Module XV·Article I·~3 min read

VaR Concept

Portfolio Risk Management

Turn this article into a podcast

Pick voices, format, length — AI generates the audio

Value at Risk: Measuring Potential Losses

Value at Risk (VaR) is a statistical measure that answers the question: "How much can we lose under normal market conditions?" VaR has become the risk management standard in banks and investment funds.

Three Components of VaR

ComponentDescriptionTypical Values
Loss AmountMaximum expected loss$1M, 5% of portfolio
HorizonTime period1 day, 10 days, 1 month
Confidence LevelProbability not to exceed95%, 99%, 99.5%

Interpretation of VaR

Example: VaR (1-day, 95%) = $1M

  • "On 95% of trading days, the loss will not exceed $1M"
  • "On 5% of days (~1 time per month), the loss will be more than $1M"
  • "VaR DOES NOT say how much more than $1M the loss will be on bad days"

Methods for Calculating VaR

MethodApproachProsCons
Parametric (Variance-Covariance)Assumption of normal distributionFast, simpleUnderestimates tail risk
Historical SimulationUsing historical dataRequires no assumptionsPast ≠ future
Monte CarloSimulation of thousands of scenariosFlexible, accounts for nonlinearityComputationally intensive

Parametric VaR: Formula

$ \text{VaR} = \text{Portfolio Value} \times \sigma \times z \times \sqrt{T} $

VariableDescription
$\sigma$Portfolio volatility (annualized)
$z$Z-score for confidence level (1.65 for 95%, 2.33 for 99%)
$T$Horizon in years (1 day = 1/252)

Example Calculation

Portfolio = $100M, $\sigma$ = 15%, 1-day 95% VaR:

$ \text{VaR} = $100\text{M} \times 0.15 \times 1.65 \times \sqrt{\frac{1}{252}} = $1.56\text{M} $

Scaling VaR Over Time

$ \text{VaR}(T~\text{days}) = \text{VaR}(1~\text{day}) \times \sqrt{T} $

HorizonMultiplierVaR (if 1-day = $1M)
1 day1.00$1.00M
5 days2.24$2.24M
10 days3.16$3.16M
21 days (1 mo)4.58$4.58M
252 days (1 yr)15.87$15.87M

VaR Limitations

LimitationProblemSolution
Does not show tailVaR = boundary, not average tail lossExpected Shortfall
Normality assumptionFat tails in realityHistorical, Monte Carlo
Backward-lookingUses past dataStress testing
Not subadditivePortfolio VaR can be > sum of partsES is subadditive
Model riskDifferent methods = different resultsMultiple approaches

VaR Limits in Risk Management

LevelVaR LimitAction Upon Breach
GreenNormal operations
Amber80–100% of limitEnhanced monitoring
Red>100% of limitMandatory risk reduction

CIO Recommendations

  • VaR as one of the tools — not the only metric
  • Supplement with ES and stress tests — VaR cannot see tails
  • Calibrate to reality — check via backtesting
  • Multiple horizons — 1-day, 10-day, monthly
  • Report consistently — standardized reporting

§ Act · what next