Risk Analytics
Value at risk
What does a ₹1 lakh one-day VaR actually promise?
Value at risk estimates a loss threshold at a specified horizon and confidence level.
VaR at level α = α-quantile of modeled losses
At a 95% confidence level, order the modeled losses and find their 95th percentile. The ₹100,000 threshold describes a loss quantile. It leaves the size of more extreme losses unspecified.
A model with 95% VaR of ₹100,000 gives a threshold, not a bound on possible loss.
At a 95% confidence level, order the modeled losses and find their 95th percentile. The ₹100,000 threshold describes a loss quantile. It leaves the size of more extreme losses unspecified.
Which number tells you about average losses in the modeled worst 5%?