I have a portfolio of interest rate derivatives, say plain vanilla swaps, caps and floors, and I have an OGarch(1,1) covariance matrix for 13 points of the euro interest rate curve. Now I want to compute 1day and 10days VaR and CVaR (or Expected Shortfall) of my portfolio.
I'm thinking about using historical bootstrapping technique to generate the P&L distribution for the portfolio returns, this should not create problem for 1day horizon and I will check if it creates problems with 10days horizon, since bootstrapping destroy any mean reversion property. My data (from which I will bootstrap) are historical returns of interest ratesI have an interest rate curve composed by 13 points, starting from overnight rate up to 25year rate. Thus I generate scenarios for interest rates, then in each scenario I price my portfolio of interest rate derivatives and calculate the difference between each simulated portfolio value and the original portfolio value to have a P&L distribution of portfolio returns from which to calculate VaR and CVaR. My questions are:
a. What is the input volatility to be used in each scenario to price caps and floors? Should I simulate the volatility in each scenario or should I keep constant for each scenario my initial estimate of the covariance matrix of interest rates? What do practitioners do in practice? Does anyone have any reference on this?
b. Suppose I have in my portfolio some exotic that must be priced by simulation. Does this imply that in each scenario I must simulate the price of the exotic? That is, suppose I price an exotic on a Hull White tree; is it correct that when I calculate VaR with historical simulation, in each scenario I have to build the tree to price the exotic?
c. Should all these issues suggest me to assume that returns on interest rates are normally distributed and then use parametric VaR and CVaR? Basically, does anyone know with which model I am less wrong?
Thank you very much!
Gaia
VaR for interest rate option portfolios
Forum rules
DISCLAIMER: We do not warrant or represent that this forum or its content is free of viruses, worms or other code that might be contaminating or destructive. We cannot guarantee that documents or files downloaded from the Site will be free from viruses and we do not accept any responsibility for any damage or loss caused by any virus. Accordingly, for your own protection, you must use viruschecking software when using the forum. You must not post or provide to us via the forum, any document or file which you believe may contain a virus. You must virus check any document or file which you intend to post or provide to us via the forum. You must ensure that any document or file you intend to post to the forum does not contravene any applicable laws or contravene any person's legal rights. We do not accept any responsibility for any damage or loss you may suffer.
DISCLAIMER: We do not warrant or represent that this forum or its content is free of viruses, worms or other code that might be contaminating or destructive. We cannot guarantee that documents or files downloaded from the Site will be free from viruses and we do not accept any responsibility for any damage or loss caused by any virus. Accordingly, for your own protection, you must use viruschecking software when using the forum. You must not post or provide to us via the forum, any document or file which you believe may contain a virus. You must virus check any document or file which you intend to post or provide to us via the forum. You must ensure that any document or file you intend to post to the forum does not contravene any applicable laws or contravene any person's legal rights. We do not accept any responsibility for any damage or loss you may suffer.

 Posts: 815
 Joined: Sun Sep 28, 2008 10:30 pm
Re: VaR for interest rate option portfolios
Hi Gaia
Lots of questions, so my answers are repeated after them for ease of reading this reply.
Gaia: .... I will check if it creates problems with 10days horizon, since bootstrapping destroy any mean reversion property.
COA: Note that using the GARCH model in the bootstrap, as in the filtered historical simulation method described in Section IV.3.3.4 will not destroy the meanreversion property. I attach an extract from Section IV.3.3 (Title of Section: Improving the Accuracy of Historical VaR) and an excel example that does filtered historical simulation.
Gaia: My data (from which I will bootstrap) are historical returns of interest ratesI have an interest rate curve composed by 13 points, starting from overnight rate up to 25year rate. Thus I generate scenarios for interest rates, then in each scenario I price my portfolio of interest rate derivatives and calculate the difference between each simulated portfolio value and the original portfolio value to have a P&L distribution of portfolio returns from which to calculate VaR and CVaR. My questions are:
a. What is the input volatility to be used in each scenario to price caps and floors? Should I simulate the volatility in each scenario or should I keep constant for each scenario my initial estimate of the covariance matrix of interest rates? What do practitioners do in practice? Does anyone have any reference on this?
COA: You need historical data on cap IMPLIED volatilities as well as on interest rates. The OGARCH matrix is used only for the simulation of interest rates (with volatility and correlation clustering). Indeed it is generally true, for any option VaR, that you need data on two types of risk factors – both the underlying price and the implied volatility. All this is explained (it takes 60 pages!) in Chapter IV.5.
Gaia: b. Suppose I have in my portfolio some exotic that must be priced by simulation. Does this imply that in each scenario I must simulate the price of the exotic? That is, suppose I price an exotic on a Hull White tree; is it correct that when I calculate VaR with historical simulation, in each scenario I have to build the tree to price the exotic?
COA: Yes. Although because of the difficulty and the time this takes, we often use a Taylor series to map a whole portfolio using its Greeks rather than fullrevaluation. See Section IV.5.5.7. This is quicker and easier than full revaluation of every product, but it is open to serious model risk. See Chapter IV.6.
Gaia: c. Should all these issues suggest me to assume that returns on interest rates are normally distributed and then use parametric VaR and CVaR? Basically, does anyone know with which model I am less wrong?
COA: Analytic VaR models for options portfolios based on normally distributed factors are highly inaccurate. Some methods are outlined, and their limitations discussed in Section IV.5.3. The question of the model risk inherent in each different approach is discussed in detail in Chapter IV.6, and throughout the empirical examples of Vol IV I highlight this very important issue.
Hope this helps!
Carol
Lots of questions, so my answers are repeated after them for ease of reading this reply.
Gaia: .... I will check if it creates problems with 10days horizon, since bootstrapping destroy any mean reversion property.
COA: Note that using the GARCH model in the bootstrap, as in the filtered historical simulation method described in Section IV.3.3.4 will not destroy the meanreversion property. I attach an extract from Section IV.3.3 (Title of Section: Improving the Accuracy of Historical VaR) and an excel example that does filtered historical simulation.
Gaia: My data (from which I will bootstrap) are historical returns of interest ratesI have an interest rate curve composed by 13 points, starting from overnight rate up to 25year rate. Thus I generate scenarios for interest rates, then in each scenario I price my portfolio of interest rate derivatives and calculate the difference between each simulated portfolio value and the original portfolio value to have a P&L distribution of portfolio returns from which to calculate VaR and CVaR. My questions are:
a. What is the input volatility to be used in each scenario to price caps and floors? Should I simulate the volatility in each scenario or should I keep constant for each scenario my initial estimate of the covariance matrix of interest rates? What do practitioners do in practice? Does anyone have any reference on this?
COA: You need historical data on cap IMPLIED volatilities as well as on interest rates. The OGARCH matrix is used only for the simulation of interest rates (with volatility and correlation clustering). Indeed it is generally true, for any option VaR, that you need data on two types of risk factors – both the underlying price and the implied volatility. All this is explained (it takes 60 pages!) in Chapter IV.5.
Gaia: b. Suppose I have in my portfolio some exotic that must be priced by simulation. Does this imply that in each scenario I must simulate the price of the exotic? That is, suppose I price an exotic on a Hull White tree; is it correct that when I calculate VaR with historical simulation, in each scenario I have to build the tree to price the exotic?
COA: Yes. Although because of the difficulty and the time this takes, we often use a Taylor series to map a whole portfolio using its Greeks rather than fullrevaluation. See Section IV.5.5.7. This is quicker and easier than full revaluation of every product, but it is open to serious model risk. See Chapter IV.6.
Gaia: c. Should all these issues suggest me to assume that returns on interest rates are normally distributed and then use parametric VaR and CVaR? Basically, does anyone know with which model I am less wrong?
COA: Analytic VaR models for options portfolios based on normally distributed factors are highly inaccurate. Some methods are outlined, and their limitations discussed in Section IV.5.3. The question of the model risk inherent in each different approach is discussed in detail in Chapter IV.6, and throughout the empirical examples of Vol IV I highlight this very important issue.
Hope this helps!
Carol
Re: VaR for interest rate option portfolios
Carol,
thank you very much, your answers are always so clear and useful!
Gaia
thank you very much, your answers are always so clear and useful!
Gaia
Who is online
Users browsing this forum: No registered users and 1 guest