This document discusses approaches for calculating value at risk (VAR) for derivative positions. It outlines the delta normal and full revaluation methods. The delta normal method approximates VAR using deltas and gammas and is suitable for linear derivatives. The full revaluation method uses Monte Carlo simulation to regularly revalue positions and is needed for non-linear derivatives and those with cross-partial effects. The document provides examples of calculating VAR for an option position and bond position using both approaches.
This document discusses approaches for calculating value at risk (VAR) for derivative positions. It outlines the delta normal and full revaluation methods. The delta normal method approximates VAR using deltas and gammas and is suitable for linear derivatives. The full revaluation method uses Monte Carlo simulation to regularly revalue positions and is needed for non-linear derivatives and those with cross-partial effects. The document provides examples of calculating VAR for an option position and bond position using both approaches.
This document discusses approaches for calculating value at risk (VAR) for derivative positions. It outlines the delta normal and full revaluation methods. The delta normal method approximates VAR using deltas and gammas and is suitable for linear derivatives. The full revaluation method uses Monte Carlo simulation to regularly revalue positions and is needed for non-linear derivatives and those with cross-partial effects. The document provides examples of calculating VAR for an option position and bond position using both approaches.
VAR of a VAR= Delta Approach± Gamma Effect Got to specify a pricing function (like Black Scholes Model) and derivative= As per Tailor’s approximation, then use SMS( Structured Montecarlo Simulation) to regularly 2 Delta* VAR of Risk ∆y=f ‘(x) ∆x +1/2f ‘’(x)( ∆x) revalue the position. Factor To be used in the case of Non-Linear misbehaved series like To be used in the Gamma Effect option-embedded bonds i.e. callable bonds,puttable bonds & rd case of linear For well-behaved non-linear derivatives,the 3 & mortgagebacked securities.Also to be used when we use derivatives like higher order terms are negligible. cross-partial effects. fwd,futures & So,gamma effect= Suppose,assume that stock-prices are log-normally distributed. 2 swaps. 1/2*Gamma*(VAR of risk factor) St+1=Steµ+zσ ( where µ=drift & σ is shock) Delta- Where z=Random draw from the standard normal distribution Normal To be used in the case of non-linear derivatives & time steps = daily for 100 days. VAR of a which are well-behaved like options, non-option Now calculate stock price using the above equation for 1 long- embedded bonds. run(i.e. 100 days)& the computer is made to carry out 10,000 position is For C+,Gamma effect is in our favour.So, runs. always VAR=VAR by Delta Approach-Gamma S overstated Effect. . For C-, Gamma effect is against us.So, VAR=VAR by Delta Approach+Gamma 22.75 Effect. For P+,Gamma effect is in our favour.So, 20.19 VAR=VAR by Delta Approach-Gamma 18.22 Effect. For P-,Gamma effect is against us.So, Days VAR=VAR by Delta Approach+Gamma Now, out of 10,000 terminal stock prices,arrange them from worst to Effect. best.Thus, the 99% VAR=Loss corresponding to the 100 th th The answers provided by this approach are NOT observation.Suppose 100 obs=18.64 100% accurate. ∴ = 20 − 18.64 = 1.36 But, since call & put are egs of well-behaved non- NOTE: For a long straddle position i.e. P+ & C+ at the same E,the risk linear derivatives,it gives a reasonable of this strategy is Share Price not changing too much.So,we cannot approximation of the true ans that can be calculate VAR using Delta-Normal Approach.Instead use Full manufactured by the full-revaluation approach. Revaluation Approach using SMC i.e. assume S follows a certain µ+zσ distribution say Log Normal.So, St+1=Ste . Using random draws of CALCULATING VAR OF A BOND POSITION Z,we have various possible (St+1)’s.Value the straddle at each of the Delta –Normal Method ( at 99% confidence level) possible Share Prices.Now, arrange straddle values from worst to best =z* yield volatility*Modified Duration* Price & then slice away the required quartile to get VAR. (where Modified Duration= Duration/(1+r/m) Suppose ytm is compounded semi-annually,Higher the frequency of Price compounding,higher the volatility and higher the VAR. Full Full Revaluation Method Revaluation At a 99% confidence level,yield can change by z*yield volatility. So,new yield=old yield+( z*yield volatility) Calculate old price(by using old yield) & new price(using new yield). 637.63 Delta Normal So, VAR=Old Price-New Price Method
Delta-Normal gives a higher answer for a long position,since actual VAR
enjoys gamma effect(enjoys convexity which means lower risk/VAR). Delta-Normal gives a lower answer for a short position,since actual VAR 5% 6.63% Yield suffers gamma effect(negative convexity i.e. higher risk/VAR)