known large number of degrees of freedom. It achieves the asymptotically optimal rate of convergence for nonparametric testing problems formulated by Ingster (1993a, 1993b) and Spokoiny (1996). In this paper, we propose a class of new tests based on loss functions, which measure the discrepancies between the fitted values of the null and alternative models, and are more relevant to economic applications. Like the GLR test, the new tests are generally applicable and enjoy many appealing features of the GLR test, such as the Wilks phenomena. Most importantly, they are asymptotically more powerful than the GLR test in terms of Pitman’s relative efficiency criterion. This holds even when the true likelihood function is available.

Yan He, Hai Lin, Xiamen University, China, cfc@xmu.edu.cn, Junbo Wang,

Chunchi Wu, University of Missouri, Columbia, USA

Abstract: We evaluate the efficacy of price discovery in the after-hours market using a comprehensive intraday transaction database of U.S. Treasury securities. Using this dataset allows us to control the trading process to ascertain the pure informational role of trades over the 24-hour day. We find that information asymmetry is generally the highest in the preopen period and the lowest in the post-close period whereas information asymmetry in the overnight period is comparable to that in the regular trading period. However, on days with macroeconomic announcements, information asymmetry peaks shortly after the news release at 8:30. Moreover, information asymmetry is higher immediately before than after the opening of U.S. Treasury futures trading. Although volume is low after hours and trading cost is relatively high, overnight trading generates significant price discovery. Results suggest that overnight trading activity is an important part of the Treasury price discovery process.

Yiguo Sun, University of Guelph, Canada

Zongwu Cai, University of North Carolina at Charlotte, USA

Qi Li, Texas A&M University, USA, qi@econmail.tamu.edu

Abstract: In this article, we propose a simple nonparametric test for testing the null hypothesis of constant coefficients against nonparametric smooth coefficients in a varying coefficient model with nonstationary data. We establish the asymptotic distributions of the proposed test statistic under both null and alternative hypotheses. A Monte Carol simulation is conducted to illustrate the finite sample performance of the proposed test statistic.

Michael McAleer, University of Western Australia, Australia, michael.mcaleer@gmail.com

Marcelo C. Medeiros, Pontifical Catholic University of Rio de Janeiro

Abstract: In this paper we propose a flexible model to capture nonlinearities and long-range dependence in time series dynamics. The new model is a multiple regime smooth transition extension of the Heterogenous Autoregressive (HAR) model, which is specifically designed to model the behavior of the volatility inherent in financial time series. The model is able to describe simultaneously long memory, as well as sign and size asymmetries. A sequence of tests is developed to determine the number of regimes, and an estimation and testing procedure is presented. Monte Carlo simulations evaluate the finite-sample properties of the proposed tests and estimation procedures. We apply the model to several Dow Jones Industrial Average index stocks using transaction level data from the Trades and Quotes database that covers ten years of data. We find strong support for long memory and both sign and size asymmetries. Furthermore, the new model, when combined with the linear HAR model, is viable and flexible for purposes of forecasting volatility. KEYWORDS: Realized volatility, smooth transition, heterogeneous autoregression, financial econometrics, leverage, sign and size asymmetries, forecasting, risk management, model combination.

Joon Y. Park, Texas A&M University, USA, jpark@tamu.edu

Abstract: In the paper, we develop a general methodology to estimate and test for a conditional mean model given in continuous time. Our model specifies the conditional mean of instantaneous change of a given stochastic process as a function of other predictable covariates. The model yields a continuous time regression for the instantaneous change of an underlying process on its conditional mean change with the error process given by a general martingale. We call it a martingale regression, since the parameter in the model is identified by the residual process being a martingale. Upon an appropriate time change, the martingale regression can always be transformed into a regression with the error process given by Brownian motion. We use this property and apply a minimum distance method to estimate the parameters in the model. More specifically, the samples are collected at random time intervals so that the errors become independent normals, and the estimates are defined as the parameter values which make the empirical distribution of the residuals closest to independent normals. We include several illustrative examples, for which the proposed methodology can be useful.

