News impacts on financial return distributions: long memory and regime switching approaches
2017-02-28T04:58:45Z (GMT) by
Critical roles of return higher moments in financial activities, which have been increasingly documented, suggest that it is worthwhile to analyze the behavior of the financial return distributions under various market conditions. The aim of this thesis is to model the responses of stock and currency return distributions to exogenous shocks under various forms of news which hit the financial markets. Specifically, this thesis is concerned with three scenarios: (i) when each of the return higher moments is shocked; (ii) when the hidden information arrives; and (iii) when the overall sovereign credit ratings change. Chapter 2 examines the linkages within-between stock and currency (FX) markets via three higher moments: realized volatility, skewness and kurtosis using the Generalized Impulse Response within a Fractionally Integrated Vector Autoregressive (FIVAR) framework. We find evidences of positive linkages within stock and FX markets via all three higher moments in both emerging and developed groups. However, the spread of the FX markets linkages via their 2nd and 4th moment is broader in the developed regions compared with the emerging regions. For the cross-assets linkages, the stock and FX markets in emerging groups are more likely to be negatively linked through the 3rd moment; whereas, those in developed groups are positively transmitted through the 2nd and 4th moment. Finally, in developed markets, the cross-assets linkages are often found to be weaker than the same asset linkages in terms of the magnitude. Limitations of methodology used in Chapter 2, where the endogenous variables in a FIVAR model need to be fractionally differenced before using the impulse response analysis of a VAR model, lead us to develop a new approach in Chapter 3. We based on the spirit of Peseran and Shin (1998) to derive a generalized impulse response function for the FIVAR model. Chung (2001) has the same purpose but he makes use of the orthogonalized approach proposed by Sims (1980). Our method is different from the methodology shown in Chung (2001) in a sense that it does not require us to orthogonalize the error vector and, therefore, is independent of the ordering of the variables in the system. Consistent with Chung (2001) and the long memory behavior, we show that generalized and orthogonalized impulse responses of FIVAR evolve slowly at the same hyperbolic rates. However, we also note that they are different in a number of aspects. For the purpose of statistical inference in empirical studies, we derive asymptotic theories for both functions. We summarize the results for two scenarios associated with one- and two-step estimation methods, respectively. However, our simulations’ results support an application of the two-step estimation procedure in generating the generalized and orthogonalized impulse responses of a FIVAR model. Chapter 4 utilizes the methodology developed in Chapter 3 to reassess influences of trading volume on stock and FX return distributions while allowing the possibility of interactions among return higher moments. Given the evidence of the higher moments’ inter-relationship, the chapter extends the analysis by exploring how trading volume affects the dynamic structure of higher moments’ inter-relationship. Our reassessment of volume – volatility interaction supports a complementary property among information theories and further contributes evidence of cross – market relations between volume and volatility. The result for the volume – skewness relationship in conjunction with previous studies leads to a hypothesis that direct impact of volume on the level of negative skewness is less significant for a better diversified portfolio. We further find that the negative interaction between volume and kurtosis can be explained by the differences of opinion hypothesis. Although behavior of the inter-relationship towards significant events and new policies are robust, its strength is mostly decreased by the trading volume. Fundamentally, this finding is consistent with the prominent result found in the volume – GARCH effect literature, which suggests that trading volume is a source of heteroskedasticity in the return volatility. In Chapter 5, we investigate the effects of credit rating agencies (CRAs)’ sovereign credit assessments on stock and currency return distributions by developing a framework that allows a multivariate system of long memory processes to be conditional on specific credit rating regimes. We find heterogeneous effects of sovereign rating actions across regimes, implying the usefulness of our proposed model in accommodating both long memory and regime switching features. Furthermore, we reveal that the total effects (both direct and indirect forces) of sovereign credit assessments on the realized moments can be different to their direct effects. Hence, we develop an impulse response of a transfer function, which can capture these total effects, to investigate which agency has the greatest impact on the EU financial return distributions. We find that the rank orders of CRAs are not unique across rating regimes and even in each realized moment.