Page 1 of 1

SABR/LMM for dissertation?

Posted: February 4th, 2011, 3:49 pm
by Mebiles
Hello everyone,I am new in the forum so I am not sure if this is a valid post but anyways,I am currently doing MSc in Financial Engineering and I want to do something about interest rates for my dissertation. I have been searching the literature and I also asked few people and I have decided to do something with SABR/Libor market model.The thing is that I don't know much about the industry and from what I have read there are a lot of people here who do.Does anybody have anything to say about SABR/LMM as a dissertation topic ( probably calibration and pricing of interest rate derivatives ) ?Thanks.

SABR/LMM for dissertation?

Posted: February 4th, 2011, 11:23 pm
by Gmike2000
The problem with the classic SABR is that you can hardly control the far out strikes...their vol becomes too high. Also it may allow arbitrage for very low strikes. You will end up mispricing CMS products for example. In my opinion, a 2 factor short rate model with stoch vol can do the same job and is less complicated. Having said that, SABR with LMM is a good exercise for a student and suitable for a dissertation but you need to be mindful of its limitations.

SABR/LMM for dissertation?

Posted: February 5th, 2011, 6:08 pm
by Mebiles
QuoteOriginally posted by: Gmike2000The problem with the classic SABR is that you can hardly control the far out strikes...their vol becomes too high. Also it may allow arbitrage for very low strikes. You will end up mispricing CMS products for example. In my opinion, a 2 factor short rate model with stoch vol can do the same job and is less complicated. Having said that, SABR with LMM is a good exercise for a student and suitable for a dissertation but you need to be mindful of its limitations.Thanks a lot for the reply. I still have some time to decide so I will keep in mind. I am currently reading the review paper about interest rate modeling by Rebonato that I believe should be helpful. Can you suggest any material for model calibration? Because all of the material I found are model specific I need some kind of a generic explanation.

SABR/LMM for dissertation?

Posted: February 7th, 2011, 12:13 pm
by MaxCohen
The vladimir piterbarg books have all you need to know about fixed income.They are very expensive but worth the investment if you plan to work in fixed income.They are not as pleasent to read as wilmott, or hull, but i think will be and invaluable reference if you are doing your dissertation on fixed income.

SABR/LMM for dissertation?

Posted: February 7th, 2011, 12:36 pm
by Hansi
QuoteOriginally posted by: MebilesCan you suggest any material for model calibration? book covers the implementation and calibration of the model but doesn't include any code. It's written in a pretty clear style and is easy to follow but I think it would have been more useful with code.Here is a comment on the model from Piterbarg regarding the LMM extension he and Andersen offer in their book vs. the SABR/LMM model:QuoteOriginally posted by: Hansi@Vladimir: In chapter 14 there is a mention that the chapter provides a series of extension to the LMM in order to better capture observed volatility smiles. Any thoughts on the differences of your suggested model in comparison with the SABR/LMM promoted by Rebonato, McKay and White given that I have not had time to read through the chapter? (if you have ever looked into it that is)QuoteOriginally posted by: piterbargHiThank you for your comments. I have added a pdf of common notations to the downloads area of the book website ( here), hopefully it will be useful to you and othersas for my opinion of LMM/SABR, do not really look to start a religious war here, but my own view of LMM/SABR is rather negative as I see numerous drawbacks, e.g.1. Having a separate stochastic vol per Libor rate seems to imply that simulated shapes of yield curves would be quite weird (i.e. jagged and non-smooth)2. SABR formula does not match SABR Monte Carlo to any reasonably level of accuracy ; why not a big problem for interpolating smiles for vanillas, becomes a huge problem if you are really trying to calibrate an LMM model that is MC-based3. The accuracy of approximation formulas for swaptions looks like 5-7 years behind the corresponding formulas for LMM with (Heston) stoch vol (and this is not even considering the MC issue)If anybody disagrees feel free to start a new thread about it :-) For the record we spend precisely zero time on LMM/SABR in our book and if this is what you fancy, you should _NOT_ buy our bookSpeaking of buying -- Wilmott bookshop always has a good stash on hand, so they probably have the fastest turnaround time in the UK. Amazon.co.uk lists about 7-11 days delay. Unlike in the UK, Printing on demand seems much faster in the US where Amazon.com is doing a brisk trade and can turn around your order within a couple of days, judging by people's comments. Some initial reviews are also starting to appear on Amazon.com Vladimir

SABR/LMM for dissertation?

Posted: February 8th, 2011, 2:51 pm
by Mebiles
Thank you very much for the replies they are quite helpful. About the code, I have some coding experience I think I can handle it. However about the model, I now have some question marks because I don't want to end up having done my dissertation about a poor model. Anyway I wait my professor's reply.Thanks again.Bilgin

SABR/LMM for dissertation?

Posted: February 8th, 2011, 3:21 pm
by Hansi
Well I agree with Gmike2000 that the SABR/LMM model is a good exercise for a dissertation and providing a critique of it's limitations can be a great topic and if you do it well will lead to a good grade. It will also give you exposure to the basic SABR and LMM models independent of each other as you'll need to cover those before jumping into the fused model.My main advice for a MSc dissertation is: "It's going to take much longer than you think"

SABR/LMM for dissertation?

Posted: February 8th, 2011, 10:15 pm
by Mebiles
QuoteOriginally posted by: HansiWell I agree with Gmike2000 that the SABR/LMM model is a good exercise for a dissertation and providing a critique of it's limitations can be a great topic and if you do it well will lead to a good grade. It will also give you exposure to the basic SABR and LMM models independent of each other as you'll need to cover those before jumping into the fused model.What I had in mind was to compare SABR/LMM to some another model in terms of calibration and pricing performance. But then I thought that' might be too much both in terms of effort and time, however a critique sounds great.

SABR/LMM for dissertation?

Posted: February 8th, 2011, 10:48 pm
by Hansi
QuoteOriginally posted by: MebilesBut then I thought that' might be too much both in terms of effort and time, however a critique sounds great.It's feasible but I think there are sure to be time issues and it might reflect in a less meaningful paper given writing length and depth concerns. If you decide to go with the model I recommend no directly attacking the model but rather introduce it and provide a detailed analysis of it's inner workings in each possible use and then identify the cases where it's may fail along with the reasons for it and solutions and alternatives.