wilmott.com

Hi,

For a cross currency swap,

I have generated a 250 days price series keeping swap parameters constant and using 250 days market data.

What methods can be applied to calculate VaR ?

How does one apply Historical Simulation to this ?

TIA.

Apply the shocks to the market data, revalue and use the resultant vector of P&Ls? Or am I missing something?

How to use the resultant vector of P&L's ? calculate a returns series and take 99 percentile ??

I think market shocks would be applied for stress testing....if i am not mistaken..

As Rmax said - take today's market, prepare as much scenarios based on historical changes in the market, calculate changes in fair value and get the required percentile. The real question is how do you calculate historical scenarios. Which type of changes do you apply on rates, spreads and fx. Absolute or relative? Historical simulation for VaR is said to be free of distributional assumptions, which is a wrong statement, because by the way you calculate historical scenarios you take the assumption on the nature of the stochastic process generating the movements.

The other thing is how do you model spread dynamics - typical changes are very small, even in the tail, but it has very strong autocorrelation and explodes in crisis, so there is a very strong skew and heavy tail, which you will not capture in 250 scenarios.

You are right, it does look similar to stress testing, but level, magnitude, origin and purpose of the scenarios is different.

xccy basis swap levels can hit 0 sometimes so believe need to use absolute changes for this particular riskfactor