SungYong Park, Xiamen University, China, park.sungyong@gmail.com

Abstract: Quantile regression model provides excellent tool to estimate well known risk measure. In this paper, a simple version of quantile generalized autoregressive conditional heteroskedasticity (QGARCH) model is proposed to estimate conditional Value at Risk. Because of high degree of nonlinearity, implementation of quantile regression into GARCH-type model is complicate. We construct QGARCH model using the statistical characteristics of the asymmetric Laplace distribution. An application to the index stock returns illustrates the usefulness of our approach.

Peter M.Robinson, LSE, UK, p.m.robinson@lse.ac.uk

Abstract: Moving from univariate to bivariate jointly dependent long memory time series introduces a phase parameter (ャ), at the frequency of principal interest, zero; for short memory series ャ=0 automatically.   The latter case has also been stressed in the long memory case, along with the “fractional differencing” case ャ=(ヤ₂- ヤ₁)ヰ/2, whereヤ₁, ヤ₂ are the memory parameters of the two series. We develop time domain conditions under which these are and are not relevant, and relate the consequent properties of cross-autocovariances to ones of the (possibly bilateral) moving average representation which, with martingale difference innovations of arbitrary dimension, is used in asymptotic theory for local Whittle parameter estimates depending on a single smoothing number. Incorporating also a regression parameter (モ) which, when non-zero, indicates cointegration, the consistency proof of these implicitly-defined estimates is nonstandard due to the モ estimate converging faster than the others. We also establish joint asymptotic normality of the estimates, and indicate how this outcome can apply in statistical inference on several questions of interest.  Issues of implemention are discussed, along with implications of knowing モ and of correct or incorrect specification of ャ, and possible extensions to higher-dimensional systems and nonstationary series.

Liangjun Su, Peking University, China, lsu@gsm.pku.edu.cn 

Halbert White, UCSD, USA

Abstract: We consider two tests of structural change for partially linear time-series models. The first tests for structural change in the parametric component, based on the cumulative sums of gradients from a single semiparametric regression. The second tests for structural change in the parametric and nonparametric components simultaneously, based on the cumulative sums of weighted residuals from the same semiparametric regression. We derive the limiting distributions of both tests under the null hypothesis of no structural change and for sequences of local alternatives. We show that the tests are free of nuisance parameters asymptotically under the null. Our tests thus complement the conventional instability tests for parametric models. To improve the finite sample performance of our tests, we also propose a wild bootstrap version of our tests and justify its validity. Finally, we conduct a small set of Monte Carlo simulations to investigate the finite sample properties of the tests.

Chor-yiu (CY) Sin, Xiamen University, China, cysinhkbu@gmail.com

Abstract: In this paper, we consider primitive assumptions that suffice for stationarity, ergodicity, and finite moments of the logarithm autoregressive conditional duration (Log-ACD) model. Two types of models, namely (a) Box-Cox ACD model; and (b) Log-ACD model type I, are considered. The former type is said to be a limiting case of the latter by Fernandes and Grammig (2006). While ACD model resembles the GARCH model, the Log-ACD model resembles the Exponential GARCH (EGARCH) model. As pointed out by Nelson (1991), the probabilistic properties of this kind of models need to be discussed with primitive distributional assumptions of the standardized error. In this paper, four families of distributions, namely (i) generalized gamma; (ii) Burr; (iii) Pareto type I; and (iv) Pareto type II, are considered. Our primitive assumptions complement those proposed by Carrasco and Chen (2002) and Fernandes and Grammig (2006). Interestingly, for the Box-Cox ACD model with generalized gamma distributed error, finite moments require not only the parametric restrictions on the model, but also those on the distributions. The result contrasts to that of the power ACD model (with Engle and Russell (1998)’s ACD model as a leading case), in which only finite moments of the standardized error is assumed.

