typora-copy-images-to: Risk Management and Financial Institution

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach


  • 模型构建法也是风险指标衡量的方法,又被称作为==variance–covariance approach==

  • 模型构建法理想的是用于衡量底层市场变量与组合价值线性相关的情境下

  • 模型构建法是马科维茨先生组合理论的拓展

  • 非线性相关的情况下,可以使用蒙特卡洛模拟进行估计,同样花时间

  • 本章介绍两个重要的模型:

    • Standard Initial Margin Model(SIMM),用于衍生品决定初始保证金
    • SA approach 决定Trading Book 的capital

14.1 The Basic Methodology

  • 一般假设expected change 是 zero,因为相对于change 的标准差来说,预期变化太小了
  • 回忆ES的计算公式

14.1.1 两种资产的情况

  • 考虑连个资产回报之间的相关性ρ
  • 两个资产的标准差计算方法:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 其他的计算方式同上,假设是两种资产回报服从二元正态分布

14.1.2 The benifits of diversification 分散化的好处

  • 如果两项资产的相关关系是完全正相关,那么VaR等于各自VaR的总和
  • 如果相关关系不是完全正相关,则存在分散化的好处

14.2 Generalization 普遍推广

  • Suppose that we have a portfolio worth P that is dependent on n market variables

  • 暂时只考虑股价、商品价格、外汇汇率等风险因子,那么组合价值与风险因子之间存在线性相关:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 假设风险因子xi的变动是多元正态分布,而delta P服从正态分布,则delta P 的标准差是:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 上述式子使用矩阵标注:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • the VaR for a T-day time horizon and an X% confidence level is :Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • ES =Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 例题:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.3 四资产组合的再回顾(详细见EXCEL——Model Building)

  • Equal weight
  • EWMA is using

14.4 关于利率的处理 Handling Term Structure

  • interest rates , credit spreads, volatilities

  • The term structure shows the value of the variable as a function of time to maturity

    • the term structure shows the zero-coupon interest rate as a function of its maturity
    • credit spread, the term structure shows the credit spread applicable to a bond or credit default swap as a function of its maturity
    • volatilities, the term structure shows the volatility used to price an at-the-money option as a function of its maturity
  • 两种常用的处理方式解决利率期限结构风险:

    • PCA 主成分分析
    • multiple-vertices approach
    • 以上两种方式更适用真实变化而不是比例变化

14.4.1 PCA

  • The first two or three principal components account for most of the variation in a term structure that is observed in practice

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • The first PCA exposure:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • The second PCA exposure:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 因此,delta P = Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • σ delta P = Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.4.2 The Multiple Vertices Approach

  • A delta is calculated for each point on the term structure by shifting that point by one basis point while keeping the others unchanged. We will refer to deltas defined in this way as “node deltas.”
  • Node deltas are therefore a way of dividing the DV01 into 10 component parts

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 例子:一个组合,对连个不同的利率期限结构暴露,利率波动的每日标准差以及delta如下图:两年期利率不同期限结构的标准差为5.6(0.056%),以及11.4(0.114%),利率期限结构每波动一个基点,影响组合价值分别为85 与 65 million,期限结构1与期限结构2的不同期限相关关系图都一致,两个期限结构的相关关系系数为0.4

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.5 基本步骤的延伸拓展

  • 对于风险因子线性相关的拓展探索

14.5.1 压力方法 Stressed Measure

  • 使用压力时期的波动率数据以及相关关系来衡量VaR

14.5.2 Non - normal distrutions 非正态分布

  • n 个资产形成的组合的VaR的聚合

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • The result suggests a way the model-building approach can be extended to allow for return distributions being non-normal

14.6 风险权重以及权重敏感性

  • When equations such as (14.2) and (14.4) are used to calculate VaR or ES, the standard deviation of the daily change in the portfolio is multiplied by a constant

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • W-parameters are referred to as risk weights, and the δ-parameters are referred to as weighted sensitivities

  • 这个公式用来决定市场风险的资本留存

  • 用来决定初始保证金的数量

14.7 非线性资产组合的模型构建法处理

  • 指包含期权的组合,存在vega 以及gamma
  • vega 一般可以进行线性的估计,因此可以使用波动率期限结构估算
  • gamma涉及多项式,因此更难估计,使用泰勒展开式:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 最后一项在计算VaR 以及ES 时不可省略
  • 例如一个long call opthon,long call 有positive 的gamma,如果底层资产价格服从正态分布,那么option的价格相对而言是有偏的,当gamma为正时,组合价值概率分布为正偏分布,当gamma为负时,组合价值概率分布为负偏分布,如下图:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • a positive gamma portfolio tends to have a less heavy left tail than the normal distribution则会导致VaR值高估,如果假设正态分布的话
  • ==a negative gamma portfolio tends to have a heavier left tail than the normal distribution.==If the distribution is assumed to be normal, the calculated VaR will tend to be too low导致VaR的低估

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.7.1 蒙特卡洛模拟

  • 使用蒙特卡洛模拟来生成delta P 的概率分布

  • 步骤如下:

    • 利用当前市场变量对交易组合定价
    • 由delta xi服从的多元正态分布中进行一次抽样
    • 由delta xi的抽样计算出交易日末的市场变量
    • 利用新产出的市场变量对交易组合进行重新定价
    • 将第四步产生的组合价值,减去第一步的组合价值,得到一个delta P 的抽样
    • 重复第二步到第五步,得到概率分布
  • 蒙特卡洛模拟的缺陷是:

    • computationally slow,一种加速计算的方式是利用delta P 与delta x的关系,可以跳过第2步到第5步,这一技巧叫做局部模拟法

14.7.2 Extension 拓展

  • 蒙特卡洛的一个优点,我们无需假设风险因子是服从正态分布,使用多元copula来定义变量之间的相关关系
  • 在第二步和第三步之间增加如下模拟过程:

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.7.3 Cornish -Fisher expansions

  • 一个概率分布的有偏性取决于它的第三阶距,正偏代表右侧厚尾,负偏代表左侧厚尾,

  • 偏度的估计公式:Risk Management and Financial Institution Chapter 14 —— Model-Building Approach,正态分布的有偏性为0

  • 峰度的估计取决于它的第四阶距,估计公式为:Risk Management and Financial Institution Chapter 14 —— Model-Building Approach正态分布的峰度为3

  • 因此,q分位数的delta P分布为:Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • 其中,Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

  • zq is q-percentile of the standard normal distribution

Risk Management and Financial Institution Chapter 14 —— Model-Building Approach

14.7.4 Isserlis’Theorem

  • 用来计算高阶距(略)

14.8 模型构建法与历史模拟法

  • 模型构建法的优点在于:

    • 结果可以很快得出
    • 相关关系与波动率更新的情况下易于应用
  • 模型构建法的缺点在于:

    • 当市场变量与组合价值线性相关时,以及风险因子的多元正态分布的假设
  • 模型构建法经常用于投资组合,而不是用来计算VaR,因为gamma很难考虑

相关文章: