Lecture 2

Measuring Market Risk

Introduction to VaR
Dr HK Pradhan
XLRI Jamshedpur

John C Hull Chapter 12

 Session Goals
• Measuring risk with Var
• Portfolio VaR computations
• Marginal, Relative & Incremental Var
• Expected shortfalls
• Using Var in risk reporting risk
 Origin of VaR
– J P Morgan’s famous 4.15 Report by then CEO
Dannis Weatherstone, somewhere in the late 1980s,
following the 1987 global market crash
– DEaR (Daily Earnings at Risk, basically worst case at 95%
level): a 1-page snapshot on positions and risk levels in
every major markets across all products (comprising more
than 1.5 million positions and 240,000 pricing series
involving securities prices, interest rates, foreign exchange
rates, etc)
– VaR = μ - z σ
– RiskMetrics Technical Document (1994)
– The 4.15 Report Became the First Value At Risk (VaR)
Report
 Measuring Risk
• Risk relates to volatility
– “How much can we lose tomorrow, with 99% confidence,
under normal market conditions?
• Notional exposures, sensitivity analysis, scenario
analysis, stop-loss limits have limitations in
quantifying risk in a futuristic/probabilistic sense
• Standard measures such as variance, Sharpe ratios,
Sortino ratios have limitations
• We need to measure expected tail losses
• Risk has to be measured in a probabilistic sense, at
portfolio levels, aggregated over assets, and capital
requirements assessed
 How much you lose?

Risk = What you can possibly lose if the market moves against you

= Amount of Position * Volatility of Price

Volatility
is the key

Example: you have a long bond position

Amount: Rs 1,000,000
Volatility per day: 0.95% (expected to drop 0.55% on average)

Risk = Rs 1 million * 0.0095 = Rs 9, 500

Need to estimate the potential loss under probabilistic scenario

VaR as Risk Measurement Tool
– VaR was pioneered as risk factors exhibited
considerable volatility
• When banks & FIs moved from accrual based to
marked to market valuation of assets
– Need for risk aggregation
• To measure risk across products, risk takers and
functions, regions
– Understand diversification benefits
– Need to relate risk to capacity of the institution
– Need to facilitate risk communication
• trading desk, corporate office and board,
regulators, shareholders, ratings agencies
 Measuring Risk (VaR)
Standard rewritten VaR equation 60

(return form): 50
1%
40
VaRp = P × ( μ − Z × σ × √t )
30

μ = Expected (mean) return

20

Z = Z-score corresponding to 10

confidence level (e.g., 2.33 for 99%) 0

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96

σ = Portfolio standard deviation

√t = Time scaling factor
Z obtained from the NORMSINV i.e NORMSINV excel function

Example: Assuming 99% confidence and a 1-day horizon, with a volatility of x% per
day, the maximum loss expected will be Rs…. And losses exceeding this will be only
1% of the time.
Var calculated from the whole distribution of portfolio daily loss(gain being counted
as negative loss)
The edge of acceptable
loss: 2.33(99%
confidence level)
standard deviations
below normal
 Assumption of Normal Distribution of Reruns

Prob(r   - 2.33 ) =1%

99%
Sd*Mean confidence level Lower
Tail Probability
1 84% 16%
1.65 95% 5%
2.33 99% 1%
• Normal distribution allows us to estimate the deviation from the mean by standard
deviation(68.27% of the observations within +/- 1-Sd; 95.45% of the observations
within +/- 2-Sd; 99.73% of the observations within +/- 3-Sd from the mean)
• Conversion of one confidence level to another easier:
• 95% confidence level VaR when translate it to the BIS standard of 99%
confidence level: 95% VaR  2.33/1.65 = 99% VaR
 Value at Risk(VaR)
How much you lose? RM 1996 Page 18

Value at Risk (VaR) - estimates the potential loss due to market

movements within a prescribed confidence level over a prescribed
time interval.
Example
• Assuming 99% confidence and a 1-day
horizon, a VaR of X million means that,
on average, the maximum loss will be X
millions with 99% confidence
1% • Losses exceeding the VaR amount X
million should occur 1% of the time
99% area
Important
Loss Profits •Potential downside exposure
Return Distribution •Given confidence level
•Over defined time horizon
•Due to market movements
 What does Var tells us?

