Serving the Quantitative Finance Community

 
User avatar
pentagram
Topic Author
Posts: 6
Joined: November 9th, 2010, 11:40 am

Rates dynamics models used by banks for pricing and risk management

February 13th, 2021, 10:11 am

I'm studying interest rates theory and in the rates world are quite a few diffusion models, Hull-White 1factor, Hull-White 2factor, Ho-Lee, Vasicek, CIR, Derman-Toy, Black-Karasinski , Chen, etc etc. Other asset classes don't have nearly as many models.

I) which of these are used by banks in production to price and risk manage (ie Greeks calc) rates exotics products ?

II) do banks tend to use short rate models for pricing and risk management or HJM models ?

III) Also, in terms of calibrating these models, in the rates world, which derivative instruments are treated as 'priced by the market' & are used to calibrate the diffusion equation parameters (is it ie swaptions, caps, floors?) which will in turn be used to price exotics. e.g. in equities, vanilla options are used to extract the vol surface that is used to price more complex instruments, there's no pricing for vanillas, they're the instrument that's used to extract the volatility surface.
 
User avatar
bearish
Posts: 5180
Joined: February 3rd, 2011, 2:19 pm

Re: Rates dynamics models used by banks for pricing and risk management

February 13th, 2021, 6:29 pm

Please don’t take this the wrong way, but your questions seem straight out of the early 90’s. We (well, some us) once thought there was some magic to the HJM formalism, but we (both academics and practitioners working in the area) had largely gotten over that by ca 1995. Of course, reasonable exceptions are made for people with something to sell. None of the models you mention are likely to be in general use, although most of them undoubtedly make an appearance here and there. For a more modern perspective on rate modeling in practice, the three volume magnum opus by Andersen and Pieterbarg is worth perusing. I also don’t think you are right regarding the richness of models in other asset classes. Google “rough volatility” for a look at some pretty cool recent models primarily focused on equity derivatives.
 
User avatar
DavidJN
Posts: 241
Joined: July 14th, 2002, 3:00 am

Re: Rates dynamics models used by banks for pricing and risk management

February 13th, 2021, 7:06 pm

There is definitely more activity in modeling equity options than rates in recent years. 

The presence of more substitutes in the rates world (simple examples - a bond can be viewed as a whole instrument or a collection of strip coupons plus the residual, and similar decompositions for caps/caplets and swaps/FRAs) seems to afford more stringent no-arbitrage conditions for rates products than in the equity world. Would you agree with that sentiment, bearish?  
 
User avatar
Cuchulainn
Posts: 20205
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Rates dynamics models used by banks for pricing and risk management

February 13th, 2021, 7:10 pm

Here's some new research on rough Heston models with Neural Networks.
The level of maths needed is a quantum jump away from normal stuff. And ANNs are no guarantee as being better than 'traditional' methods. It's always good to measure and benchmark and not believe any results until you can reproduce them yourself. Which what was don in this thesis.
https://www.datasim.nl/blogs/29/msc-the ... nance-2020
 
User avatar
bearish
Posts: 5180
Joined: February 3rd, 2011, 2:19 pm

Re: Rates dynamics models used by banks for pricing and risk management

February 13th, 2021, 9:25 pm

There is definitely more activity in modeling equity options than rates in recent years. 

The presence of more substitutes in the rates world (simple examples - a bond can be viewed as a whole instrument or a collection of strip coupons plus the residual, and similar decompositions for caps/caplets and swaps/FRAs) seems to afford more stringent no-arbitrage conditions for rates products than in the equity world. Would you agree with that sentiment, bearish?  

I think there is something to that. There is also the fact that the first order object of interest (the yield curve, however you choose to represent it) is high dimensional with a strong and moderately stable low dimensional factor structure. Then there is the annoying (yet perfectly reasonable) tendency for markets to be made and quoted in terms of yields and spreads, rather than honest prices, which means that you have some convexity issues to sort out even before addressing those caused by particular pay-off functions. And, of course, the sort of universal act of discounting future cash flows invariably gets tangled up in the modeling of said cash flows in this context, whereas you can often abstract that out of the problem in many other asset classes.
 
User avatar
pentagram
Topic Author
Posts: 6
Joined: November 9th, 2010, 11:40 am

Re: Rates dynamics models used by banks for pricing and risk management

February 14th, 2021, 9:50 am

Please don’t take this the wrong way, but your questions seem straight out of the early 90’s. We (well, some us) once thought there was some magic to the HJM formalism, but we (both academics and practitioners working in the area) had largely gotten over that by ca 1995. Of course, reasonable exceptions are made for people with something to sell. None of the models you mention are likely to be in general use, although most of them undoubtedly make an appearance here and there. For a more modern perspective on rate modeling in practice, the three volume magnum opus by Andersen and Pieterbarg is worth perusing. I also don’t think you are right regarding the richness of models in other asset classes. Google “rough volatility” for a look at some pretty cool recent models primarily focused on equity derivatives.
 In equities most banks price everything but forward starting structures with local vol (for barriers sometimes stochastic local vol is used). For forward starting constant skew models are commonly used (in some cases bergomi is used as well). Clearly there is always an ongoing debate as to whether this is best general use, if there exist better models etc but in practice the models that market participants have chosen per product family is more or less fixed.

 In the rates world isn't there an equivalent set of 2-3 models that are used in production by nearly all banks ? if so, which model per product family is used in production?