Estimation, Testing, and Model Selection in Semiparametric Copula-based Multivariate Dynamic Models

Vanderbilt University, Nashville TN

Investigators

Abstract

Because of their skewed and unusual distribution, financial time series modeling pose challenging problems that require the analyst to model the entire conditional distributions. This is especially true if the series is multivariate, possibly nonlinear, time series. One set of models that mimic the distribution of financial time series very well are copula-based models. An important development in econometrics is the emergence of nonparametric and semiparametric methods because they impose few restrictions on parameter space as compared to parametric estimation. Although practitioners in finance and insurance have long used copulas to model multivariate option pricing, portfolio value-at-risk, correlated default and credit risk, contagion, and the time-varying asymmetric non-linear co-movements among different series, existing methods fail to successfully address the issues of estimation, testing, and model selection in the semiparametric multivariate time series via copulas. This proposal consists of several projects on the specification, estimation, testing, and model selection of a new broad class of Semiparametric Copula-based Multivariate Dynamic (SCOMDY) models. The model is based on the idea that all macroeconomic data are non-linear beyond the second moment. The modeling combines parametric and nonparametric methods to reduce the problems of dimensionality inherent in nonparametric methods. This is an innovative and ambitious approach to modeling the conditional distribution of multivariate time series and the results would advance our understanding of the behavior of financial and macroeconomic time series. The proposal offers a flexible and practically feasible way to model the entire multivariate conditional distributions of multiple economic and financial time series that typically have nonlinear, asymmetric co-movements, and fat tails. The research outputs will make significant, original contributions to the current econometrics and statistics literature on the estimation, testing and model selection of semiparametric copula-based multivariate models. The outputs of this proposal will guide applied researchers and practitioners to perform statistically reliable economic policy evaluations, financial forecasts, and risk managements. This research makes an important contribution to the estimation of multivariate financial and macroeconomic time series.