Wei Sun and Svetlozar Rachev, University of Karlsruhe, Germany, sun@statistik.uni-karlsruhe.de

Stoyan V. Stoyanov FinAnalytica, Inc., USA,

Frank J. Fabozzi, Yale School of Management, USA

Abstract: Analyzing comovements in equity markets is important for risk diversification in portfolio management. Copulas have several advantages compared to the linear correlation measure in modeling comovement. This paper introduces a copula ARMA-GARCH model for analyzing the comovement of indexes in German equity markets. The model is implemented with an ARMA-GARCH model for the marginal distributions and a copula for the joint distribution. After goodness of fit testing, we find that the skewed Student’s t copula ARMA(1,1)-GARCH(1,1) model with Lⅴevy fractional stable noise is superior to alternative models investigated in our study where we model the simultaneous comovement of six German equity market indexes. This model is also suitable for capturing the long-range dependence, tail dependence, asymmetric correlation observed in German equity markets.

Valentina Corradi1, University of Warwick, UK

Andres Fernandez and Norman R Swanson, Rutgers University, USA, nswanson@econ.rutgers.edu

Abstract: In this paper we first develop two statistical tests of the null hypothesis that early release data are rational. The tests are consistent against generic nonlinear alternatives, and are conditional moment type tests, in the spirit of Bierens (1982,1990), Chao, Corradi and Swanson (2001) and Corradi and Swanson (2002). We then use this test, in conjunction with standard regression analysis in order to individually and jointly analyze a real-time dataset for money, output, prices and interest rates. All of our empirical analysis is carried out using various variable/vintage combinations, allowing us to comment not only on rationality, but also on a number of other related issues. For example, we discuss and illustrate the importance of the choice between using first, later, or mixed vintages of data in prediction. Interestingly, it turns out that early release data are generally best predicted using real time datasets constructed using all of the latest available information, much as is done in the Federal Reserve models, for example. Thus, the standard practice of using “mixed vintages” of data when constructing prediction models appears to yield the “best” predictions. One reason that simpler models based solely upon the use of first release data may not perform better is the existence of “definitional change problems” associated with using only first (or later) releases of data for prediction. Furthermore, we note that our tests of first release rationality based on ex ante prediction find strong evidence that the data rationality null hypothesis is rejected for a variety of variables, in accord with recent findings by Aruoba (2006), for example. Moreover, nonlinear data irrationality appears to characterize money, as linear rationality tests fail to reject for our money dataset. Output and price datasets reject the null of rationality regardless of whether linear or nonlinear tests are implemented. We argue that the notion of final data is misleading, and that definitional and other methodological changes that pepper real-time datasets are important. Finally, we carry out an empirical illustration, where little evidence that money has marginal predictive content for output is found, regardless of whether various revision error variables are added to standard real-time vector autoregression models of money, output, prices and interest rates.

Viktor Todorov, George Tauchen, Duke University, USA, george.tauchen@duke.edu

Abstract: This paper conducts a non-parametric analysis of the fine structure of market volatility moves using high-frequency data on the VIX index. We find that the stock market volatility is a very “active” process, i.e. it evolves with a lot of “small” moves and occasional “big” jumps. However, our empirical results suggest that jump-diffusion volatility models are misspecified and pure-jump models are better suited for the purposes of volatility modeling. We also discover that most of the market volatility jumps arrive at the same time as the jumps on the stock market and there is a strong dependence between the two.

Chien-Ho Wang, National Teipei University, Taiwan, wangchi3@mail.ntpu.edu.tw

Robert M. de Jong, Ohio State University, USA

Abstract: In this paper we expand recent results of de Jong (2000). We consider time series regression when the regressors are periodic transformation on scaled I(1) process. When transformation is periodic function on scaled I(1) process with symmetric distribution, the convergence rate will be slower than de Jong (2000). The methods developed in this paper can be useful for developement of periodic transformation of nonlinear regression for nonstationary time series.