November 10th, 2012, 7:58 pm
I haven't gone through any such article but here are my suggestions:1. As we estimate VaR using different techniques(historical, Monte carlo etc), lets we begin with Monte Carlo simulation of CMO pricing.2. We know that major risks in MBS are interest rate/prepayment/default risk. So we can do MC simulations for these risk factors.3. Use any short interest rate model e.g. Hull white and predict short rates.4. Use a model for prepayment prediction. This is not easy and firms use many in house developed models.5. Using predicted prepayment we get predicted cash flows. We find OAS(Option adjusted spread) for our instrument and using appropriate discount rate(spread + OAS), we find PV of future cash flows. We take average price of MC simulated prices. This is CMO price for a given interest rate yield curve that we obtained from our short rate model.6. Depending upon short rate model, we find a distribution of MBS/CMO price of our instrument from above step.7. Now we can find VaR for required quantile(e.g. 95%)I hope this gives some basic idea. Again, I don't claim anything here to be perfect and this procedure comes from my general understanding.