| The integration of economic globalization,financial integration and internet technology,as well as the strategy of "One Belt and One Road" has made the modes of financial investment and the portfolio move toward an internationalized and diversified trend.At the mean time,the internationalization of RMB and the liberalization of capital have accelerated the arrival of portfolio in globalwide.So far,the research on international portfolio investment mainly focuses on the correlation analysis of stock markets,but there is still no consensus on the influencing factors and mechanism of correlation.At the same time,most of the research just stays in the time-series correlation analysis of stock markets.As for spatial correlation of stock markets,there is less research.In a small number of studies based on spatial correlation of stock markets,the research concludes that correlations of stock markets are closely economically dependent.However,in reality,some markets that are not closely economically dependent also have linkage to other stock markets,and the linkage shows different degrees of agglomeration and regionalization.This phenomena involving geographical space has gone beyond the traditional econometrics research,and they can no longer explain the reasons.Therefore,it is necessary to introduce spatial econometric models to study the linkage of stock market.This dissertation will systematically and comprehensively analyzes the correlations of international stock markets from the perspective of spatial correlation.Specifically,the analysis involves influence factors and interaction mechanism of correlation,the impact of correlation on portfolio returns or risk as well as some portfolio risk forecast models constructed from the view of investors.Logically,this paper is divided into two parts,one is to study mechanism of correlation of stock markets with constant space weights and spatial effect with variable weights,the other one is to construct risk prediction models.Specifically,the main content and the innovation of this paper are summarized as follows:(1)This paper is to study the degrees of correlations and influence factors,as well as mechanism of correlation of stock markets using the spatial econometric model.The selected samples invole 43-country stock markets,including developed markets and emerging ones from December 2006 to June 2015.Our research is to to discover and make some strategies for cross-border stock indices investment.(2)In order to overcome the shortcomings that the existent spatial weight can not accurately describe the interaction of different stock markets,several variable weight functions and their corresponding weights are constructed under the guidance of variable weight theory.First,we construct a kind of nested weight as constant weight,and then put forward three types of variable weight functions.Among them,these functions include completely punitive function,completely motivated one and incentives dominate the hybrid one,which are all based on the already exist one that punishment plays a leading role.Then,four types of functions are employed to construct variable weights.Finally,the spatial panel regression model established by 43 stock markets is adopted to test the rationality and effectiveness of variable weights.(3)Constructe a Space-GJR-GARCH model based on spatial variable weights to forecast investment risk.In order to provide a more accurate risk prediction investing in cross-border stock indices,this paper firstly proposes a risk forecasting model(hereinafter referred to as Space-GJR-GARCH model)for transnational stock indices portfolio with space-variant weighting.The model takes time sequence-related and spatial correlation of portfolio returns into account,it also involves the asymmetry of volatility of portfolio returns.Then,when using the maximum likelihood estimation to estimate the parameters of the model,we set the initial value of the parameter to be estimated into two parts: one is the estimation result of a similar model(such as Space-GARCH model),and the other one is seted with some unknown parameters.We set the initial value of the parameters in this way is to make the convergence of the estimation iteration process faster and the result more accurate.Finally,an example is given to illustrate the superiority of the proposed model over the other several reference models in risk prediction.(4)Constructe a Space-GJR-GARCH-EVT-Copula model with spatial variable weight and tail correlation for multinational portfolio risk forecast.This model solves the tail-related problem that volatility of portfolio returns of stock indices is not considered in the Space-GJR-GARCH model.Meanwhile,in the process of modeling,the variables of selected Copula function are expanded from two dimensions in the existing literature to six dimensions.The types of Copulas are also expanded from the elliptic Copulas(including Gaussian Copula and t Copula functions)to elliptic Copulas plus Archimedes Copulas.That is,adding the Clayton Copulas and Gumbel ones.Thus,the proposed Space-GJR-GARCH-EVT-Copula model can be applied in a broader and a universal scope for investment in multinational equity indices.Empirical evidence shows that the new proposed model has more advantages than the other reference models in predicting the minimum portfolio risk,whether with the equal weight or non-equal weight. |
PDF Full Text Request