Why using "Basis Point Volatility" is better than "Relative Rate Volatility"? Are they really useful? and How to use them properly?
'Cause in the world of interest rates, things are normal, rather than lognormal. Nobody cares about the percentage move in rates. Instead people care about the absolute number of basis points, which is why, in the world of rates, it's all about bp vol, aka normal vol, and not about Black-Scholes, percentage vol, aka lognormal vol.
Very interesting Martin. Are you saying that you would use a model assuming a normal distribution to trade options on a contract denominated in basis points, such as eurodollar futures?
You guys have no idea for what is basis point volatility talking about! I'd better save my time here!
OT, now you be talking loco... If I were to guess, you got the two terms out of Burghardt's Eurodollar book. Point is that there's actually three types of vol people in the world of rates talk about: lognormal, normalized, aka bpvol (lognormal * forward), and normal. Now normalized vol is actually a rough approximation to proper normal vol, which is, generally, the most fundamental basic concept that modern practitioners in the world of rates still refer to. to dmo: In practice, modern mkts conventions have gone beyond simple normal distributions wherever possible (e.g. vanilla rates products, such as Eurodollar options, swaptions, caps/floors etc). SABR (which is a very computationally tractable stoch vol model) and its various flavors is what's most commonly used nowadays. One exception is the bond futures mkt, where it's quite difficult to deal with the various options embedded in the actual underlying, so people would still, most commonly, rely on Black-Scholes.
I should mention that there's an important caveat that I really should have remembered. Historically, the IR vol mkt has been used to rates significantly above 0, where normal really works a lot better than lognormal. However, Japan has always been a very interesting exception that now provides important lessons for other mkts. The lesson is that as rates approach the zero bound, things behave more and more lognormally, as one would expect. Luckily, SABR (see my previous post) is actually able to parameterize this gradual normal -> lognormal transition, which means you can tweak the distribution to your liking. There are also methods to specifically address the zero bound by 'absorbing' the distribution, but I am now starting to wax all poetic, so I'll stop.
That all makes perfect sense. With rates this low, makes it hard to ignore the nearby limit of zero. BTW, when I was in T-bond options in the eighties, the skew was such that equidistant otm puts and calls traded at about the same price - as if they were implying a normal distribution. At one point I decided to look into running sheets with normal distribution, and called Myron Scholes to ask him how to modify his model so it would generate normal-distribution prices. He was HIGHLY incensed at my inference that his model could be improved, and not cooperative in the least!
Hi Martin, what if your pnl is calculated using lognormal % movement? I'm thinking more from a strategy angle, say a swaption RV trade looks very good on a chart on a normalised basis but not so good in lognormal. How can you cater for this?
You should have presented him first a 12 yr malted Whiskey ! Thereafter it´s simply easier talk to him !