One for the quants...

[jamescummins]

Hi,

I am looking at trying to put together model that will value the following portfolio:

Well, thinking about it, that is badly worded, but let me try and explain what I want to do...

you have a portfolio of products, that show some correlation -say, nickel and steel, or oil and gas, or whatever (works better with financial products). What I am trying to do is build into my valuation model the correlations to get a better worst case picture.

For example, say I model a 10% fall in nickel, I want to reflect the fact that as a result of the correlation, my steel positions will likely suffer a fall in value. These correlations are hard coded at the moment, and I want to be using realtime correlations. I can get the correlation easy enough themselves, but I am wondering if I am missing something fundamental, because it is not looking right!! All of a sudden the gas desk is basically running an enormous arbitrage position which I am assuming means either something very odd happened in the market, or more likely, I have completely smashed up my valutions

Has anyone come across this type of risk measurement before? It is inherently simple, but I am getting really bogged down trying to make it work, and it is stressing me out. Hence i keep posting on here rather than trying to fix it I guess!!

Anyway, any suggestions greatly appreciated

Thanks,
James

 

[gooseman]

futures, options? check out rebonato, the fox and the perfect hedge. no good for programming but the theory behind it is very good.

 

[jamescummins]

It is a whole portfolio valuation model - so it is across all sorts, but mainly swaps, forwards, and options. The Valuations themselves are fine, all straighforward. But essentially, if you calculate a margin for each position and add them up, it will come to a much higher figure than if you were to correlate all the indiviual positions to each other.

I can do this using the static correlations, but this breaks easily, and does not offer a reliable mtm value if that makes sense. This is not going to be used by the risk department as the 'official' risk management tool, it is for the desk that would like to have an idea of overall what their risk profile is at any given moment.

There are a whole load of flaws, in that a lot of the trading they do is correlation based, so using a model that incorporates only a fixed rate would not really do the job, although it is better than nothing.

Sorry for being stupid, but is that a book you are referring to above? I have used most of the standard texts in the past, and the initial model was hard enough to build, but thanks, I will have a look into this.

One of the main problems is that the more I try to model the market accurately, to produce a live risk model, the more assumptions need to be made in the model. That is inherent to it, and I cant see any way around that. The thing is, I am not sure if what I want to do can actually reasonably be done, or if the numbers you get at the end of the day, are just going to be spurious and not telling you a lot...

Thanks for your help, any other comments on how I could go about this, or where I am going wrong greatly appreciated.

James