Three Essays on Dynamic Asset Allocation Models with Downside Risk Control

近幾年，風險管理受到大家廣為重視，Value-at-Risk (VaR)則是最常用來衡量風險的工具。Basak and Shapiro (2001)是首位將涉險值（VaR）的限制式納入效用函數內，再極大化投資人之效用函數而求出最適資產配置。依據他們的方法，本文的第一部分（見第二章）探討當資產報酬分配呈左偏和肥尾時，對風險管理者資產配置之影響。許多實證研究顯示資產報酬分配呈左偏和肥尾。本文採用Gram-Charlier expansion近似資產報酬分配，探討當資產報酬分配在非常態分配下，其資產配置的變化。對風險管理者而言，最重要的工作就是準確預測損失與發生損失的機率。瞭解資產報酬的型態將有助於準確的預測損失，我們無法降低損失，但可以降低發生損失的機率，本文建議可以降低 值（期末財富損失大於VaR之機率）來達成，而降低 值會使期末財富在好的狀態與壞的狀態的財富稍減。利率是影響使用金融工具的主要因素，本文的第二部分（見第三章）探討VaR風險管理者當考慮利率風險時如何配置其資產，本文採用Vasicek-type模型描述隨機利率，探討在隨機利率的情況下，財富配置於現金、股票與債券之比例。本文將這些參數以數值代入，分析VaR風險管理者期末財富的分配情況以及期中現金、股票與債券之配置情形。本文的第三部分（見第四章）探討VaR風險管理者當考慮利率與通膨風險時如何配置其資產。本文採用correlated Ornstein-Uhlenbeck過程描述隨機實質利率與通膨率，探討當考慮利率與通膨風險的情況下，VaR風險管理者財富配置於現金、股票與債券之比例。對風險管理者而言，最重要的工作就是準確預測期末財富與損失。研究發現忽略通膨風險將使風險管理者嚴重低估期末財富與損失。Risk management has received much attention in the last few years. Value-at-Risk (VaR) is widely used by corporate treasurers, fund managers and financial institution (Hull, 2000). A vast amount of literature considered a simple one-period asset allocation problem under VaR constraint. Furthermore, the aggregation of single-period optimal decisions across periods might not be optimal for multi-period as a whole. Basak and Shapiro (2001) were the first to address VaR-related issue in a dynamic general equilibrium setting. This dissertation builds upon the work of Basak and Shapiro (2001) to discuss three issues about dynamic asset allocation. The first topic focuses on how deviations from normality affect asset choices made by risk managers. This study utilizes the Gram-Charlier expansion to approximate asset returns with negatively skewed and excess kurtosis. This work examines how negatively skewed and excess kurtosis affects asset allocations when investors manage market-risk exposure using Value-at-Risk-based risk management (VaR-RM). It is important for risk managers to precisely forecast the loss. The analytical results imply that the impact of leptokurtic asset returns is based on the shape of asset returns, and a correct measurement of leptokurtic asset returns is helpful to risk managers seeking to precisely forecast the loss. A risk manager cannot reduce the loss in bad states, but can reduce the value of , the probability that a loss exceeds VaR, and the agent will suffer from reduced terminal wealth in both the good and bad states. The second topic solves an optimal investment problem involving a VaR risk manager who must allocate his wealth among cash, stocks and bonds. This study incorporates a stochastic interest rate process into the optimization problem. A Vasicek（1977）one-factor model governed the dynamics of the term structure of interest rates and risk premia are constant. Closed form formulate for the optimal investment strategy are obtained by assuming complete financial markets. Moreover, this study provides numerical examples to analyze the optimal terminal wealth and portfolio weights in stocks and bonds of the VaR risk manager. This work demonstrated the bond-stock allocation puzzle of Canner et al. (1997) that the bond-to-stock weighting ratio increases with risk aversion in popular investment advice in contradiction with standard two fund separation. Finally, this work derives the optimal portfolio selection of the VaR manager by assuming complete financial markets and that the inflation and real interest rates follow correlated Ornstein-Uhlenbeck processes. This study provides numerical examples to analyze the optimal terminal real wealth and optimal portfolio in stocks and two nominal bonds with different maturities. Furthermore, this work studies the influence of the parameters of inflation on the solution. This work illustrated that the younger VaR agent who has a long investment horizon invests the fraction of wealth in stock varies with the state price. It is not consistent with the Samuelson puzzle