Market Risk Measurement – Complete Study Guide
Prepared for Professional Risk Manager (PRM) Examination – Market Risk Measurement (PRMIA
Handbook)
3.1 Value-at-Risk (VaR) – Concept and Calculation
Value-at-Risk (VaR) quantifies the maximum potential loss of a portfolio over a specified time
horizon at a certain confidence level. It is a core risk measure used by banks to estimate market
risk capital and control limits.
Formula:
P(VT − V0 ≤ −VaR) = 1 − α
VaR = zασ■V■
where:
• zα = quantile value (e.g., 2.33 for 99%)
• σ■ = portfolio standard deviation
• V■ = current portfolio value
Example Calculation:
A portfolio valued at $100 million has a daily volatility (σ■) of 1.5%. Compute the 99% one-day
VaR.
VaR = 2.33 × 1.5% × 100,000,000 = $3,495,000
Interpretation: There is a 1% chance the portfolio will lose more than $3.5 million in a day.
Marginal, Incremental, and Component VaR
Type Formula Explanation
Marginal VaR MVaR■ = ∂VaR■/∂V■ = z(σ■■/σ■) Change in portfolio VaR from a small change in asset i
Incremental VaR IVaR■ = VaR■■■ − VaR■ Change in VaR from adding a new position a
Component VaR CVaR■ = MVaR■ × V■ Contribution of asset i to total VaR
Example:
For a $100M portfolio with σ■=1.5%, a sub-asset has σ■=2.0% and correlation ρ■■=0.8.
MVaR■ = 2.33 × 2.0% × 0.8 = 3.73%
CVaR■ = 3.73% × $10M = $373,000
Volatility Estimation Models
1. EWMA (Exponential Weighted Moving Average)
σ²■ = λσ²■■■ + (1 − λ)r²■■■
λ = decay factor (usually 0.94). The model gives higher weight to recent market volatility.
Example: σ■■■=1%, r■■■=2%, λ=0.94 → σ²■=0.94(0.01²)+0.06(0.02²)=0.000196 → σ■=1.4%.
2. GARCH(1,1) Model
σ²■ = ω + αr²■■■ + βσ²■■■
Captures volatility clustering; α and β represent reaction and persistence.
Parametric VaR Calculation Example
Assume portfolio standard deviation σ■=1.83%, z=1.96 (95% CL):
VaR■d = 1.96 × 1.83% × $100M = $3.6M
VaR■■d = $3.6M × √10 = $11.4M
Historical Simulation VaR
1. Use past returns (e.g., 500 days)
2. Compute daily P&L;
3. Rank from worst to best
4. Take 1% quantile as VaR
Example: 500-day P&L; → 5th worst = −$1.8M → VaR■■ = $1.8M.
Monte Carlo Simulation VaR
Simulate correlated risk factor shocks using Cholesky decomposition. For each simulation k:
f■■■ = f■ exp[(−½σ²■)∆t + σ■ε■√∆t]
The 97.5% quantile of 10,000 simulated losses = VaR■■.■.
Principal Component Analysis (PCA)
Used to reduce correlated yield curve risk factors to independent components: Level, Slope,
Curvature.
σ²■ ≈ δ²_PC■λ_PC■ + δ²_PC■λ_PC■ + δ²_PC■λ_PC■
Example: 3 PCs explain 98% variance of yield curve; VaR derived using only these 3.
Extreme Value Theory (EVT)
Models tail losses beyond VaR threshold.
GEV: G(z)=exp[−(1+ξz)^(−1/ξ)], valid if 1+ξz>0
POT VaR Formula: VaRα = u + β/ξ[(1−α)^(−ξ) − 1]
Example: ξ=0.2, β=1000, u=5000, α=0.99 → VaR■■ = 5000 + (1000/0.2)[(0.01)^(-0.2) − 1] ≈
$11,000.
Expected Shortfall (ES)
ES = E[−X | X ≤ −VaR]. It measures the average loss in the tail beyond VaR.
Example: If 99% VaR=$3.5M and the average of the worst 1% outcomes = $4.8M → ES■■ =
$4.8M.
Backtesting VaR Models
Regulators require testing VaR accuracy by comparing predicted vs. actual P&L.;
Method Formula / Description
Kupiec Test Λ = (p■/p)^x × ((1−p■)/(1−p))^(n−x); tests correct coverage.
Traffic Light Approach Green (≤4 exceptions), Yellow (5–9), Red (≥10).
Kolmogorov–Smirnov KS = √n max|F■(x) − F(x)|; tests distribution match.
Example: 250 trading days, 3 exceptions → within expected (Green zone, K=3.0).
Basel 2.5 / FRTB Enhancements
• Incremental Risk Charge (IRC): 99.9% CL, 1-year horizon (Vasicek model)
• Stressed VaR (SVaR): calibrated to crisis period
• Expected Shortfall replaces VaR in Basel IV (97.5% CL)
Final Summary
VaR quantifies loss risk, Expected Shortfall improves tail accuracy, PCA reduces dimensionality,
and EVT and copulas enhance modeling of extreme and dependent risks. Backtesting ensures
model reliability per Basel standards.