Interest Rate Models: Theory and Practice D. Brigo, F. Mercurio

Hazard · Post by **Hazard** » Wed Oct 13, 2004 2:00 pm

Best book EVER!

asd · Post by **asd** » Wed Oct 13, 2004 9:38 pm

I agree this is best not only in contents, but has a high (No. of pages)/(Book Cost) ratio...... Applause

I had heard that that Plesser's book is better than Brigo for LMM implementation, but found that is is very thin for almost the same price

Dariusz · Post by **Dariusz** » Fri Oct 15, 2004 5:41 pm

LT: A while ago, there was a rumour that a 2nd edition was in the making. Apparently that project has been abandoned.

LT, I've got a good news for you, it will appear in the next year.

karabouchi! · Post by **karabouchi!** » Tue May 13, 2008 6:17 am

It's now 2008 and the second edition has been out for over a year now. I'm going to pick it up on wednesday and dig into it. Excited...

morphic.poly · Post by **morphic.poly** » Wed May 14, 2008 4:01 am

Hi,

since you bring this up again, here are a few words. It's a good book, but I am always appalled by people calling it "the best". That's a crass overstatement, besides the fact that there isn't any need for "a best".

There are a couple of things that I really dislike about this book. I don't think it reflects very well applied interest rate modeling as it is going on in banks nowadays. It is far too much focused on the libor market model and squarely doesn't spend enough time on shortrate models. the treatment of the HJM model is slim and basically useless. No connection between HJM and shortrate models is spelt out. The chapter on change of measure is unusable, because so general that it will never be a useful toolkit - at least not to me.

The thing is that LMM isn't as important in practice as people say it is. For one thing it's a bitch in terms of risks, I mean really, and it's friggin slow. Besides there isn't a market for truly multifactor sensitive interest trades, I mean marginally yes, but other than that not really. So, from a quant perspective LMM is important, but who cares about what quants think anyway. From a business perspective, if you can stay away from it, good for you.

The thing is that the guys who wrote this book, I am really not sure that they are employed by a shop that has an exotics derivative franchise that is worth mentioning. If they had, then they would not have time to churn out two editions of this book and they probably wouldn't use the type of implementation of LMM they write about.

So in terms of interest modeling books there are quite a few which are positively more useful than BM.

mp

mingon · Post by **mingon** » Wed May 14, 2008 10:10 am

Would you mind throwing a few names ?

IM · Post by IM » Mon Jan 03, 2011 4:54 pm

Morphic

I'm thinking about doing my master's thesis on rates, and can't decide whether it should be on LMM or the short rate models. Could you tell me which books you were thinking about?

Thanks!

sv507 · Post by **sv507** » Mon Jan 03, 2011 9:25 pm

IM

I would recommend Piterbarg and Andersens book ( try and get your library to get it!).

Brigo and Mercurio's book is value for money, but as morphic.poly said, its not clear that they were trading with the models. To me its great for derivations of formulas, but not so good for developing an understanding of the models (eg why is the hull-white model sufficient to value bermudan swaptions?)

I would recommend looking over papers by andersen, andreasen and piterbarg.

Its hard to say whether to go for short rate models or lmm - how good at coding/how much time do you have...how much do you already know?

I would just advise you to try to gain an understanding of the models rather than just implementing.[ The problem is that I suspect you will be hard pressed to do more than just implement Hull-White within the time limits of your masters]

for LMM - the two problems are calibration of swaptions, and bermudan type payoffs (using Longstaff-schwartz). So you could eg look at a TARN payoff (see piterbarg RISK article) and identify the factor dependence...

Perhaps another question to look at is hedging performance of eg SABR for vanilla swaptions on real market data.

you might also want to look at Rebonato's books which develop more insight (but have fewer implementation details)

IM · Post by IM » Wed Jan 05, 2011 2:21 pm

SV

Thanks for the detailed response! My thesis is due this August, and I can work on it about 1 day a week till then (maybe more in the last few months), so not too much time, but should be reasonably sufficient.

I've just started flipping to Brigo Mercurio, and also grabbed a copy of Rebonato's SABR LMM book. I've heard good things about Andersen Piterbarg, so I'm going to check that one out as well. I'm also taking a course under Pelsser which uses his book, so hopefully the book (and he) will be helpful.

As far as implementation goes, I'm a pretty decent coder, but I don't know too much about numerical optimization just yet (that course is in the next block) so I think I'll have a real hassle calibrating my model. I'm also worried about getting good market data.

What's the best way to develop an understanding of the models? I usually tend to just implement one and then play around with the parameters and plot stuff. Do you have any suggestions or a reference for how to "analyze" these models?

Thanks!

sv507 · Post by **sv507** » Thu Jan 06, 2011 11:58 pm

Getting an understanding of the models is really hard!

Basically you need to choose a product and play around with it (as you say)

it helps to have two models to compare! eg investigating factor dependence of berm swaptions- changing number of factors...

However- a crucial point made repeatedly by andreasen/andersen/piterbarg is that you have to calibrate your models appropriately, so that they match to the relevant market data (and then do comparisons/analysis). eg for pricing a bermudan swaption you need to calibrate the so called coterminal swaptions (rather than eg caplets/doing global fit etc).

The other aspect is to look at hedging performance on real data and identify where does the product have (cross-)gamma. eg if you look at a tilt of the yield curve are there significant nonlinear effects?[ since a one factor model will not price these in] (for which you need to have real data/statistics to identify typical moves)

[the same 'analysis' can be done to assess the smile exposure]

I would personally go for a two factor PDE model (where you can investigate the factor dependence) and look at bermudan products straightforwardly, but try to use off the shelf components as much as poss (matlab etc) to do the finite differences/ optimisation.[ Arguably this is closer to the development that takes place within a quant group anyway]

[you can always choose components you do in C++ afterwards according to the time you have]