VaR = adverse loss on

a normally bad day

You are capturing 1% of the worst cases, with 99% confidence level
• Maximum potential loss that a portfolio can suffer in the 1% worst
cases in n-days??
• Minimum potential loss that a portfolio can suffer in the 1% worst
cases in n-days??
• Maximum potential loss that a portfolio can suffer in the 99% best
cases in n-days??
 Let’s Compute VaR
• Let us define daily return on an asset as follows:
• The change in the logarithm of the daily closing price of
the asset
 St 
rt = Ln ( S t ) − Ln ( S t −1 ) = Ln  

S
 t −1 
• The arithmetic return is instead defined as

rt +1 = (St +1 − St ) St or rt +1 = St +1 St − 1

• Log return & percentage return are close on a daily basis

 Advantage of Log Return
• From the log return we can easily calculate the
compounded return over K-days as the sum of daily
return
K K
Rt +1:t + K = ln(St + K ) − ln(St ) =  ln(St + k ) − ln(St + k −1 ) =  Rt + k
k =1 k =1
K
ln(S t + K ) / ln(S t ) =  Rt + k
k =1

• Called the geometric average of single period return

• Log return does not consider negative prices

 VaR measurement steps:
Setting Parameters:
Set the base currency, measure of risk, probability of loss and
confidence level and time horizon
Risk Measure:
What is the statistical measure we are using to quantify our
exposure to risk? What is the measure of volatility
Level of Confidence:
With what level of confidence one tries to measure and
quantify the level of risk given probability? What
assumptions of distribution?
Time Horizon:
Over what period of time are we concerned to consider our
exposure to risk? How volatility scales with time?
Holding period:
How many days it requires a longer period to unwind a
position, that is, if markets are less liquid.
:
 Confidence Intervals

• The value for X: .99, .95, .90…

– Regulatory requirements
– Backtesting history
– Environment of risk
– Liquidity and other market variables
– Capacity of the institution
 The Time Horizon
N - Day Var = 1 − Day VaR x t
• Time Horizon is the key:
– Over what period of time are we concerned to
consider our exposure to risk? How volatility
scales with time?
• It depends on the holding under consideration:
– How many days it requires a longer period to
unwind a position, that is, if markets are less
liquid.
 Measure VaR
• Var defined as the possible loss in value from “normal
market risk” as opposed to all other risk

Varp = P σp N ( x) t
−1

– Risk is measured by volatility of returns

– Given normality assumptions
– Given the confidence level
– Over a predetermined time horizon
Portfolio Risk
• Like portfolio diversification enhances
return, so also it reduces risk
• Importance correlation in asset portfolio
needs to be recognized
• Risk concentration in certain maturity
buckets, sectors, regions, or instruments
or special class of instruments, and
underlying exposure
– Example: Sub-prime crisis was also
due to large concentration of
exposure in sub-prime mortgages
 Portfolio VAR
For normal portfolio return distributions

Varp =MVp * z * σp * t
MVp = Market value of the portfolio
z = Lower quantile of standard normal distribution
 p = standard deviation of daily portfolio returns
t = time horizon/holding period

