4.2.1 Measures of Financial Risk

Market Risk Measurement and Management

1. Measures of Financial Risk

1.1 Coherent Risk Measures

Monotonicity: If (regardless of what happens) a portfolio always produces a worse result than another portfolio, it should have a higher risk measure.

Translation Invariance: If an amount of cash $K$ is added to a portfolio, its risk measure should decrease by $K$.

Homogeneity: Changing the size of a portfolio by multiplying the amounts of all the components by $\lambda$ results in the risk measure being multiplied by $\lambda$.

Subadditivity: For any two portfolios, A and B, the risk measure for the portfolio formed by merging A and B should be no greater than the sum of the risk measures for portfolios A and B.

1.2 Mean-Variance Framework

In mean-variance(or standard deviation) framework, mean and variance is used to model financial risk and return follows normal distribution.The mean return is used to measure the expected return, and the standard deviation is used to measure the risk.

The normal distribution is often assumed for the returns on financial variables. It is a convenient choice because all we need is the mean and standard deviation of the returns.

In practice, the distribution of financial variables tend to be skewed or have much fatter tails, then the normality assumption is inappropriate and the mean-variance framework can produce misleading estimates of risk. The use of the standard deviation as a risk measurement is not appropriate for non-normal distribution.

1.3 Value at Risk

1.2.1 Definition of $\text{VaR}$

$\text{VaR}$ is the loss that we do not expect to be exceeded over the time horizon at the specified confidence level.

‘10-day $99\%$ $\text{VaR}$ is $10$ million’ means that we are $99\%$ certain that the loss during the next 10 days will be less than USD $10$ million, or there is a probability of only $1\%$ that the loss over the next 10 days will be greater than USD $10$ million.

Assume that asset returns conform to a normal distribution.

$$\text{VaR}(X\%)=|E(R)-Z_{X\%}\times \sigma |$$

$$\text{VaR}(X\%{)}_{\text{dollar}}=|E(R)-Z_{X\%}\times \sigma |\times \text{Asset Value}$$

$$Z_{X\%}=\{\begin{array}{ll}1.28,& \text{when}X=10\\ 1.65,& \text{when}X=5\\ 2.33,& \text{when}X=1\end{array}$$

Cindy is a trader and she is using $\text{VaR}$ to measure the risk of her portfolio. She assumes the asset return distribution is normally distributed. The portfolio daily standard deviation is $2.5\%$ and has a current value of $ $7.8$ million. Please calculate the daily percentage $\text{VaR}$($5\%$) and daily dollar $\text{VaR}$($5\%$) of this portfolio.

$\text{VaR}(5\%)=2.5\%\times 1.645=4.11\%$ （ 默认均值是$0$）

$\text{VaR}(5\%{)}_{\text{dollar}}=0.0411\times 7.8\text{million}=320,775$

The VaR for an investment opportunity is a function of two parameters:

• The time horizon: $\text{VaR}$ increases when the holding period is longer.
• The confidence level: $\text{VaR}$ increases when the degree of confidence increases.

Assuming return is normally distributed and $E(R)=0$, $\text{VaR}(X\%)=|Z_{X\%}\times \sigma |$

Time horizon: Square root rule can be used.

$${\sigma }_{\text{J-periods}}={\sigma }_{\text{1-period}}\times \sqrt{J}\to \text{VaR}(X\%{)}_{\text{J-periods}}=\text{VaR}(X\%{)}_{\text{1-periods}}\times \sqrt{J}$$

The confidence level

$$\text{VaR}(\text{New}\%)=\text{VaR}(\text{Old}\%)\times (Z_{\text{New}\%}/Z_{\text{Old}\%})$$

A risk analyst has to adjust the $95\%$ daily $\text{VaR}$ into $99\%$ weekly VaR to meet the new risk management policy. The $95\%$ daily $\text{VaR}$ is $ $3,300,000$. Please compute the $99\%$ weekly $\text{VaR}$. Assuming daily mean return of zero and $5$ trading days in a week.

