貝氏門檻模型在財務議題之探討

Abstract

這項研究包含兩篇貝氏門檻模型在金融市場議題探討之論文。 在第一篇論文中，我們提出四因子的貝氏門檻模型去比較基金經理人面對市場走空或走多時，對於系統風險調整是否存在非對稱性。我們證明不僅經理人有非對稱的擇時能力而且三區間的模型比二區間模型有更顯著的擇時能力。另外，我們使用縱橫資料模型檢查基金投資者的行為與基金績效和特性之間的關係。實證結果說明投資者的行為與過去基金選股績效和基金規模正相關，而與過去基金周轉率，銷售費，費用率負相關。另外，有較大即時現金流量的基金將有較好的預測上升趨勢市場的能力及較差的預測下降趨勢市場的能力。 第二篇論文提出穩健多變量門檻VAR-GARCH-DCC模型，這個模型可以描述在金融資產其條件平均值，波動，與相關存在的非對稱性。另外，我們把門檻變數假設成所有內生變數的線性組合。因為這樣不僅可以消除過分主觀的選取門檻變數，而且還可以作為決定哪一市場是價格領先者。我們用MCMC方法去估計模型中的參數。而且，介紹幾個有意義的準則去評估條件共變異矩陣的預測績效。最後，我們使用每日的S&P500 期貨和現貨價格，和S&P500與Nasdaq100現貨價格作為實證研究。

This study contains two essays on the Bayesian threshold model in financial markets. In essay 1, we propose a Bayesian three-regime threshold four-factor model to compare the asymmetric risk adjustment between the transitions from neutral to downside markets and those from neutral to upside markets and investigate the performance of mutual funds in changing market conditions. We show that not only fund managers have asymmetric timing ability but three-regime models are more powerful and exhibit significant timing ability more often than two-regime models. In addition, we use panel data model to examine fund investors’ behavior and the relationships between fund performances and characteristics. Empirical results suggest that investor’s behavior is positively associated with past selectivity performances and fund sizes, while it is negatively correlated to past turnover, load charges and expenses. In addition, funds with large contemporaneous net cash flows will results in better upside market timing ability but worse downside market timing ability. Essay 2 proposes a robust multivariate threshold vector autoregressive (VAR) model with generalized autoregressive conditional heteroskedasticities (GARCH) and dynamic conditional correlations (DCC) to describe conditional mean, volatility and correlation asymmetries in financial markets. In addition, the threshold variable for regime switching is formulated as a weighted average of endogenous variables to eliminate excessively subjective belief in the threshold variable decision and to serve as the proxy in deciding which market to be the price leader. Estimation is performed using Markov chain Monte Carlo (MCMC) methods. Furthermore, several meaningful criteria are introduced to assess the forecasting performance in conditional covariance matrix. The proposed methodology is illustrated using two data sets including daily S&P500 futures and spot prices, and S&P500 and Nasdaq100 spot prices.