19
 Deriving Portfolio Volatility
With more than one asset, and the portfolio variance (volatility) is computed as:

 P2 (1, 2 ) = w12 12 + w22 22 + 2 w1w2 12

 12  12   w1 
= w1 w2   2  
 12  2   w2 

 P = ( wi i ) 2 + ( w j j ) 2 + 2wi w j  2 ij

Since Covariance  2 ij = ij  i j

 2 P =  wi w j  i j ij ij =1 for i = j

i j
 VaR Decomposition
• Diversified & undiversified Var
• Marginal, Incremental &Relative Var
• Conditional Var
• CFaR, CaR
 Diversified & undiversified Var
• All levels of risk management, and can capture
diversification effects, thereby providing the link
between aggregated risk and individual risk
• Sum of single position VAR ± portfolio effect
• When correlations assumed to be negligible, we add
component VaRs
 Marginal VaR
• Portfolios have sub-asset portfolios. Measures change in a sub-
portfolio leads to change in VaR is called as marginal Var
• Marginal VaR differs from VaR because of it offsets between the
position and rest of the portfolio
– Compare the stand-alone VaR with that of its contribution
within a portfolio
dVaR
• Marginal VaR dx i
• How much portfolio Var changes w.r.t every $ change in value of
position
• Marginal VaR can be negative meaning that the position
reduces the portfolio risk i.e., if it is a hedge. Say a forward
contract eliminates the position volatility from the portfolio
 Incremental VaR(IVaR)
• Incremental effect of a sub-portfolio
• What is the impact on Var “with or without the sub-
portfolio”
– Effect of a new investment decision that adds to total
business risk/sector risk
– Or by changing the position weights within the
portfolio and computing the resultant Var (due to the
increment effect of a new trade or the effect of closing
out an existing trade)
• The following approximation for incremental Var
dVaR
xi
dx i
 Relative VaR
Measures risk of underperformance relative
to a pre-defined benchmark, such as the Var Risk scenario
of a stock as relation to BSE Sensex or
Risk
NIFTY Index.
Example:
Relative VaR of Rs1million would mean
only 1 in 20 days you expect to under
perform your benchmark more than
Rs1million due to market movements given Time
95% level of confidence
• Var facilitates risk benchmarking: Portfolio managers would
have benchmarks based on market wide indicators, and
comparison can be made for the portfolio’s own risk
 Component VaR
• The component VaR is approximately similar to
that of incremental VaR
N
d (VaR)

i =1 dx i
xi

Where N is the number of components, so each is

d (VaR)
Ci = xi
dx i
Var = C1 + C4 + C3 + C2 + ... Euler’s theorem
The component VaR therefore provides a sensible way of
allocating VaR to different activities
 Coherent Risk Measures
• Properties of coherent risk measure
– If one portfolio always produces a worse outcome than
another its risk measure should be greater (Monotonicity)
– If we add an amount of cash K to a portfolio its risk
measure should go down by K (Translation Invariance)
– Changing the size of a portfolio by  should result in the
risk measure being multiplied by  (Homogeneity)
– The risk measures for two portfolios after they have been
merged should be no greater than the sum of their risk
measures before they were merged (Sub-additivity)

Refer an article by John Hull:

[Link]
paper/1506669/var-versus-expected-shortfall 8.27
 Other @Risk Measures
• Cash Flow at Risk(CFaR)
– Worst shortfall in cash flows due to
unfavorable movements in market risk factors
(say, the quantity & earnings impact of
exchange rate changes on exports)
• Cost at Risk (CaR)
– Expected maximum cost given a specified
probability and time period due to interest and
exchange rates changes
 Would Var Limit Risk Taking?
• You can have a situation where the position
risk is maintained well below the Var limit
• But it does not imply the probability of
worst case loss is being minimized?
• The right measure would be the this worst
case loss(which is expected shortfall)
 Expected shortfall
(Conditional Var-CVaR)
• Other measures of Var would
be necessary, gven market
conditions, distribution pattern
of the asset returns, etc
• One such measure considered VaR
is expected shortfall, i.e.
expected value of losses when
it exceeds the Var given
confidence level