$\text{VaR}(1\%{)}_{\text{daily}}=\text{VaR}(5\%{)}_{\text{daily}}\times (2.33/1.65)=4,660,000$

$\text{VaR}(1\%{)}_{\text{weekly}}=\text{VaR}(1\%{)}_{\text{daily}}\times \sqrt{5}=10,420,777$

1.2.2 Limitations of $\text{VaR}$

The $\text{VaR}$ only tells us the most we can lose if a tail event does not occur. (e.g., it tells us the most we can lose $95\%$ of the time.) If a tail event does occur, we can expect to lose more than the $\text{VaR}$, but the $\text{VaR}$ itself gives us no indication of how much that might be.

$\text{VaR}$ is not subadditive. We can only make the $\text{VaR}$ subadditive by imposing the severe restriction that returns distribution is elliptically distributed.

Suppose there are two bonds, Bond X and Bond Y. Each has a probability default of $3\%$ and $ $1000$ face value, If default occurs, zero value will be recovered. These two bonds are identical and independent. Calculate the $95\%$ VaR of single bond and then sum then up. Compare the sum of single bond VaR with portfolio $VaR$.

Portfolio of Bond X and Bond Y

Number of Defaulted Bond	Loss	Probability	Cumulative Probability
$2$	$2000$	$C_2^2(3\%{)}^2(97\%{)}^0=0.09\%$	$0.09\%$
$1$	$1000$	$C_2^1(3\%{)}^1(97\%{)}^1=5.82\%$	$5.91\%$
$0$	$0$	$C_2^0(3\%{)}^0(97\%{)}^2=94.09\%$	$100\%$

Single Bond ( Bond X or Bond Y)

Loss	Probability	Cumulative Probability
$1000$	$3\%$	$3\%$
$0$	$97\%$	$100\%$

${\text{VaR}}_{5\%}(\text{Portfolio})=1000$

${\text{VaR}}_{5\%}(\text{Single Bond})=0$

${\text{VaR}}_{5\%}(\text{Portfolio})>{\text{VaR}}_{5\%}(\text{Bond X})+{\text{VaR}}_{5\%}(\text{Bond Y})$

1.4 Expected Shortfall

Expected Shortfall($ES$) is a risk measure that does take account of expected losses beyond the VaR level, which is also called conditional $\text{VaR}$ ($\text{C-VaR}$) or tail loss.

It is the expected value of the worst $(100-X)\%$ of losses.

$ES$ gives an average amount of unfavorable events over a specific period while $\text{VaR}$ tells nothing about the exact amount of loss. $ES$ is the coherent risk measure, and it meets the property of sub-additivity, while $\text{VaR}$ does not.

A trader accumulated $100$ daily returns for a $2$ million portfolio. After sorting the returns into ascending order, the lowest ten returns were provided

$-12.3\%$, $-8.7\%$, $-6.5\%$, $-6.34\%$, $-3.2\%$, $-2,87\%$, $-1.2\%$, $-0.59\%$, $-0,05\%$, $0.47\%$

Please calculate the daily Expected Shortfall at $5\%$ significance level.

Return	Probability	Cumulative Probability
$-12.30\%$	$1\%$	$1\%$
$-8.70\%$	$1\%$	$2\%$
$-6.50\%$	$1\%$	$3\%$
$-6.34\%$	$1\%$	$4\%$
$-3.20\%$	$1\%$	$5\%$
$-2.87\%$	$1\%$	$6\%$
$-1.20\%$	$1\%$	$7\%$
$-0.59\%$	$1\%$	$8\%$
$-0.05\%$	$1\%$	$9\%$
$0.47\%$	$1\%$	$10\%$

$E{S}_{\text{daily}}(5\%)=(12.3\%+8.7\%+6.5\%+6.34\%)/4=8.49\%$

1.5 Spectral Risk Measures

Spectral risk measures provide the weighted means of the quantiles of the loss distributions.

$ES$ is a special case. $ES$ assigns an equal amount of weight to all loss levels greater than the $\text{VaR}$.

$\text{VaR}$ is also a special case. All the weight is assigned to one specific percentile (takes no account of losses more than the $X$ percentile point).

If a risk measure is coherent, the weights should be a non-decreasing function of the percentile of a loss distribution.