VaR

When combined with VAR, ES gives a measure of the cost of insuring portfolio losses
 Expected Shortfall
• Expected Shortfall asks 'if things do get
bad, what is our expected loss’??
• ES is more sensitive to the shape of the tail
of the loss distribution
• Given two distributions with the same VaR
but can have different Expected Shortfalls
 VaR as Risk Measure
• Allow risky positions to be directly compared and
aggregated in P&L framework
• Risk aggregation is attained
– Several dimensions of risk can be integrated into a
single figure
– All risks can be expressed with the same currency
units
• Benefits of diversification measured
• Simulation of stressed risk factors can be conducted
 VaR uses in banks
• Treasury Risk Management (Bond VaR, Funding Risk) Trading
Book Risk Management (Trading VaR, Position Risk)
• Limit Setting & Risk Control (Desk Limits, Exposure Limits)
• Market Risk Capital Management (Basel VaR, Capital Charge)
• Stress Testing & Model Validation (Stressed VaR, Backtesting)
• Regulatory Capital Management (Capital Adequacy, Regulatory
Compliance)
• Economic Capital Management (Unexpected Loss, Capital
Buffer)
• Portfolio Risk Aggregation (Enterprise VaR, Diversification
Risk)
 Trading Risk Management

Daily Trading Limits

↓
Value at Risk (Normal Loss Estimate)
↓
Stressed VaR (Crisis Conditions)
↓
Expected Shortfall (Tail Risk)
↓
Stress Testing (Extreme Scenarios)
 VaR Applications in Indian MIIs
• Equity Market Margining (Initial Margin, Volatility Margin, Exposure Margin)
• Clearing Corporation Risk Management (Portfolio VaR, Extreme Loss
Margin, Default Fund Contribution)
• Government Securities & Repo Market Risk (Bond VaR, Interest Rate VaR,
Repo Exposure Measurement)
• Commodity Derivatives Risk Management (Commodity Price VaR, Futures
Margining)
• Collateral & Haircut Framework (Collateral VaR, Haircuts, Margin Adequacy
Testing)
• Settlement Guarantee Fund Management (SGF Sizing, Default Risk Buffer)
• Stress Testing & Crisis Management (Stressed VaR, Scenario Analysis)
• Foreign Exchange Settlement Risk (FX VaR, Exposure Limits, Settlement Risk
Control)
• Securities Lending & Borrowing Risk (Collateral Exposure, Counterparty
Protection)
• Liquidity Risk Monitoring (Liquidity VaR, Liquidation Risk, Funding Stress
Estimation)
 Risk waterfalls Mechanism by MIIs
Trade Executed
↓
Initial Margin (VaR Based)
↓
Mark-to-Market Margin
↓
Extreme Loss Margin
↓
Collateral Haircuts
↓
Settlement Guarantee Fund
↓
Default Waterfall
 VaR Summary
• VaR can play several important roles within financial institutions:
– Quantify risks, not notional exposure
• How much a position is expected to lose?
– Allow the management to make effective decisions as to how much return to
take
• Returns from diverse risky business directly comparable on a risk adjusted
basis
– Which exposures offset each other
• Highlights offsetting positions or hedges and diversification
– Where would the losses be concentrated?
• Facilitates stress testing – measures hidden concentration by business
group, region and risk type
– Dynamic actions with model backtesting & improvement
• Tracking historical P&L scenarios
– Risk benchmarking
• Comparing risk of a given position with that of the market-wide
benchmark
– A measure of the economic or equity capital required to support a given level
of risk (Basel II)
• Capital against risk in relation to the capacity of the institution
 Parametric VaR: Drawbacks
• VaR describes Maximum potential loss that a portfolio can
suffer under given market conditions
• Two parameters are considered important:
• CL factor & holding period
• Abnormal market movements, correlation breakdown,
etc not captured
• VaR does not describe the maximum losses in the left tail
– Other measures of tail losses needs to be recognised
• VaR is subject to statistical error
– estimates of volatility remains the key
 Limitations of VaR
• Statistical orientation and reliability
– Assumption of distribution
• Does not consider event risk
– Stress testing for non-normal market conditions
• Need for high quality data
Thank You